Embed Size (px)
Citation preview
Improving methods for using existing land cover databases
and classification methods
A literature review
Technical Report Series GO-10-2016
February 2016
Improving methods for using existing land cover databases
and classification methods
A literature review
Table of contents
Acronyms and Abbreviations……………………………………………………………………………. 4 Preface…………………………………………………………………………………………………………….. 6 Acknowledgments……………………………………………………………………………………………. 8 1. Introduction…………………………………………………………………………………………………. 9 2. Remote sensing application in support of agricultural statistics……………………. 11 3. Country approaches for crop area estimation using remote sensing:
an overview..................................................................................................... 18 4. Remote sensing application in support of agricultural statistics……………………. 26
4.1. Crop type classification and identification…………………………………………. 26 4.2. Crop condition (monitoring) assessment…………………………………………… 27 4.3. Crop area estimation…………………………………………………………………………. 29 4.4. Mapping soil characteristics………………………………………………………………. 30 4.5. Precision farming practices……………………………………………………………….. 30 5. Remote sensing techniques and area frames……………………………………………….. 31
5.1. The FAO LCCS approach: the contribution to the area frame……………. 35 5.2. Generating the land cover database…………………………………………………. 37 5.3. Two case studies in land cover database generation in support of area frames: Ethiopia and Pakistan…………………………………………………… 41 5.4. Mapping agricultural land cover features using remote sensing – derived products: some considerations……………………………………………. 43
6. New products.................................................................................................. 52 7. Gaps and expected developments and implementation……………………………….. 54 8. Conclusion and recommendations……………………………………………………………….. 57 Annex I…………………………………………………………………………………………………………….. 59 References……………………………………………………………………………………………………….. 61
4
Acronyms and Abbreviations
AEZ Agro – Ecological Zone AF Area Frame AFS Area Frame Sampling
AGRISTARS A joint program for Agriculture and Resources Inventory Surveys Through Aerospace Remote Sensing (NASA)
AGRIT Programma Statistico AGRIT, Ministero Politiche Agricole Ambientali e Forestali
ALOS Advanced Land Observing Satellite APhs Aerial Photographs ASABE American Society of Agricultural and Biological Engineers ASTER Advanced Spaceborne Thermal Emission and Reflection Radiometer AVHRR Advanced Very High Resolution Radiometer AWIFS Advanced Wide Fide Sensor CC Cubic Convolution CDL Cropland Data Layer CFF Crops for the Future CGIAR Consultative Group for International Agricultural Research CLC Corine Land Cover CORINE Coordination of Information on the Environment CSA Central Statistic Authority DMC Disaster Monitoring Constellation EMA Ethiopia Mapping Agency EO Earth Observation EOSAT Earth Observation Satellite ESA European Space Agency ETM Enhanced Thematic Mapper FAO Food and Agriculture Organization of the United Nations FEWSNET Famine Early Warning Systems Network GIS Geographic Information System GLCV Temporal Coverage for Fraction of Green Vegetation GNSS Global Navigation Satellite System GPS Global Positioning System HRV High Resolution Visible Sensor HRVIR High-Resolution Visible & Infrared Sensor IFAD International Fund for Agricultural Development IRRI International Rice Research Institute IRS Indian Remote Sensing ISO International Organization for Standardization LACIE Large Area Crop Inventory Experiment LANDSAT Land Remote – Sensing Satellite LCCS Land Cover Classification System LCML Land Cover Meta - Language LISS Low - Imaging Self Scanner LS Legumes Survey
5
LUCAS Land Use and Cover Area frame Survey LULC Land Use Land Cover MAP North Map Orientation MARS Monitoring Agricultural Resources, Ec MERIS Medium Resolution Imaging Spectrometer MLC Maximum Likelihood Classification MMU Minimum Mapping Unit MODIS Moderate Resolution Imaging Spectroradiometer MoA Ministry of Agriculture MSF Master Sampling Frame MSS Multispectral Scanner NASA National Aeronautics and Space Administration NAS National Agricultural Survey NDVI Normalized Difference Vegetation Index NDWI Normalized Difference Water Index NED National Elevation Data NIR Near Infrared NOAA National Oceanic and Atmospheric Administration OLI Operational Land Imager PSU Primary Sampling Unit RS Remote Sensing RVI Ratio Vegetation Index SAC-C Satélite de Aplicaciones Científicas-C SAR Synthetic Aperture Radar SAVI Soil Adjusted Vegetation Index SEEA System of Environmental Accounting SPOT Satellite pour l’Observation de la Terre SRTM Shuttle Radar Topography Mission SSU Secondary Sampling Unit SUPARCO National Space Agency of Pakistan SU Survey Unit TIRS Thermal Infrared Sensor TM Thematic Mapper TNOVI Transformed Normalized Difference Vegetation Index TVI Triangular Vegetation Index USA United States of America USDA United States Department of America USGS United States Geological Survey UTM Universal Transverse Mercator VGT Vegetation WGS World Geodetic System WRSI Water Requirement Satisfaction Index WSEAS World Scientific and Engineering Academy and Society XS Multispectral
6
Preface
This literature review presents the main findings of bibliographic research on
the application of remote sensing to agricultural and rural statistics. It was
carried out in support of the Global Strategy Action Plan, a programme that
aims to provide guidelines for national and international statistical authorities.
The research focused on remote sensing applications in agriculture – in
particular, on the aspects and activities related to land cover database generation
and classification. A list of the relevant literature and any other related
information has been created and may be consulted in the “References” section
at the end of this document.
Through this work, it was sought to understand how various authors have
exploited the available remote sensing products and technologies (GIS, GPS,
etc.) in the different phases of statistical analysis.
The sources of information were derived mainly from papers retrieved from
online catalogues and research engines. While this collection is not to be
considered complete, due to the wide scope of the topics analysed, the
fundamental approaches adopted in applying remote sensing technology in
agricultural statistics have been collected and presented.
The first part of the consultancy was devoted to research; in the second part, the
acquired documentation was analysed. Finally, general considerations on the
papers that presented the state-of-art application of remote sensing in
agriculture and operative experiences were drafted. The results of the research
are reported in the References section of this document.
The first part of this literature review presents the remote sensing products that
are or will be available, including information on their approximate costs. The
second part provides an overview of the various statistical approaches applied
around the world, emphasizing the applications of remote sensing products and
techniques. The third part examines remote sensing and its applications to
agricultural statistical analysis in further detail, especially in relation to land use
and land cover as a basis for statistics extraction, stratification and sample
allocation. In particular, attention is paid to the aspects linked to the generation
of an LCCS 3-style land cover database as a tool to support area frames.
Finally, the main gaps and recommendations identified from this preliminary
analysis are discussed.
7
In Annex I, the criteria adopted for researching and listing the main scientific
journals consulted are described.
8
Acknowledgments
The report was developed by John S. Latham and Ilaria Rosati, FAO Geospatial
Unit Deputy Director Office for Natural Resources, DDNS, DDN Division.
Valuable inputs and comments were provided by Elisabetta Carfagna (Full
Professor of Statistics, University of Bologna, Department of Statistical
Sciences); during the peer-review stage by Ray Chambers (Statistical Institute
of Australia), Jacques Delince’ (ESS) and Keita Naman.
The following FAO officers also provided contributions: Christophe Duhamel,
Flavio Bollinger, Angela Piersante, Carola Fabi.
The document was edited by Sarah Pasetto.
9
1
Introduction
The assessment of cultivated areas has long posed a challenge for the accurate
and timely monitoring of agriculture. In recent times, it has become possible to
support and facilitate the evaluation of the two main components of crop
production (area and yield) by means of innovative satellite remote sensing
technologies, which can provide timely, precise, objective, and accurate
information.
It is generally agreed that the use of remote sensing technology can
significantly improve the efficiency and accuracy of traditional agricultural
statistics and systems for monitoring agricultural conditions. The use and
exploitation of such technology are advocated and promoted as a powerful tool
in many phases of assessment. Several authors discuss and report the relevant
aspects and limitations (among many, see Carfagna, Gallego, Latham, etc.; see
the References for a more complete list).
The application of geospatial sciences has gained considerable credibility over
the past few decades. GIS, remote sensing and Global Positioning Systems
(GPS) are the most frequently applied tools. They have evolved rapidly to
address a wide range of scientific problems arising within the agricultural
monitoring field. In agricultural production and agronomy, remotely sensed
imagery is beneficial in several interrelated activities, such as:
• crop type identification;
• crop classification;
• crop condition assessment;
• crop yield estimation;
• mapping of soil characteristics;
• mapping of soil management practices; and
• compliance monitoring (farming practices).
Remote sensing products have been successfully applied in various phases of
area frame construction. Sampling allocation on the ground also requires the
use of aerial photos or images, as well as crop type discrimination and
monitoring.
10
Remote sensing is also a source of information from which it is possible to
generate analyses of the territory and of its features. An accurate land cover
database provides an updated baseline that can be useful in several phases of
agricultural statistics, especially in relation to stratification. In this regard, a
vector database enables the generation of high-quality digital outputs that can
improve the efficiency of sample allocation and reduce sample size, thereby
improving overall efficiency. Generally, a supervised automatic classification
should not be used as a direct tool to estimate the cover, including the area
covered by crops.
This literature review focuses on the possible applications of optical images to
support the development of improved agricultural statistics. Thermal Infrared
Images and Microwave imagery have more complex and specific procedures
and methods of applications and processing. Due to their specificity, therefore,
this publication will mention them only briefly, and a more detailed discussion
will take place in a separate module.
The main characteristics of the products and general principles of remote
sensing and its applications are described in several excellent books and
manuals (Lillesand et al., 2014; Chuvieco et al., 2009; Sabins, 1986) and lecture
notes and tutorials available online. There are also many websites with
resources on remote sensing; a selection of the most interesting of these is set
out in Table 1 below.
Table 1 – Remote sensing websites.
http://satellite.rsat.com/rsat/tutorial.html
http://auslig.gov.au/acres/referenc/abou_rs4.htm
http://landsat.gsfc.nasa.gov/education/tutorials.html
http://www.ccrs.nrcan.gc.ca
http://earthengine.google.org/#intro
11
2
Remote sensing products:
an overview
Satellite companies offer a wide range of products, which feature a variety of
characteristics.
Table 2 illustrates the main features of some of the major optical remote
sensing sensors (medium, high, and very high). These products, which are well-
known to remote sensing end users, are widely used in research and
applications.
Recently, a number of new products have become accessible free of charge, e.g.
Landsat 8 (USA-NASA) and Sentinel (EU; operated by ESA). Tables below 4
and 5 provide additional technical details.
Landsat 8 is already available to download, while Sentinel products will be
fully operative in 2016. Interestingly, they are both proposed as free-of-charge
products; this represents great potential, especially when the surveys and
analyses require large territories to be covered at national or regional level. In
addition, the repetitive acquisitions will provide ample data sets, which are
useful for monitoring and long-term analysis.
Sentinel is arranged in five missions that carry a range of different technologies,
such as radar and multi-spectral imaging instruments for monitoring the Earth.
Sentinel 1 and 2 are missions that are potentially of interest to agriculture; the
others are dedicated to sea-surface topography, sea- and land-surface
temperature, ocean colour and land colour (Sentinel 3) and atmospheric
monitoring (Sentinel 4-5).
12
Table 2 - Technical characteristics of selected satellite sensors
Sensor
Spectral Resolution Spatial Resolution
(m)
Re-visit time
(days)
Recommended max. working
scale (approx.) Approx. Cost
Channel Wavelength range (µm)
Low / medium
NOAA-AVHRR
Channel 1 0.58-0.68
(red)
1,100 x1,100 1 1:1.5 M 0.00015 US$/km2
Channel 2 0.725-1.1 (near IR)
Channel 3 3.55-3.93 (mid IR)
Channel 4 11.3-12.6 (ther. IR)
Channel 5 11.4-12.4 (ther. IR)
MODIS Channel 36 From 0.620
To 2.155 250-1,000 1 1:1.5 M free
MERIS Channel 15 From 0.39
To 1.04 300-1,200 3 1:1.5 M free
SPOT VGT
Channel 1 0.50-0.59 (green)
1,150 x1,150 1 1:1.5 M free
Channel 2 0.61-0.68
(red)
Channel 3 0.79-0.89 (near IR)
Channel 4 1.58-1.75 (sh. w. IR)
High
Landsat Channel 1 0.5-0.6 (green)
80 x 60
18
1:200,000 / 1:100,000
free
MSS
Channel 2 0.6-0.7 (red)
Channel 3 0.7-0.8
(near IR)
Channel 4 0.8-1.1
(near IR)
TM/ETM
Channel 1 0.45-0.52
(blue)
30 x 30 16
Channel 1 0.52-0.60 (green)
Channel 3 0.63-0.69
(red)
Channel 4 0.76-0.9 (near IR)
13
TM/ETM
Channel 5 1.55-1.75 (sh. w. IR)
30 x 30
16
1:200,000 / 1:100,000
free
Channel 6 10.4-12.5 (ther. IR)
120 x 120
Channel 7 2.08-2.35 (sh. w. IR)
30 x 30
ETM Pancro 0.52-0.9 15 x 15
Landsat 8 See Table 3
SPOT-XS
Band 1 0.50-0.59 (green)
20x20 5- 2.5
26
1:100,000
1:5,000
Band 2 0.61-0.68
(red)
Band 3 0.79-0.89 (near IR)
SPOT Pancro 0.51-0.73 10x10 1:50,000
ASTER
3 Bands VNIR
(near IR) 15
16 1:100,000 0.0152 US$/km2 6 Bands SWIR
(near IR) 30
5 Bands TIR
(ther. IR) 90
IRS Pancro 0.5-0.75 5.8 x 5.8
24
1:15,000 0.33 US$/km2
IRS
LISS
Band 2 0.52-0.59 (green)
23.5 x 23.5 1:100,000
0.114 US$/km2
Band 3 0.62-0.68 (red)
Band 4 0.77-0.86 (near IR)
Band 5 1.55-1.7 (near IR)
IRS WiFS
Band 3 0.62-0.68
(red)
188 x 188 5 1:500,000
Band 4 0.77-0.86 (near IR)
DMC
Band 1 0.52-0.60 (green)
22 3/1 1:100,000
Ortho-rectified US$ 592 per image US$ 0.12 per km2
Band 2 0.63-0.69 (red)
Band 3 0.77-0.90 (near-IR)
14
DMC Nigeria Sat-2 2015
Pancro 0.52-0.898 2.5
Now on demand
1:100,000 1:5,000
Ortho-rectified US$ 592 per image US$ 0.12 per km2
Band 1 0.448-0.517
(blue)
5
Band 2 0.527-0.606
(green)
Band 3 0.63-0.691
(red)
Band 4 0.776-0.898
(near-IR)
DMC3 2015
Pancro 1
R,G,B,NIR 4
ALOS
Band 1 0.42-0.50
(blue)
10 46 1:100,000 US$ 0.1004 /km2
Band 2 0.52-0.60 (green)
Band 3 0.61-0.69
(red)
Band 4 0.76-0.89 (near-IR)
Very High
Pleiades
Pancro 0.47-0.83 0.7
1 day (43° off-nadir); 4 days (30° off-nadir); 5 days (20° off-nadir); 13 days (5° off-nadir)
Very detailed < 1:5,000
13 US$/km2
Band 1 (B0) 0.43-0.515
(blue)
2.8
Band 2 (B1) 0.50-0.62 (green)
Band 3 (B2) 0.59-0.71 (red)
Band 4 (B3) 0.74-0.94 (near IR)
RapidEye
Band 1 0.44-0.51
(blue)
5 5 Very detailed
< 1:5,000 1.28 US$/km2
Band 2 0.52-0.59 (green)
Band 3 0.63-0.685
(red)
Band 4 0.76-0.85 (near IR)
15
GeoEye-1
Pancro 0.40-0.80 0.5
21 Very detailed
< 1:5,000 25-40 US$/km2
Band 1 0.45-0.51
(blue)
2.0
Band 2 0.51-0.80 (green)
Band 3 0.655-0.69
(red)
Band 4 0.78-0.92 (near IR)
Quickbird
Pancro 0.45-0.90
0.6-2.4
1-3.5 days, depending on latitude (30° off-nadir)
Very detailed < 1:5,000
28-40 US$/km2
Band 1 0.45-0.52
(blue)
Band 2 0.50-0.60 (green)
Band 3 0.63-0.69
(red)
Band 4 0.76-0.90 (near IR)
IKONOS
Pancro 0.45-0.90
0.8-3.2 Approx.
3 days at 40° latitude
Very detailed < 1:5,000
20-35 US$/km2
Band 1 0.45-0.52
(blue)
Band 2 0.50-0.60 (green)
Band 3 0.63-0.69
(red)
Band 4 0.76-0.90 (near IR)
Sentinel See Table 4
16
Table 3 – Landsat 8 technical characteristics.
Landsat 8
Launched February 2013
Instruments on board
Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS)
Processing Level 1T - Terrain Corrected
Pixel size
OLI Multispectral bands: 30 m OLI Panchromatic band: 15 m TIRS Thermal bands: 100 meters (re-sampled to 30 m to match multispectral bands)
Spectral bands
Landsat OLI image data 9 spectral bands (band designations) with a spatial resolution of 30 m for Bands 1 through 7 and Band 9. Band 8 (panchromatic) is 15 m Landsat TIRS image data will consist of two thermal bands with a spatial resolution of 100 meters (re-sampled to 30) for Bands 10 and 11. The TIRS data will be packaged with the OLI for data distribution.
Scene size 170 km north-south by 183 km east-west
Data characteristics
GeoTIFF data format Cubic Convolution (CC) re-sampling North Up (MAP) orientation Universal Transverse Mercator (UTM) map projection (Polar Stereographic for Antarctica) World Geodetic System (WGS) 84 datum 12 m circular error, 90% confidence global accuracy for OLI 41 m circular error, 90% confidence global accuracy for TIRS 16-bit pixel values
Data Delivery HTTP download within 24 hours of acquisition
Table 4a – Sentinel-1: technical characteristics.
Sentinel-1
Launched 2013
Instruments on board
Sentinel-1 is a polar-orbiting all-weather, day-and-night radar imaging C-band imaging radar mission
Processing
Pixel size 5 × 20
Spectral bands Expected to provide coverage over Europe, Canada and main shipping routes in 1–3 days, regardless of weather conditions
Scene size Swath width of 250 km
Data Characteristics
Data Delivery Radar data will be delivered within an hour of acquisition
17
Table 4b – Sentinel-2: technical characteristics.
Sentinel-2
Launched 2014
Instruments on board
Sentinel-2 is a polar-orbiting, multispectral high-resolution imaging mission
Processing
Level 1 image processing includes:
- a) Radiometric corrections: straylight/crosstalk correction and defective pixels exclusion, de-noising, de-convolution, relative and absolute calibration - b) Geometric corrections: co-registration inter-bands and inter-detectors, ortho-rectification.
Level 2 image processing includes:
- a) Cloud screening - b) Atmospheric corrections: including thin cirrus, slope and adjacency effects correction - c) Geophysical variables retrieval algorithms: e.g. fAPAR, leaf chlorophyll content, leaf area index, land cover classification.
Level 3 provides spatio-temporal synthesis
Simulation of cloud corrections within a Level 2 image
Pixel size < 1 ha MMU (Minimum Mapping Unit) fully achievable with 10 m 4 bands at 10 m, 6 bands at 20 m and 3 bands at 60 m spatial resolution (the latter is dedicated to atmospheric corrections and cloud screening)
Spectral bands Optical payload with visible, near infrared and shortwave infrared sensors comprising 13 spectral bands). 0.4-2.4 µm (VNIR + SWIR)
Scene size Swath width of 290 km
Data characteristics
Revisit time of five days at the equator (under cloud-free conditions) and 2–3 days at mid-latitudes
Info https://earth.esa.int/web/sentinel/user-guides/sentinel-2-msi/resolutions/spatial
A number of smaller start-ups are poised to enter the market. Generally, their
approach is to develop a larger number of smaller satellites that provide
imagery more frequently – thereby reducing the “revisit time” – and to deliver
images of the same location every day or multiple times per day. Skybox,
Planet Labs, Dauria Aerospace (USA) – in joint venture with Elecnor DEIMOS
(E), operational in 2016 – and OmniEarth are some of the promising new
companies that will provide several new satellite products.
18
3
Country approaches for
crop area estimation using
remote sensing: an overview
Since the 1970s, the United States and other countries have taken the lead in
conducting acreage and yield estimations of major food crops (wheat, corn,
rice, etc.). These were implemented in the early 1970s and 1980s through the
LACIE (first) and AGRISTARS (later) programmes, and operationalized
nationally through a series of major projects. These now form a regularly
operating technology and monitoring system.
Remotely sensed satellite data for crop area and yield estimation was used
extensively in the United States and the EU, although significant applications
have also been implemented in several developing countries (e.g. in Libya,
Latham, 1981-83; in Ethiopia, Wigton, et al., 2009).
In the first work on agricultural monitoring by means of remote sensing
measurements, the United States National Aeronautics and Space
Administration (NASA) and National Oceanic and Atmospheric Administration
(NOAA) completed a landmark experiment (Jerry et al., 1979) on measuring
the area, yield and production of the world’s major wheat-producing areas. The
program, titled Large Area Crop Inventory Experiment (LACIE), was
conducted from 1974 to 1978. After the LACIE program, the United States
launched AGRISTARS, Ag20/20 and other agricultural resources survey
projects.
In 1989, the EU launched the MARS program. This program, which
experimented with surveying crops in a large area with remote sensing
techniques, ultimately became a practical operating system. MARS is capable
of monitoring and forecasting crop information on three levels: community,
regional and national (Gallego, 1999). After the MARS program, the EU
launched the LUCAS program, which examined the land use situation based on
an area sampling frame. LUCAS was implemented on national, Member State
and international scales (Delincé, 2001; Gallego, 2004).
19
FAO has established a global early warning food information system to monitor
and forecast global crop yield. Canada uses the NOAA satellites for crop water
measurement, disaster warning, and yield forecasting in large areas. Russia uses
MODIS to estimate crop acreage in the south of Russia. South Africa estimates
crop area with the support of remote sensing data and unmanned aerial vehicles.
Brazil monitors crop acreage and yield in the south of Brazil using remote
sensing data.
In 2008, the Group on Earth Observation prepared a useful outline of the best
practices for Crop Area Estimation
(https://www.earthobservations.org/documents/cop/ag_gams/GEOSS%20best%
20practices%20area%20estimation%20final.pdf) adopted by several national
agencies. The main parameters analysed are listed below.
Scale of coverage
Input data
Spatial resolution
Sampling design
Image analysis
Accuracy evaluation
User acceptance
Costs/benefits
Software
Manual
Sources
Table 5 summarizes the most important features of the systems analysed,
integrated with case studies from a collection of different publications and
papers.
20
Table 5 – Agricultural statistical approaches in different countries.
Country Scale of
crop covered
Input data spatial
resolution
Sampling design
Image analysis
Argentina Regional / provincial
Landsat and Medium Resolution (SAC-C, 175 m pixel size) imagery
Stratified sampling based on digital cadastral maps and other ancillary Data
Digital analysis
Belgium &Poland
National Meris, SPOT & Landsat
Full area coverage will be input to Area Frame Sampling (AFS)
Arable land cover mask generation based on CORINE land cover (CLC) product. Decision rule-based classification using Temporal coverage for Fraction of Green Vegetation (GLCV).
Brazil
State level for rice, wheat and soya bean
NOAA/AVHRR, Landsat TM. Recent switched to (MODIS 250 m res.).
Full area coverage
NDVI image thresholding based on decision rule, on image differencing of NDVI images
Canada
National level, using major agricultural states. Potatoes as single crop.
Optical and SAR (VV-VH) data, future plan for polarimetric SAR data from Radarsat-2
Sampling of remote sensing data followed by regression estimator
Multi-temporal classification based on decision rule
Prince Edward Island (PEI),
Canada
Province level for Potato
Medium resolution, Earth Observation (EO) satellite data
Stratified sampling combined with a regression Estimator
China
National level to global level (26 countries, representing 80% grain production). Eight crops - winter and spring: wheat, early rice, semi-late rice and late rice, spring and summer maize and soya bean.
NOAA/AVHRR, SPOT-VGT, Landsat TM, ALOS and Radarsat presently switched to MODIS,
IRSP6/AWIFS，BJ-
1 and HJ-1a/b
Full area coverage with Remote Sensing data. Transect sampling for crop types.
Multi-temporal satellite data. Spectral profile based on NDVI.
21
Colombia
National level: National Agricultural Survey (NAS) and Legumes Survey (LS)
Not mentioned
ASF with coverage of 37,900,546 ha (169,587 SSUs in 31,588 PSUs). LS coverage of 10,874 ha (588 SUs).
Ethiopia, Niger and Zimbabwe
National-level, pilot area in Ethiopia. For crop and non-crop areas
Landsat ETM+ / AWiFS. Other inputs are samples of high-resolution images. The moderate-resolution images and other ancillary information include: SDRN-FAO processed SRTM 90 m digital elevation model; FEWSNET primary livelihood zones for Niger; agro-ecological zones defined by the Water Requirement Satisfaction Index (WRSI) for Zimbabwe; and the International Research Institute’s woody biomass map for Ethiopia as co-variables.
Sampling frame at 2 km grid and Sample area at 0.5 km.
Field survey/GT by photo-interpretation on Ikonos and Quickbird
Ethiopia National SPOT 5m From 2008, AF
Land cover using LCCS and object-based stratification methodology
EU-1: EU-MARS
1988-93
EU countries at regional levels
Landsat TM, SPOT
Stratified sampling, sample segment of 25-200 ha.
Supervised/unsupervised/decision rule mask(non-agriculture)/multi-stage classification
22
EU-2: Land Use/Cover Area frame
Survey (LUCAS)
EU 25
11 countries at regional level. Land use/cover area frame survey (LUCAS).
Landsat-TM images
Two-way sampling, 2-km grid for stratification. Each stratum is sampled at different rates based on the agricultural area sampled, 5 times to non-agric. area.
EU-3: MARS –
Rapid Area Estimate
Regional/national (EU) level. Inter-annual acreage changes. For main crops: winter cereals, summer crops, grassland, sunflower
SPOT-XS (20 m), Landsat-TM and IRS-1C data
Multi-stage stratification based on computer-assisted photo-interpretation and pixel clustering (unsupervised classification)
A fuzzy classification and a priori information was used where each cluster of pixels was linked to one or several land cover classes through a correspondence table. This was later used to compute estimates that provided the flexibility to combine the image classification with general information obtained from published data.
EU-4: EU_ TUSRAD
Regional-level (Tuscany). EU. Crops covered: common wheat, durum wheat, rapeseed, maize
SAR data Full area coverage
Multi-stage and multi-date using classification, using mask (non-agriculture areas, etc.); for multi-date SAR data, using decision rule-based classification.
Hungary
COUNTRY level, For 8 main crops including wheat, soya bean, colza
Landsat TM, SPOT, IRS-1C/1D/NOAA/ AVHRR, EOSAT
Full area coverage
Multi-temporal digital data analysis based on decision rule
India
Local-regional to national. Crops covered: wheat, rice, sugarcane, mustard, cotton, sorghum, groundnut,
potato.
IRS–/III/AWiFS 23/56 m. 24/5 days. Microwave. Radarsat SAR.
Stratified sampling with 10-15% sample
Supervised maximum likelihood at single date data for local- to state-level crop acreage estimation. Multi-stage classification using mask (non-agriculture areas); for multi-date AWiFS, using decision rule-based classification
Italy
Regional (20 regions ) or provincial level For major crops including wheat, soya bean, colza
Landsat TM
Two way Sampling: 1. Stratified area frame sampling, 500 m grid (each segment of 50 ha). 2. Regression estimator
23
Iran
Country and provincial level (rice, wheat & barley, and other annual crops)
Not mentioned
Stratified two-stage subsampling for crop survey
Libya
National & Regional (Area Gefera Plain, Jebel Achdar, 1981/82/83)
Landsat MSS/TM Aerial photos
Multi-year/Multi-seasonal. Stratified Random Sample Area Frame
Digital Image Analysis
Kazakhstan
Local level, pilot project. Cereals: winter wheat, spring barley Together
MODIS 250 m, IRS LISS-III 23 m
Full area coverage
Morocco Country SPOT 5-10m
AFS. Stratified sampling based on image interpretation
Visual interpretation using 10 strata (land cover classes)
Norway Country level for natural resources
APh for fieldwork.
Lucas linked to Eurostat. AF sampling at intersection of 18 x 18 km. PSU 1,500 x 600 m within 10 SSUs.
No detail on how the RS data are used.
Pakistan
District level (Rahim Yar Khan District) Crops covered: winter crops, summer crops, double cropping, and dominant pluri-annual and permanent crops.
SPOT VEGETATION data for stratification and SPOT4 and 5 HRV for digital analysis
Stratified sampling based on NDVI profiles at district level
Multi-stage and multi-date classification using mask (non-agricultural areas, etc.) for multi-date SPOT classification and
regression estimator.
Pakistan National SPOT 5m
Land cover using LCCS and object-based stratification methodology
Russia-1
ISPRA
County level leading to national level.
For winter crops and sunflower.
Multi-date weekly composited MODIS/SPOT –VGT, Landsat TM /ETM+, NOAA/AVHRR data
Full area coverage Decision rule-based classification
24
Russia-2
Ministry of Agriculture
Russian Federation
National level.
For winter crops and sunflower.
Terra/Aqua-MODIS (250 m) and SPOT 2&4 - HRV/HRVIR, NOAA-AVHRR, SPOT-Vegetation
and Landsat TM/ETM+
Full area coverage Decision rule-based classification
South Africa
Provincial leading to national.
Maize, sunflower, soya beans and sorghum are cultivated in summer. The second group of crops are grown during the winter and include wheat, barley, oats and canola.
Landsat 5, SPOT 2 & 4, Aphs
Stratified systemic random sampling
stratification based on cultivation intensities; GT
by aerial survey of sample area
Supervised based on maximum likelihood classifier combined with the parallel-piped functions.
Turkey
National level. Need to produce sub-regional and provincial level statistics.
For crop acreage
prediction studies and household surveys.
No specification, but extensive use of RS products for stratification baseline.
Small area estimation, household Surveys, area frame sampling, survey-based estimates,
synthetic and model-based estimates
Image interpretation (generic) and ground survey
USA-1: USDA
National level using major agricultural states of USA.
For winter and spring wheat, corn, soya bean and durum wheat, rice and cotton.
Landsat TM/ETM+ 30 m resolution from 2004 onwards IRS-P6 AWiFS data 56 m resolution, multi-date RS data plus ancillary data of 30 m national elevation data (NED), LULC
NASS area sampling frame sampling
Multi-stage classification based on decision rule. Mask (non-agriculture) generation based on 2006 cropland data layer (CDL), forest canopy, impervious surfaces, NED re-sampled to 56 m using NN transformation
25
USA-2: USDA
Global, for major crops of economic
importance to the US such as wheat, coarse grains, rice, oilseeds, and cotton
MODIS 250 m and recently converted from Landsat to AWiFS
Supervised/unsupervised/decision rule semi-automated
algorithms for multi-stage classification
Zambia National for key crops
List frame (census 1990)
List frame proposed sampling strategy
No application of Remote Sensing
26
4
Remote sensing application
in support of agricultural
statistics
A consistent body of research has been undertaken to develop and improve the
techniques for identifying and classifying crops by means of multispectral and
multi-temporal Remote Sensing data. Several fields of application – such as
crop area estimation, crop condition monitoring and crop yield estimation may
benefit from the spectral reflectance of these data. Theoretically, the data should
link a unique spectral signature to a specific crop type, to aid crop identification
and support consistency in classification. However, when two crops have the
same or similar spectral signatures on a given date, a multi-date data set is
required to identify them. In this connection, a considerable volume of work has
been performed. This literature review emphasizes that several studies have
been performed on testing and developing automatic techniques, which can be
considered as fast, effective, timely, and cheaper mapping tools for the
classification, identification and monitoring crops that may have various
agricultural applications, even though accuracy may not be very high.
Sections 4.1-4.5 examine the applications of remote sensing in various
agricultural contexts
4.1. Crop type classification and identification
National and international statistical agencies and other agriculture-related
organizations have several reasons to identify and report on the cropped area
(see the References section) and yield, and to estimate production. As explained
above, remote sensing imagery is capable of supporting rapid estimates of this
information and assist in survey design and efficiency. Remote sensing
techniques play a vital role in identifying and classifying different crops and
determining their associated acreage, conditionality and potential yield. For this
purpose, production is often estimated using multispectral and multi-temporal
data.
27
Scientific investigators have often had to address the challenge of automatically
identifying and classifying crop types in a timely and cost-efficient manner.
Crucial factors appear to be the availability of imagery in coincidence with field
surveys; the possibility of managing two or three different images acquired in
different periods and of conducting an adequate field verification programme as
part of the validation exercise; the complexity and heterogeneity of the
agricultural landscape; and the presence or absence of homogeneous crops in
the area under study. These factors greatly contribute to the quality, timeliness,
efficiency and ultimate accuracy of the final estimate.
An example is given by Pardit et al. (2006), in which the authors consider
LISS-IV and IKONOS images to discriminate between sugarcane and onion in
a restricted area using Maximum Likelihood Classification (MLC). Particular
attention is given to the date of image acquisition to improve the definition of
the crop types during the different stages of growth. The overall accuracy in
discrimination ranges from 82 to 86 per cent for sugarcane crops, and from 91
to 94 per cent for onion crop fields; it also appears possible to obtain a better
discrimination level for mixed cultures and monocultures.
Finally, the correct and accurate identification of crop types can be attributed to
the following familiar factors:
spectral and spatial resolution linked to the remote sensing product
chosen and/or available;
percentage of occurrence of pure or mixed pixels, which influences the
spectral signature;
field crop size;
multi-date data set availability; and
field validation activity.
4.2. Crop condition (monitoring) assessment
An important group of applications takes place in the context of crop condition
monitoring. Recurring passages of the satellite are fundamental to monitor
crops on a timely and accurate basis. Crop condition monitoring is also very
useful in crop acreage and crop yield estimation. The monitoring of crop
condition is important to identify location-specific sites at the farm level where
growth has decreased or is impaired. This can provide information that is
necessary to improve the quality and timeliness of decisions on crop
management. Remote Sensing imagery delivers the required spatial overview of
the farmland and can be helpful in identifying the crops that are affected by less
28
than optimal or stress-related conditions (e.g. scarcity of water and nutrients,
pest attacks, disease outbreak and weather conditions, physical parameters and
crop indices). Early detection of these conditions can help farmers to mitigate
their effects and to monitor the outcome of their treatment.
Some researchers emphasize that a clear definition of “crop condition” has yet
to be established.
Vegetation indexes for direct monitoring
Many useful indexes for crop condition monitoring are now available. They are
widely used in agricultural applications of remote sensing and are relatively
easy to derive. Indeed, the different band value ratios do not require complex
computation efforts; however, the interpretation of the output must be analysed
carefully. Table 6 illustrates some of the most common and useful vegetation
indexes.
Table 6 - Vegetation Indexes
Vegetation index Formula Bands
RVI
(Ratio Vegetation Index) RVI=NIR/RED NIR and RED
NDVI
(Normalized Difference Vegetation Index)
NDVI=NIR-RED / NIR+RED NIR and RED
NDWI
(Normalized Difference Water Index) NDWI=NIR-SWIR / NIR+SWIR NIR and SWIR
SAVI
(Soil Adjusted Vegetation Index) SAVI=(1+L)( NIR/RED)/NIR+RED+L NIR and RED
TVI
(Triangular Vegetation Index) TVI=0.5(120(NIR-GREEN)-200(RED-GREEN)
NIR, RED and GREEN
TNDVI
(Transformed Normalized Difference Vegetation Index)
TNDVI=[( NIR-RED/ NIR+RED)+0.5]½
NIR and RED
The vegetation indexes (differences and/or ratios) are used to compare different
years and thus monitor the growing state of crops. Some applications consider
same-period comparison data outputs.
29
The crop growth profiling monitoring method is an NDVI analysis that
compares crop growth profiles over time within the duration of the growing
season, for different years. NDVI time series at provincial or regional level are
used. Using NDVI, it can be seen that different crops have different
characteristics in their respective growth profiles, and that the same crops have
different NDVIs in different environments. VITO has developed specific
software (namely, SPIRIT) to analyse larger areas (at national or regional
scale).
Image classification methods
Unsupervised and supervised classification methods are also widely applied to
monitor crop growth status, using multi-temporal satellite imagery. However,
these methods should not be used to derive statistics on crop area in areas that
are dominated by small fields or where there are mixed cropping patterns, as
they would achieve only approximate results.
See the References section for further examples of applications and details.
4.3. Crop area estimation
Crop area estimation is the “backbone” of agricultural activities. It is also a
critical parameter for yield estimation. Reliable and timely crop acreage
estimation is very useful in agricultural planning and to support decision-
making. To estimate crop areas, three main methods of remote sensing may be
applied:
1. Pixel counting. This is the fastest and simplest method, but one that
encounters several limitations due to biased results. The bias in area
estimation may contain the same errors of omission or omission errors
(Gallego, 2008).
2. Regression, calibration, and small area estimation. These methods are
based on a combination of remote sensing data and accurate sample
information (ground surveys). The regression estimator of survey
estimates is a standard technique for estimating the mean of a variable Y
known for n samples using the auxiliary variable X known for the n
elements of the entire population and correlated with Y (Latham
1981/82/83; Carfagna et al., 2005). The calibration estimator is a
statistical family of procedures in which the biased pixel counting
estimator is corrected with the help of the confusion matrix computed
on samples (Carfagna et al., 2005). Small area estimation is a family of
30
statistical techniques in which auxiliary data are used to downscale
sample results to smaller populations, when the sample itself is unable
to provide sufficient statistical support.
3. Area frame surveys. Satellite images can support the area frame in
several ways: i) they can help to define the sampling unit for
stratification and ii) as graphic documents for ground surveys and
quality control. The application of remote sensing imagery to area
frame preparation is extensively described in Section 5 below. In
particular, Gallego (2008) has emphasized that stratification based on
coarse photo-interpretation of satellite images was cheap and cost-
efficient. Further analysis performed using a land cover database
prepared with the CORINE legend has confirmed that the method
makes it possible to achieve high efficiencies.
4.4. Mapping soil characteristics
Agricultural management and development benefit from information on soil.
Remote sensing imagery provides a synoptic overview of the study area,
making it possible to map soil types and related characteristics quickly and
more effectively than with traditional soil samplings and laboratory analyses
which, although very precise, require a great deal of time and are costly. A
significant body of research exists on the use of microwave (active and passive)
Remote Sensing for mapping soil characteristics (soil salinity, ravine erosion
inventory, sand dune characterization, soil texture and hydraulic proprieties,
soil drainage and soil surface roughness).
4.5. Precision farming practices
Precision farming is an emerging methodology that is designed to link
management actions to site-specific soil and crop conditions, placing inputs of
fertilizer, herbicides, and pesticides where necessary to maximize farm
efficiency and minimize environmental contamination. Precision agriculture
requires locating specific positions in the field based exclusively on remote
sensing, GIS, GPS and GNSS technology.
Several papers and articles illustrate many of these applications in detail and are
listed in the References section.
31
5
Remote sensing techniques
and area frames
This section describes the standard procedure for constructing the area frame.
Particular emphasis is placed upon the steps in which remote sensing products
and technologies are involved. The literature contains several examples that
refer to area frame construction.
The conventional area frame sampling survey is a type of agricultural
probability sample survey activity that is introduced as a vehicle for conducting
surveys on crop acreage, cost of production, farm expenditures, yield and
production, livestock inventories and other agricultural items. It is a costly and
time-consuming method that requires a basic data set (aerial
photographs/images) and topographic and thematic (land cover/use) maps or
databases. Meticulous and laborious work must be undertaken to delineate
strata, primary survey units (PSUs) and then segments, and qualified staff with
a strong background in statistics must be employed. This section presents the
steps in which remote sensing products are involved and attempts to evaluate
the efforts required.
Area frame sampling consists in “dividing the total area to be surveyed into N
small blocks (Segments) without any overlap or omission, furthermore select a
random sample of n small blocks and get the desired data for reporting units of
the population that is in the sample blocks” (Madana, 2002). When stratifying
the survey area into a number of strata (which are then divided into PSUs), it is
important to ensure homogeneity within a stratum: this will make it possible to
reduce the sampling variance and to obtain more accurate estimates.
Stratification and the correct definition of strata are, therefore, crucial steps. An
accurate land cover database can facilitate and improve the process.
In the final stage, the sampling units of an area frame are defined. These land
areas are called segments. They should not overlap and should cover the entire
survey area. These land parcels can be determined on the basis of factors such
as ownership, on easily identifiable boundaries or on square segments defined
by straight lines that form squares the endpoints of which are established by
32
map coordinates (FAO, 1996). To gather agricultural information, segments
should be visited by an interviewer (Houseman, 1975).
Surveys are very expensive undertakings that involve significant logistic
efforts. Enumerators collect field data, completing a questionnaire by means of
personal interviews with holders or other respondents who can provide
information on the tract included in each sample segment selected. It is vital
that the enumerators visit the field at the correct time to gather the information
sought. Enumerators use topographic or road maps showing the segment
locations to identify routes of access and reach the segment. They verify that
the boundaries have been delineated appropriately and commence the surveys
with the nearest occupied dwelling, or by approaching the nearest visible
worker in a field within a segment. Then, the enumerators identify the holder of
a tract within the segment. The holder helps enumerators to delineate the tract
on a transparent overlay of the segment photo.
In addition to these questionnaires, data collection involves the identification
and measurement of cultivated areas. For each sample segment, the enumerator
uses an enlargement of an aerial photo (or a map) that shows the boundaries of
the segment. This is called the segment photo. For each tract within a given
sample segment, an enumerator delineates the boundaries of the tract and of all
the fields included therein. The enumerator verifies the crops planted and other
land uses for each field, and may also seek this information from the holder.
During the interview, the enumerator may also superimpose a transparent grid
over the segment photo for an approximate verification of the fields’ reported
area. The agricultural areas identified in each sample segment can be measured.
It is crucial to verify the area estimates made by holders and/or enumerators if
data reliability is to be ensured (Houseman, 1975; Latham, 1982).
Area frames are critical to the production of quality estimates, because they
provide complete coverage and land areas are represented in a probability
survey with a known chance of selection (Cotter and Tomczak, 1994). This
frame does not suffer rapid obsolescence unless the population extends into
areas that are not covered by the frame (FAO, 1996).
To prepare an area frame, the first requirement is the availability of up-to-date
cartographic materials. Their resolution must be sufficient to enable
stratification and the subsequent subdivision of these strata into PSUs, which
must be demarcated by recognizable permanent physical boundaries. PSUs are
usually constructed from photographs or satellite images in which the
boundaries of the strata are visible and are then transferred to maps for
33
measurements. Each PSU must be measured and assigned a target number of
segments. The number of segments in each stratum is added again to provide
the total within the frame. Then, a two-stage probability sample of segments is
selected from each stratum with a replicated selection procedure. Each sample
segment is constructed on small mosaics of aerial photography, upon which the
boundary of the corresponding PSU has been transferred. Finally, the selected
sample segments are located on the aerial photo enlargements used to control
field data collection (FAO, 1996).
The above steps can be summarized as follows:
1) Prepare frame materials, delineate frame limits, and measure total
frame area.
This is the basic step for stratifying large areas. Here, it is necessary to select
the product to be used (photo-mosaics, available ortho-photomaps, satellite
images, maps, or a combination of any or all of these). Once the material is
ready, boundaries are drawn to delineate the potential sample units and frame
limits. In particular, it is necessary to:
delineate and measure areas covered with water (lakes, large rivers,
etc.), heavy forest, high mountains, national parks, military reserves,
and other non-agricultural land, except urban areas.
outline and measure the urban and agro-urban strata; the city’s central
portion is defined as non-agricultural, while an area with high
population density that also includes patches of agriculture is indicated
as agro-urban agriculture.
delineate strata in agricultural lands; in agricultural areas, the strata are
defined by the proportion of cultivated land, the predominance of
certain crops, the average size of cultivated fields and special sites of
agricultural activities.
2) Transfer strata boundaries to map; measure and field-verify strata.
The strata boundaries are then transferred from satellite images or photo-
mosaics to maps. In this step, the strata should be measured and their total area
compared with the data available.
34
3) Construct PSUs.
The PSUs are constructed. The PSUs should reflect, on a small scale,
those that are considered the stratum’s characteristics.
The PSUs are transferred to maps and ordered.
PSU area is measured.
Measures of size are assigned to PSUs, strata and the total frame.
As PSU areas are determined and approved, they are entered on another listing
sheet. Next, segments are assigned to the PSUs as determined by PSU area and
the target size of segments in that stratum. The number of assigned segments is
equal to the area of the PSU divided by the target size of the segments in the
stratum, approximated to the nearest integer. The PSU size measure should be
an integer.
Generally, the stratification is carried out with images, Aerial Photographs
(APhs) or maps; the different land cover classes are identified on a generic and
approximate basis. The availability of a reliable and consistent land cover
database can be an important element in optimizing, improving and accelerating
the various steps of area frame construction.
When stratification is homogeneous, the margin for sampling error is lesser. If
stratification is less homogeneous, sampling errors are more likely to occur.
The availability of a consistent land cover database can provide necessary
support in correctly delineating the strata and reducing sampling variance.
Latham (1982) examines an example from Libya, while Gallego (2008)
discusses EU surveys using a land cover map prepared with the CORINE
legend.
All outputs described in Step 1 above are available immediately and with a high
degree of accuracy if a valid land cover database is easily accessible. The
percentage of agricultural types of crops and many other features of interest can
be readily “extracted” from the land cover database, and the accuracy of the
land cover data can be found through a quality assessment of the database. In
addition, the digital database can be reprocessed or updated easily when new
images or data become available. If necessary, systematic reviews of the land
cover or agriculture changes can be implemented every year, whenever a new
image data set is acquired and a digital database is available.
Steps 2 and 3 are standard procedures that can usually be performed quickly
and accurately in a GIS environment. Any GIS software can provide sound
35
support to the correct overlap of different data sets, as well as to the
optimization of the routes and calculation of the boundaries.
Furthermore, prior to the survey, a digital (object-based) segmentation of the
PSUs and segments on the image/APhs can be generated. The resulting geo-
referenced vector database can be useful to enumerators in the field. The
segmentation can rapidly generate an accurate vector file in which the physical
boundaries are delineated and the main features and their linked attributes (e.g.
area and perimeters) are vectorized. If the data set of updated images is
available before the fieldwork, the crop extensions can be calculated in advance
with very little effort in terms of cost and time.
In addition, the digital database is a strong framework for later surveys.
5.1. The FAO LCCS approach: the contribution to the area
frame
Land cover maps derived from satellite imagery provide a suitable model of the
real world. However, there are several differences in how the databases or maps
are produced. This is reflected in the accuracy and intrinsic meaning of the
database’s existing classes. Whatever the production process, the data are
immediately handled by users who are unaware of the geographic information’s
intrinsic meaning.
However, experts will have different basic perceptions of the information,
depending on their disciplinary sensitivities. In other words, there is confusion
as to how different users treat land cover information, a confusion that
originates partly from how land cover information is defined and generated
(Comber et al., 2004).
Land cover and land use information is transferred from its producers to its end
users, providing a useful baseline with which “land” analyses can be performed.
In addition, there is widespread misinterpretation of the concepts of land cover
and land use among users.
Land cover is the observed (bio) physical cover of the Earth’s surface. It
includes vegetation and man-made features, as well as bare rock, bare soil, and
inland water surfaces.
36
Land use is a description of how people utilize the land (Comber, 2006). Land
use is characterized by the arrangements, activities and inputs that people
undertake in a certain land cover type to produce change or to maintain it. Land
use establishes a direct link between land cover and the actions of people in
their environment.
From satellite interpretation, it is possible to capture the land’s land cover
features. Its use, instead, is linked to many other factors (e.g. social, cultural,
economic) that can be defined only when ancillary information becomes
available.
A number of classification schemes have been devised to classify and record
land cover and land use features and to monitor their changes.
The FAO-NRL geospatial unit has extensive experience in land cover mapping.
It has also defined and developed a new standardized classification system. In
2002, the unit proposed the Land Cover Classification System (LCCS2)
scheme, which later evolved into LCCS3 - Land Cover Meta-Language
(LCML). The ISO standardization approval for Geographic Information ISO
TC-211 was granted in July 2 (see
http://www.iso.org/iso/iso_technical_committee?commid=54904 ). LCCS3 is
based on the idea that it is more important to standardize the attributed
terminology than the final categories in land cover classification. LCCS works
by creating a set of standard diagnostic attributes (called classifiers) to create
and describe different land cover classes (ISO19144-2-2012; DiGregorio,
2011). The classification method introduces an important innovation with
respect to previous classification systems (e.g. CORINE Land Cover, USGS
Anderson Classification, etc.): the LCML is based on the use of classifiers as
class elements that are defined in unambiguous terms throughout the world, and
provides the classification rules to describe land cover types.
Several advantages can derive from the use of a land cover database classified
using LCCS (2-3). For example, it ensures:
an ISO standardization of the land cover database;
a clear definition of the land cover classes through a systematic
description of the land cover elements, using an object-oriented
approach that ensures a high degree of flexibility of the thematic
content;
that the biotic and a-biotic features are identified and linked to the
database;
37
the clear semantic meaning of the classes used, and no limitations in
describing the elements present in the ground;
scale- and pixel-independent;
consistent and interoperable outputs;
the classification can be done “per elements” or “per classes”— LCCS
enables a detailed description of the elements, the level of accuracy
depending only on the sources of information;
consistent and effective structures that enable aggregation at element
level, in accordance with specific needs;
the possibility to define the level of detail; this can be applied in a very
simple classification, such as "Agriculture/Not Agriculture", or in a
more sophisticated and comprehensive land-cover database; and
the final digital database can be linked in a GIS environment with the
land’s other elements (social, economic and environmental).
The innovation proposed by LCCS3 is the object-oriented database. It provides
excellent input for describing basic elements that, linked with high or very high
remote sensing sources, could dramatically improve the stratification phase and
the sample point allocation.
In addition, it can be developed as a multipurpose database and can be used in
many environmental applications.
5.2. Generating the Land Cover Database
This section briefly presents FAO’s mapping approach to land cover database
preparation, from pre-processing to generation.
This is a multifaceted procedure, created on the basis of decades of experience
in constructing detailed national multipurpose databases. Specific and
standardized mapping procedures were established to ensure a reliable,
consistent, and comparable product.
38
The scheme below illustrates the principal steps of the FAO methodology.
The selection of appropriate and suitable remote sensing coverage is essential.
The interpretation is based on the possibility of discriminating features in the
image. Therefore, keeping in mind the ultimate purpose of mapping, it is
fundamental to perform an accurate selection of the source data.
Many types of commercial satellite data are available at different costs and
resolutions. Taking into consideration the elements of interest and the
territory’s characteristics, a correct balance of these parameters is key to a
successful mapping.
The next step in the correct preparation of data sets is the pre-processing
procedure. This set of procedures is familiar to remote sensing users and is
widely presented and described in specific publications. When a mapping
activity is intended to generate or update cartographic coverage on a national or
regional scale, a great number of data sets must be prepared and pre-processed.
Pre-processing is performed to improve the images’ geometric and spectral
characteristics. Very often, data providers deliver images that are ready for use,
and pre-processing focuses mainly upon mosaic creation and band composite.
The FAO approach is vector object-oriented; the vector layer is generated
during the automatic analysis of the remote sensing image. This process, called
“segmentation”, is performed using ad hoc software (e.g. Madcat by FAO or e-
Cognition by Trimble), and generates objects that represent regions of images
39
having similar spectral characteristics (reflectance pixel values, texture). This
process facilitates analysis of all common data sources, minimizes image-
analyst subjectivity and increases cost/benefit efficiency.
The experts then interpret the segmented layers — usually visually, on screen –
to achieve a finer distinction of the different land cover units. Several automatic
or semi-automatic procedures for clustering polygons with similar land cover
characteristics have been developed to assist and accelerate interpretation.
These modules are available in dedicated software, including the FAO-NRL
tool.
The legend is the land cover database’s explanatory key. Its preparation is an
important and critical step. It must reflect and synthesize the land cover classes
present in the area of interest as much as possible. The information must not be
redundant and must be arranged correctly. Therefore, where possible, its
preparation is performed in collaboration with country experts who are familiar
with the specific land cover characteristics of the area of interest and are able to
enrich the legend in detail.
LCCS is a comprehensive, a priori, pixel-independent land cover classification
system that is well-designed to meet user requirements. The preliminary version
of the legend must be created with LCCS at the beginning of the project.
During interpretation, this version is modified whenever a new element is
identified. The final version of the legend is obtained only upon completion of
the interpretation activity. To avoid subjective interpretation, and to obtain a
common reference of the land cover types present in the study areas, an attempt
to create a photo-keys album must be performed at an early stage for all the
interpreters involved in the mapping activity.
Once these preliminary steps are completed, the work focuses upon image
interpretation. This is the core of the database creation and is the longest and
most complex part of the entire process. It entails control of a massive number
of polygons (40,000-100,000 per Landsat scene) and the identification therein
of similar land units and/or complex associations of land cover types. It also
requires consistent labelling, using the thematic categories present in the
legend.
Upon completion of the preliminary interpretation, it is strongly recommended
to conduct a fieldwork campaign, because this can enhance database reliability
and minimize interpretation errors.
40
Alternatively, the validation can be performed using Google Earth (or any other
available high-resolution data set), making use of the high-resolution images
that are available online.
The final review of the interpretation is a crucial step in consolidating the
database. The harmonization work must generate a consistent land cover model
of the area, minimizing the differences arising from the subjectivities of the
various interpreters. For this purpose, the final phase must be controlled by a
restricted number of remote sensing experts. In addition, the technical and
semantic aspects should be verified with great attention. The final database
must have a correct topology (the correct geometric arrangement of point, line,
and polygon features) and arrangement of the attribute’s table. For this purpose,
FAO has developed several procedures that require skilled expertise in GISs.
The duration, accuracy and quality of the results are closely linked to various
factors, such as the number of total images covering the study areas, their
resolution, and the skill of the remote sensing and GIS experts.
Moreover, the synergic involvement of national experts is an essential part of
the process: local knowledge and familiarity with the study area and its features
are considered key factors in creating a reliable national database. Different
procedures and “hands-on training” have been developed to ensure that this
vital contribution can be made to the mapping activities.
Reaching a final multipurpose land cover database is a complex and time-
consuming process that requires the involvement of a number of national
subjects and actors. The final product is a good quality and useful baseline,
which can improve area frame analysis and support several agricultural
activities.
In this connection, the following activities are also recommended:
strengthening the national capacity to map activities;
supporting the national capacity to produce the land cover baseline and
to manage remote sensing data and related issues;
supporting national institutions in finalizing a common multitasking
product; and
catalysing interest in and awareness of agriculture-related products.
41
Finally, the land cover product generated using the method described above has
a further added value: it can be revised easily as soon as new source images are
available. The vector database can be updated using FAO tools and tested
methodology. This is important in terms of the product’s life expectancy.
5.3. Two case studies on land cover database generation
in support of area frames: Ethiopia and Pakistan
In recent years, the FAO-NRL geospatial unit has assisted the Governments of
Ethiopia and Pakistan in improving agricultural statistical data to support and
implement the generation of a standardized land cover database using the LCCS
classification system. The mapping activity also includes the development of
national capacities in land cover database generation.
A first attempt was carried out in Libya in the 1980s. The production of
agricultural statistics on crop acreage was supported by the use of remote
sensing and field data. Land cover mapping was performed using MSS Landsat
imagery and an LCCS precursor schema.
Table 7 below illustrates the main features of the three projects.
42
Table 7 – Main features of Libya, Ethiopia and Pakistan projects.
Libya Ethiopia Pakistan
Extension 1,759,540 km2* 709,095 km2* 1,104,300 km2*
Technical support agency
FAO FAO FAO
Implementing agency
MoA
CSA (Central Statistical
Authority) EMA (Ethiopia Mapping
Agency)
SUPARCO (National Space Agency of
Pakistan)
Data used
Landsat MSS
80m Dated 1980
SPOT5 5m resolution 2,085 scenes
Dated 2006-2007
SPOT5 m resolution 310 scenes Dated 2012
Team 10 people without any
background in RS
12-15 persons
without any background in RS
10 persons with background in RS
Methods Digital processing
and visual interpretation
Segmentation object-based
Segmentation object-based
LC classification system adopted
Land Cover
Class’n System, precursor to LCCS
FAO - LCCS2 FAO - LCCS3
Mapping period From 1981 to 1983
From 2008 to date;
Intermittent
From 2012 to date Continuous
Training
On the job, over 2
years
Initial course of 2 weeks Initial course of 2 weeks
Supervision 1 land cover/RS
expert for 2 years
2 land cover experts,
constantly present for the first few years
After the first 6 months, 2 weeks of land cover
expert supervision
% of country mapped to date
completed
Approx. 90%
Where the main agricultural activities of the country are located
Approx. 30% Where the main
agricultural activities of the country are located
Completion of
mapping activities
1981 2016 2014
*Source: CIA World Factbook, https://www.cia.gov/library/publications/the-world-factbook/.
43
As a complementary step, the countries sought to develop a land cover database
for agricultural statistics. In Pakistan, the SUPARCO Remote Sensing centre
offers excellent support to mapping activity. The activity was executed
autonomously by expert teams; FAO was involved only in presenting the
methodology and provided minor supervision.
In Ethiopia, the team’s remote sensing background was very limited. FAO’s
effort focused mainly on developing capacity, providing constant support and
supervising the mapping. Currently, the team is able to continue the mapping
independently. Once it has been completed, FAO will follow the final module
to ensure correct harmonization and generation of the database. To date, the
national team’s mapping capacity has been consolidated and any further
updates, revisions or improvement can be conducted autonomously by the
CSAS-EMA. The land cover database can be adjusted and modified in
accordance with the CSA’s agricultural statistical requirements or any other
specific criteria.
In both countries, the activities are ongoing. For an in-depth overview, it is
necessary to await completion of the mapping.
5.4. Mapping agricultural land cover features using
remote sensing-derived products:
some considerations
Paragraph 4 presented the main issues regarding the applications of remote
sensing in the context of agricultural statistics. The demand for timely and
reliable agricultural information has prompted the scientific community to
consider all the possibilities and applications offered by the various remote
sensing products and fully exploit their potential. These benefits were briefly
outlined in Section 4 above. It is vital to improve the surveying methods to
produce agricultural statistics, especially in countries where these are very poor,
inadequate or of low quality.
One of the most obvious applications of remote sensing is the study of the land
surface. The mapping activity can be oriented to land cover in large areas,
mainly for environmental and natural resources studies or to estimate the
extension of cultivated land (the main interest in agricultural planning analysis).
Mapping land cover and area estimation are complementary objectives: they are
closely related, but correspond to different needs and assign different priorities
to different measures of accuracy (Gallego, 2004). It can be also stated that the
44
procedures, methodology and sources of information are the same, but they are
normally applied with different modalities and expectations. In addition, the
application of standards and methodologies improve the product’s quality.
Generally, certain prerequisites should be met if a land cover mapping exercise
is to be performed. In particular, the following should exist:
large area of interest;
longest period of production;
multidisciplinary production chain;
detailed legend covering all components of land surface/cover;
limited fieldwork campaign;
overall accuracy similar to that of all the land cover classes linked to the
MMU; and
multi-purpose applications by various end users.
On the other hand, the area estimation should be guided by specific agriculture-
oriented objectives, such as the following:
focus on agricultural areas;
need to generate results quickly, simultaneously or in support of
agricultural estimation analysis;
prepared by staff without any specific remote sensing skills;
combined with accurate and specific information on a sample from field
survey;
provide as much accuracy as possible for agricultural class types; and
prepared mainly for the purposes of agricultural statistics.
In addition, the reasonably precise identification of field areas must take into
account the necessary time and costs for their estimation and the manageability,
size and precision of the data sets.
However, both methods share an ability to combine information from accurate
and detailed field surveys with the paramount (but biased) vision of satellite
imagery. For land cover mapping, this process is limited to a validation
campaign to minimize interpretation errors. In land cover area estimation
methods, regression, calibration and small-area estimation combine exhaustive
but inaccurate information (from satellite images) with accurate information on
a sample (often from a ground survey) (Gallego, 2004).
45
Other operations may be undertaken to analyse how, whether and under which
conditions the gaps between these two approaches can be reduced and the
added value that could thereby be gained.
A standardized (LCCS-based) "high resolution" vector database can be used to
improve agricultural and rural statistics; its creation would affect the accuracy
of all parameters of estimation. Such a database does entail a relatively high
commitment in terms of time and resources, commensurate with higher levels
of consistency and accuracy, but would contribute greatly to the increased
quality of the results derived by the methodology. A cost/benefit analysis
should be conducted before choosing the methodology and products to be used.
However, LCCS-based databases have several known advantages, such as
being multi-purpose and multi-use systems; flexible, consistent and object-
based; easier to update, maintain and re-use; and tools that are based upon ISO
standards.
At the beginning of the statistical implementation plan in a country or region, a
feasibility study assessing the landscape characteristics should be performed,
especially when a multidisciplinary or multitasking cartographic database is one
of the objectives of the statistical operation.
Innovative options for correctly defining the environment in relation to its
agricultural characteristics can be an interesting subject of research. Its
development must take into consideration the following aspects:
the remote sensing spatial resolution (pixel size) of the products;
the complexity of the surface features of the landscape to be
investigated; and
the relationship between the above.
These elements and of their correlation could be assessed in the initial stage, to
select the optimal sensors to be adopted in light of the environment’s
complexity. This process requires an understanding of the relationship between
the sensor’s spatial resolution, the size and distribution of ground objects and
the surface measurement of interest (Turner, 1989).
The technical aspects relating to Remote-Sensing derived products are
extensively described in Section 2 of this literature review. Pixel resolution is
very important to resolve landscape elements. Indeed, “[t]he fixed pixel
resolution of a specific satellite product determines its limitations — that is,
whether or not it can be used to identify an element on the ground.” Several
46
authors have addressed this matter (Pax-Lenney and Woodcock, 1997;
Townshend and Justice, 1988; Ozdogan and Woodcock, 2006; etc.).
LCCS is a scale- or pixel-independent classification system specifically created
to capture these elements and systematically describe landscape. Regardless of
the definitions commonly used to describe the features of the study areas, the
FAO LCCS is capable of defining – with great accuracy and abundance of
information – any type of land scenario. It is capable of reporting “granularity”
of information without any limitation on the details or predefined schema.
Section above presents the main characteristics of LCCS3, which is a ISO
standard classification system that is capable of appropriately describing
environments for mapping purposes.
Landscape complexity can be described according to several parameters that
highlight specific features and differences, such as:
the Agro-Ecological Zones (AEZ), including the crop calendar, shifting
cultivations and crop intermixing, and
agricultural patterns and practices (defined using crop pattern, crop field
size, and crop intensity).
These two (ecological and geometrical) aspects can be used to describe the
landscape thoroughly. In any landscape study, the representation of land surface
properties and ecological processes are closely linked, but they are also
inherently related to the scale adopted in the analysis. This scale-dependency is
also true for stratification purposes in area frames and many other applications
in agricultural statistics. Moody and Woodcock (1995), Meentemeyer and Box
(1987) and Milne (1992) have discussed the various aspects of this subject.
It is thus important to understand how the loss of information due to the
transference of land cover elements or data across the scale affects
stratification. It is also important to quantify the reduced information content
and to understand how to mitigate it. Area estimation is often taken for granted
in remote sensing when, in fact, large errors in area estimates for classes can be
made, even when based on accurate thematic maps. These errors may be greater
when the size of the field is larger than the size of the pixels in the image and
all pixels become mixed pixels (Ozdogan and Woodcock, 2006). The map or
database is a source of information. The means of transmission of this
information to end users is important.
47
Agro-ecological characterization is defined in terms of the climate, landform,
soils and/or land cover, with a specific range of potentials and constraints for
land use. An AEZ is a division of an area of land into smaller units that have
similar characteristics in terms of land suitability, potential production and
environmental impact. These elements are very important and provide a
consistent model for farming system zones, although – due to the complex
nature of the study – its complete and correct implementation in the agricultural
context should be evaluated carefully.
Several authors have analysed the complexity of the landscape in relation to the
spatial pixel size, stressing the quality of the mapping result in terms of
accuracy.
Examples of different pattern distributions and crop arrangements are given in
Figures 1a and 1b below. The first was taken in Russia, to the east of Moscow;
the second was taken in Ethiopia, in the West Showa region near Addis Ababa.
In the geometrical and spatial conditions of Example 1, Fritz et al. (2008)
successfully utilized a MODIS13 16-day NDVI composite, 250 m resolution, to
estimate the acreage of larger fields for certain crop types with a distinctive
profile.
In Figure 1b, however, considering that the agricultural pattern was
characterized by small fields and scattered rural houses (gardens), SPOT
images with a resolution of 5m are a valid compromise for mapping source and
agriculture ground surveys.
The landscape patches on the Earth’s surface are essential to ecological and
geographical studies. Through remote sensing imagery, landscape patches can
be extracted and treated as objects and classified according to their physical or
geometrical characteristics. The accuracy of object characterization varies as a
function of the spatial resolution of the imagery used: Maud and Foody (2013)
discuss the problem of the accurate and correct classification of smaller objects
(i.e. smaller than pixel size) from coarse spatial resolution remote sensing
images. They propose specific geometric elements to define object shape (e.g.
area, perimeter, circularity, length and shape orientation, and spatial formation)
as well as a mapping algorithm for a super-resolution mapping technique, to
gain accurate information on landscape patches when finer high-resolution
mapping is impractical.
Geometric characterization can be exploited and examined in depth to describe
the landscape; the use of high-resolution mapping may be considered the best
48
solution when optimal ground conditions exist. Xie et al. (2013), when
discussing remote sensing imagery in vegetation mapping, conclude that rather
than using field samples to test the accuracy of classifications, a widely
accepted practice is to use higher-resolution satellite data to assess resolution
products, although high-resolution data are themselves subject to interpretation
and possible errors.
The influence of the scale and spatial characteristics of the landscape on land
cover mapping using remote sensing is extensively discussed by Moody and
Woodcock (1995). They propose an approach to the incorporation and
comprehension of the relationship between the scale of observation, the spatial
aggregation of land cover classes and classification errors. Spatial
measurements and a multilinear regression model are suggested. The research
demonstrates that large proportion errors arise as land cover data are sampled at
progressively coarser scales. The relationship between land cover spatial
patterns and the scaling of cover-type proportions can be understood and
modelled using fairly simple measurements of the landscape’s spatial
characteristics.
The study undertaken by Pax-Lenney and Woodcock (1997) in the Nile Delta
shows that the accurate estimation of the area of productive, cultivated land
from coarse spatial resolution data is indeed influenced by landscape
characteristics, such as small average farm size, variable cropping schedules
and cultivation of peripherical lands. The authors compared scene statistics at
different resolutions and tested the effectiveness of coarse spatial resolution
data to detect elements and changes.
These effects increase as the cell size increases, although not dramatically, and
the overall impact on area estimates is relatively low. However, the interesting
aspect of their application lies in the differences (in terms of spatial) pattern
between the classes of interest of the landscape and their variability in area
estimate accuracy. The spatial pattern of the classes of interest was evaluated
using two simple measurements of spatial cohesion: patch size and variance-
mean ratio. Patch size is defined as the number of four-way contiguous pixels
within a cluster of pixels of the same class. Old agriculture is dominated by the
existence of a few very large patches, while reclamations typically feature a
greater number of smaller patches. The variance-mean ratio is a measure of
clustering used with point data, based on the count of points within a series of
identical cells.
49
The conceptual framework presented by Duveiller and Defourny (2010)
addresses the question of spatial resolution. This framework evaluates the
spatial sampling requirements for agricultural assessment over various
landscapes with contrasting patterns, and focuses on identifying the coarsest
acceptable pixel size for both crop area estimation and crop growth monitoring.
The products or spatial resolution must be appropriate to the spatial pattern
existing in the area of study. The complexity of the landscape is also an
important element to be investigated in relation to the spatial sampling
requirements for agricultural assessment. Duveiller and Defourny present an
exercise to demonstrate the requirements (in terms of pixel size) to monitor
agriculture over different landscapes. Crop fields must be different not only in
terms of size and shape, but also in their spatial distribution. However, different
applications in crop area estimation or monitoring require different pixel size or
pixel purity. These two aspects are analysed in detail, along with the mapping
variables across the pixel size/pixel purity dimensions. This methodology
should be examined in detail for use in the final, correct definition of
guidelines.
In this section, it will be shown that it is crucial to choose the correct product
for data acquisition in several applications, especially when crop production
and acreage must be estimated accurately. Unsurprisingly, the literature on this
topic is very rich. However, specific analytical guidelines for choosing the most
appropriate product according to landscape features are yet to be established.
50
Figure 1b – Agricultural features in Ethiopia (Addis Ababa, eastern area).
Where suitable conditions for the production of a new cartographic product
exist, the environment’s specificity – including its agro-ecological
characteristics – should be studied preliminarily, to identify the best remote
sensing products to be used, the MMU, and the necessary accuracy.
Additionally, adoption of the LCCS3 standardized classification system ensures
that a valid and reliable approach is taken to correctly describe the elements on
the ground. However, the main guidelines that should apply to such an analysis
have not been clearly indicated.
There are several benefits to a geo-referenced geographic database; therefore,
its creation should be supported. A detailed description of land cover is a
baseline for describing the changes that occur over a given time frame. The
identification of areas of land-cover change in relation to a certain period is
important for the systematic updating of the cartographic database and
extension of the product’s life expectancy. The area of change must be
delimited with geometric precision, to define the exact extent and location of
changes in the study area. The conversion from one land cover type to another
must be documented, and the spatial information associated with a "From-To"
table that reports the trajectory of the changes. It is often possible to observe an
expansion of agriculture into areas where the cover previously consisted of
natural vegetation. Specific tools and methods have been developed by FAO-
NRL geospatial units and have been tested in several areas of study.
51
In agricultural statistics, the baseline must be updated with any changes that
may occur over the years. This work can be very time-consuming and the
output may not be ready in time to be used in a forthcoming statistical
campaign. A systematic and integrated approach can be considered in view of
the availability of new remote sensing products. The cartographic baseline can
be characterized using AEZs. Using a new remote sensing product at 10m
resolution available every week, an NDVI procedure to detect the hotspot area
of change can be implemented. The feasibility and practicability of
implementation is a subject that could be researched further.
Land cover is closely connected to land use. The rational description of land use
requires a holistic approach, based on the integration of land cover information
with geospatial and socio-economic data, their joint analysis, and modelling.
Reliable and up-to-date geo-referenced information on land use is required, as a
basis for the sustainable development of land resources and agriculture in both
urban and rural contexts, and to inform the development of policies across all
areas of human activity at national, regional and local levels. The need for
information on land use is evident and should be a central consideration, as it is
necessary for the purposes of:
establishing a complete national or regional system;
providing a national and consistent basis for identifying recording and
reporting the use of the land; and
correctly planning, exploiting and managing the land.
Development standards and guidelines for land use mapping is important to
enable adoption of a consistent approach at global scale.
52
6
New products
Performance of a preliminary study in a country or an area of interest may be
beyond a project’s realistic contingencies. Therefore, a detailed evaluation of
existing cartographic databases should be considered, and standard parameters
to define the quality of the product should be established. Examples of the most
basic specific parameters are data sources, dates of production, legend
characteristics, and spatial and thematic accuracy.
A viable alternative could be to use existing databases with global coverage.
GLOBCOVER, prepared using MERIS 300m pixel resolution and classified
using the LCCS2 legend, is available throughout the world. However, at
country level, its value is yet to be evaluated.
In 2013, FAO created another global product to respond to the growing need for
reliable information on the global extension and distribution of the major land
cover types. This is the Global Land Cover-SHARE (GLC-SHARE), a new
global-level land cover database established by FAO-NRL, Land and Water
Division, in partnership with and with contributions from various partners and
institutions. It provides a set of eleven major thematic land cover layers
resulting from a combination of the “best available” high-resolution national,
regional and/or sub-national land cover databases. The database is produced
with a resolution of 30 arc-second2 (approximately 1 km2).
GLC-SHARE strives to bring these data sets together, offering multiple
benefits and seeking to enable a platform for partnerships and contributions by
all. The approach implemented is based on the use of the LCCS to harmonize
the various land cover databases available. The legend adopted refers to the
System of Environmental Accounting (SEEA). Available land cover legends
are translated using the land cover classifier elements to assign adequate
classifier values and ranges at an intermediate step, followed by a second
iteration in which they are converted into the major land cover classes by
assigning the class, class unit, minimum, maximum, range, and best estimate
values for each land cover class. The results are reported at 30 arc-second grid
cells as a percentage land cover class in a later step. The outputs make use of
thematic and spatial high-resolution land cover databases for various areas
53
globally (up to 66 per cent of the total surface area is covered by high-
resolution data sets). Where national or sub-national land cover data are not
available, the information refers to the global data sets. The preliminary results
are validated and calibrated. The product is by definition available at various
accuracy levels, relating to the original data. The validation phase enriches the
final product’s reliability. This is a crucial phase, and may be obtained using the
high-resolution images available and the effort of experts— even if the
production chain is the most time-consuming part. For further details, see
Latham et al. (2013).
Another global product, which is relevant because it is generated using
Landsat (TM and ETM+) 30m resolution, is the database that was presented in
April 2013 by the People’s Republic of China. The product has quasi-
worldwide coverage (currently, only Antarctica and Greenland are excluded). A
wide time series from the 2006-2010 Landsat data set was automatically
classified to generate the final database, with over 91,000 training examples and
38,000 test samples. LCCS2 was used to classify the land cover classes.
However, the accuracy of this first release spans from approximately 64 to 54
per cent. This low precision rate can constitute a major limitation in application
of the product to agricultural statistics.
Many national databases are also available online (e.g. Africover databases for
East Africa); readers are directed to the GEONETWORK
(http://www.fao.org/geonetwork/srv/en/main.home).
54
7
Gaps and expected
developments and
implementation
Since 1970, there has been a tendency to consider the issue of agricultural
statistics as being closely linked with remote sensing. The ability of remote
sensing to present the actual ground situation is undeniable: agricultural
features are clearly visible in the images. Latham, Carfagna and Gallego (2005)
and many other authors (see Reference section) have provided a description of
the different ways to use remote sensing for agricultural statistics.
However, many factors limit the intensive application of remote sensing in
agricultural statistics, in particular as regards the production of a land cover
database in support of Master Sampling Frames (MSFs). These gaps are mainly
linked to:
standard procedures for updating data production and assessing the
accuracy of existing baselines;
technical capacity;
product costs;
appropriate technical structures (including hardware, software and
know-how); and
appropriate data policies for data dissemination and maintenance.
In general, in terms of remote sensing, these gaps may be expressed as follows:
limited staff and capacity of the statistical units assigned to the
collection, processing, analysis and elaboration of RS products and
geospatial analysis for agricultural statistics;
lack of adequate geospatial technical tools, standardized methodology
and approach to supporting agriculture;
insufficient funding allocated to remote sensing for land cover baseline
production and other applications in agricultural statistics;
lack of coordination and agreement among experts;
55
lack of reliable databases due to the absence of appropriate metadata or
any other specific indications; and
lack of capacity to analyse gaps in geo-referenced databases.
In conclusion, it is suggested that further research be devoted to analysing the
basic geographic data sets that could be used in statistics. Possible topics could
be the following:
development of standard procedures and parameters for the quality and
accuracy assessment of existing cartographic databases, in view of
the implementation of agricultural statistical analyses;
development of specific guidelines to define the database’s
obsolescence, and the criteria or methods to update it;
development of evaluation criteria to establish the feasibility of a
cartographic production, according to specific in-country conditions
and the project’s expectations or objectives;
definition of basic parameters for landscape geometric-spatial
description and of minimum requirements and specific conditions to
develop an agro-ecological study;
development of an integrated and systematic approach to quantify the
minimum scale of representation in relation to landscape
characteristics, and of the implications of the possible loss of
information due to the adoption of an inadequate scale of the source
products;
development of more efficient and robust methods for the correct
utilization of remote sensing products in relation to the specific
characteristics and complexity of the landscape and its changes;
development of specific guidelines relating to the adoption of sampling
design schemes (location and dimension);
development of standard procedures and guidelines for the acquisition
of existing reference lines for the preparation and production of land
use database;
development, at design level, of an optimal baseline using LCCS3 to
support the stratification or sample selection for any specific
agricultural landscape.
In addition, issues relating product cost and sustainability should be considered.
The costs of production are linked mainly to the source products and personnel.
As noted in this literature review, new incoming satellite products will soon be
available free of charge. This will significantly reduce costs of production and
56
increase the feasibility of generating new land cover databases. On the other
hand, personnel-related costs are more difficult to assess, as they are country-
specific. Section 8 above examines three case studies relating to developing
countries. Each case study sets out the relevant country’s specific conditions
and the time frame for generating the land cover database. The main parameters
affecting the production chain are:
involvement of international/national experts;
availability and technical background and capacity of national experts;
and
technical facilities.
To evaluate costs, guidelines on required pixel size according to landscape
characteristics are also important. The number of images to be processed and
interpreted will affect the cost of production, in the context of area accuracy or
signal variability.
The sustainability of a product is also linked to technical facilities and
personnel. Once a database is produced, increasing its life expectancy,
accuracy, and reliability requires its maintenance and updating. A core unit in a
country is also necessary and important for dissemination and correct
distribution among different end users.
Research should be performed on defining these parameters and modelling the
costs of production, maintenance, and dissemination in developing countries.
57
8
Conclusion and
Recommendations
This Literature Review presents several applications of remote sensing
technologies for agricultural statistics. However, the field is vast, and
researchers are directed to the publications listed in the References Section for
additional details.
Remote sensing is an important tool for collecting information to be integrated
with conventional statistical surveys. It has emerged that in recent years, a great
deal of technical progress in the field of remote sensing has been made. The
products available on the market demonstrate considerable improvements in
terms of resolution pixels, spectrum and revising time, and a considerable
reduction in market prices.
While this improved technology may introduce several possibilities to foster the
production of accurate, updated, and reliable agricultural databases, the
professional capacities of operators often does not develop as quickly. On the
other hand, efforts to produce ad hoc databases could exceed the immediate
scope of national projects. Therefore, two main implications may be
considered.
It is possible, advantageous, and beneficial to produce a new geo-
referenced cartographic land cover baseline, considering – among other
things – the new free-of-charge high-resolution remote sensing products
and new technology available; however, commitment is important and
is to be evaluated carefully. The land cover database can be part of the
core data as a fundamental baseline. It can be utilized to support
statistics as well as environmental analysis (e.g. land degradation).
If there are insufficient resources, then the implications of using
available products should be considered. Specific guidelines to evaluate
the products’ quality and accuracy should be defined, as well as the
implications of using medium-resolution databases in the various stages
of statistical analysis.
58
The applications of remote sensing for statistical analysis should be expanded,
especially in the following directions:
1. Expansion of applications and know-how in the use of remote sensing
for land cover production, survey and monitoring;
2. Creation and improvement on national capacity levels for the
management of remote sensing products, in close collaboration with
statistical units;
3. Defining the precision and accuracy of the remote sensing products in
relation to the landscape; and
4. Defining the relation between the land cover prepared with a specific
MMU and the sampling baseline.
The lessons learned in the Libya, Ethiopia, and Pakistan projects demonstrate
that different capacities and skills may be found onsite. However, it is also
possible to generate a land cover database with different modalities, resources
and timeframes.
59
ANNEX I
Literature review
This publication documents the main findings of a literature review that
examined various sources. The research was performed using the following
keys:
Period: 2000 to date
Keywords: Agricultural statistics; Area Sampling Frame; Remote Sensing;
Crop Acreage; Accuracy, Estimation, Terrain Mapping
The main catalogues and links consulted are listed below:
ScienceDirect
The following links are available from the FAO David Lubin Memorial
Library’s online catalogue
Scopus
Wiley Online Library
Springer Link
Taylor and Francis Online
Google and Google Scholar
The main scientific journals examined were:
Advances in Agronomy
Advances in Space Research
Agricultural Economics
Agriculture, Ecosystems and Environment
Agricultural Water Management
Applied Geography
Computer and Electronics in Agriculture
Computer and Sciences
ITC Journal
International Journal of Applied Earth Observation and Geoinformation
International Journal of Agricultural and Statistical Sciences
60
International Journal of Remote Sensing
International Society for Tropical Ecology
International Statistical Review
Journal of the Indian Society of Remote Sensing
Journal of Photogrammetry and Remote Sensing
Journal of Rural Studies
Precision Agriculture
Remote Sensing
Remote Sensing of Environment
Society of Agricultural Engineering
Transactions of the ASABE
Tropical Ecology
WSEAS Transactions on Environment and Development
The websites of the following agencies and institutions were consulted for
operative projects:
CGIAR
IFAD
IRRI
Crops for the Future Research Centre (CFF)
USDA
MARS
AGRIT
World Bank
World Bank Research Observer
Many useful links are available at:
Project EUCLID
Finally, the websites of universities and scientific venues and national statistical
bodies were explored to identify further relevant information.
61
References
AAIC. n.d. Agricultural Assessments International Corp. Homepage. Available
at: http://www.aaic.net/home.html. Last accessed on 15 January 2016.
Abkar, A.A., Sharifi, M.A., & Mulder, N.J. 2000. Likelihood-based image
segmentation and classification: a framework for the integration of expert
knowledge in image classification procedures. International journal of applied
earth observation and geoinformation, 2(2): 104-119.
AfDB Group, AU Commission, UNECA & FAO. 2011. Improving Statistics
for Food Security, Sustainable Agriculture, and Rural Development an Action
Plan for Africa 2011-2015.
AFSIS. n.d. ASEAN Food Security Information Services Homepage. Available
at: http://www.afsisnc.org/. Last accessed on 15 January 2016.
Ahamed, T., Tian, L., Jiang, Y., Zhao, B., Liu, H., & Ting, K.C. 2012.
Tower remote-sensing system for monitoring energy crops; image acquisition
and geometric corrections. Biosystems engineering, 112(2): 93-107.
Alonso, F.G., Soria, L.S. & Cuevas Gozalo, J.M. 1991. Comparing two
methodologies for crop area estimation in Spain using Landsat TM images and
ground-gathered data. Remote Sensing of Environment, 53(1): 29-35.
Amissah-Arthur, A. & Balstad Miller, R. 2002. Remote sensing applications
in African agriculture and natural resources: Highlighting and managing the
stress of increasing population pressure. Advances in Space Research, 30(11):
2411-2421.
Anderson, J.R., Hardy, E.E., Roach, J.T. & Witmer, R.E. 1976. A land use
and land cover classification system for use with remote sensor data.
Geological Survey Professional Paper No. 964. US Government Printing
Office: Washington, D.C.
Anderson, J.R. & Feder, G. 2004. Agricultural extension: Good intentions and
hard realities. The World Bank Research Observer, 19(1): 41-60.
62
Andrew, C.O. & Hildebrand, P.E. 1993. Applied agricultural research:
foundations and methodology. Westview Press, Inc.: Boulder, CO, USA.
Atzberger, C. 2013. Advances in remote sensing of agriculture: Context
description, existing operational monitoring systems and major information
needs. Remote Sensing 5(2): 949-981.
Balaghi, R., Tychon, B., Eerens, H., & Jlibene, M. 2006. Crop yield forecast
with NOOAA and SPOT-Vegetation data in Morocco. In Baruth, B., Royer, A.
& Genovese, G., ISPRS WG VIII/10 Workshop 2006: Remote Sensing Support
to Crop Yield Forecast and Area Estimates (pp. 1-2). ISPRS Archives –
Volume XXXVI-8/W48, 2006. Workshop held in Stresa, Italy, 30 November-1
December 2006.
Baltsavias, E. 2004. Object extraction and revision by image analysis using
existing geodata and knowledge: current status and steps towards operational
systems. ISPRS Journal of Photogrammetry and Remote Sensing 58(3): 129-
151.
Barrett, R., Crowther, P., Laurence, R., & Lincolne, R. 2000. Agricultural
Crop Identification using Spot and Landsat Images in Tasmania. International
Archives of Photogrammetry and Remote Sensing, 33(B7/1; PART 7): 133-139.
Barroso, C. & Monteiro, I. 2010. Monitoring Vegetation from Space.
EUMeTrain Publication. Available at: http://www.eumetrain.org/data/3/36/.
Last accessed on 15 January 2016.
Baruth, B., & Kucerab, L. Crop Masking–Needs for the Mars Crop Yield
Forecasting System. In Baruth, B., Royer, A. & Genovese, G., ISPRS WG
VIII/10 Workshop 2006: Remote Sensing Support to Crop Yield Forecast and
Area Estimates (pp. 3-6). ISPRS Archives XXXVI-8/W48, 2006. Workshop
held in Stresa, Italy, 30 November – 1 December 2006.
Bastiaanssen, W. G. & Ali, S. 2003. A new crop yield forecasting model based
on satellite measurements applied across the Indus Basin, Pakistan. Agriculture,
ecosystems & environment, 94(3): 321-340.
Bastiaanssen, W.G., Molden, D.J. & Makin, I.W. 2000. Remote sensing for
irrigated agriculture: examples from research and possible applications.
Agricultural water management, 46(2): 137-155.
63
Becker-Reshef, I., Vermote, E., Lindeman, M. & Justice, C. 2010. A
generalized regression-based model for forecasting winter wheat yields in
Kansas and Ukraine using MODIS data. Remote Sensing of Environment,
114(6): 1312-1323.
Benedetti, R. & Ciavattella, M. EBLUP estimates by using remotely sensed
data. In Baruth, B., Royer, A. & Genovese, G., ISPRS WG VIII/10 Workshop
2006: Remote Sensing Support to Crop Yield Forecast and Area Estimates (pp.
11-14) ISPRS Archives – Volume XXXVI-8/W48, 2006. Workshop held in
Stresa, Italy, 30 November-1 December 2006.
Benedetti, R., Catenaro, R. & Piersimoni, F. 2008. Generalized Software
Tools for Crop Area Estimates and Yield Forecast. In Baruth, B., Royer, A. &
Genovese, G., ISPRS WG VIII/10 Workshop 2006: Remote Sensing Support to
Crop Yield Forecast and Area Estimates (pp. 7-10). ISPRS Archives XXXVI-
8/W48, 2006. Workshop held in Stresa, Italy, 30 November-1 December 2006.
Benedetti, R., Piersimoni, F., Bee, M., & Espa, G. (eds). 2010. Agricultural
survey methods. John Wiley & Sons: Chichester, UK.
Berman, C.H. 2002. Assessment of landscape characterization and
classification methods. Center for Water and Watershed Studies. WCI
Cooperative Agreement No. PNW 99-9053-1-CA. University of Washington,
Seattle, USA.
Bhagia, N., Rajak, D.R. & Patel, N.K. 2011. Improvement in precision of
crop acreage estimation by remote sensing using frequency distribution based
stratification. Journal of the Indian Society of Remote Sensing, 39(2): 153-160.
Blaschke, T. 2010. Object based image analysis for remote sensing. ISPRS
Journal of Photogrammetry and Remote Sensing 65(1): 2-16.
Blaschke, T., Lang, S., Lorup, E., Strobl, J., & Zeil, P. 2000. Object-oriented
image processing in an integrated GIS/remote sensing environment and
perspectives for environmental applications. Environmental information for
planning, politics and the public, 2: 555-570.
Bolca, M., Kurucu, Y., Altınbaş, Ü, Esetlili, M.T., & Özen, F. 2012. A study
on the determination of electromagnetic reflection values of agricultural crop
pattern to improve accuracy of land use map by remote sensing technique.
Eurasian Journal of Soil Science, 1(2): 116-126.
64
Bolliger, F.P., Freitas, M.P.S., Antonaci, G.A. & Borges Cabrat, M.D. 2012.
Master Sampling Frames for Agricultural Surveys: Brazil overview. Paper
prepared for the High Level Stakeholders Meeting on the Global Strategy:
From Plan to Action - Expert Session: Master Sampling Frames for
Agricultural and Rural Statistics. 3-5 December 2012. Rome, FAO
Headquarters.
Bolton, D.K., & Friedl, M.A. 2013. Forecasting crop yield using remotely
sensed vegetation indices and crop phenology metrics. Agricultural and Forest
Meteorology, 173: 74-84.
Boryan, C.G., & Yang, Z. 2012. A new land cover classification based
stratification method for area sampling frame construction. In 2012 First
International Conference on Agro-Geoinformatics (Agro-Geoinformatics) (pp.
1-6). IEEE Publication: Piscataway, NJ, USA.
Bouma. J., de Vos, J.A., Sonneveld, M.P.W., Heuvelink, G.B.M.,
Stoorvogel, J.J. & Sparks, L.D. 2008. The Role of Scientists in Multiscale
Land Use Analysis: Lessons Learned from Dutch Communities of Practice.
Advances in Agronomy, 97: 175-237.
Brown, K., Hall, W.L., Snook, M.H., & Garvin, K. 2010. Sustainable land
development and restoration: decision consequence analysis. Butterworth-
Heinemann: Burlington, MA, USA.
Browne, W.P. & Cigler, A.J. 1990. U.S. Agricultural Groups: Institutional
Profiles. GreenWood Press: New York, USA.
Bush, J. & House, C. 1993. The area frame: a sampling base for establishment
surveys. In Proceedings of the International Conference on Establishment
Surveys (pp. 335-344). American Statistical Association Publication:
Alexandria, VA, USA.
Cai, R., Mullen, J.D., Wetzstein, M.E., & Bergstrom, J.C. 2013. The impacts
of crop yield and price volatility on producers’ cropping patterns: A dynamic
optimal crop rotation model. Agricultural Systems, 116: 52-59.
Cai, X. & Sharma B.R. 2010. Integrating remote sensing, census and weather
data for an assessment of rice yield, water consumption and water productivity
in the Indo-Gangetic river basin. Agricultural Water Management 97(2): 309-
316.
65
Canters, F. 1997. Evaluating the Uncertainty of Area Estimates Derived from
Fuuy Land-Cover Classification. Photogrammetric Engineering & Remote
Sensing, 63(4): 403-414.
Canty, M.J. 2009. Image Analysis, Classification and Change Detection in
Remote Sensing: With Algorithms for ENVI/IDL. 2nd ed. CRC Press: Boca
Raton, FL, USA.
Carfagna, E. 2001. Cost-effectiveness of remote sensing in agricultural and
environmental statistics. In Proceedings of the Conference on Agricultural and
Environmental Statistical Applications in Rome (CAESAR), 5-7 June, Vol. III
(pp. 617-627). ISTAT Publication: Rome.
Carfagna, E. 2012. Presentation on the priority topics to be addressed in 2013.
Delivered at Implementation of the Research Plan meeting of the Global
Strategy, 4 December 2012, FAO Headquarters, Rome.
Carfagna, E. & Gallego, F.J. 2005. Using remote sensing for agricultural
statistics. International Statistical Review/Revue Internationale de Statistique,
73(3): 389-404.
Carfagna, E. & Marzialetti, J. 2010. Continuous Innovation of the Quality
Control of Remote Sensing Data for Territory Management. In Erto, P. (ed.),
Statistics for Innovation, Statistical Design of “Continuous” Product
Innovation (pp. 172-188). Springer-Verlag Italia: Milan, Italy.
Cayrol, P., Poilvé, H., Lefèvre, V. & Duthil, P. 2006. Crop Area estimates
with MERIS data on Belgium and Poland. In Baruth, B., Royer, A. &
Genovese, G., ISPRS WG VIII/10 Workshop 2006: Remote Sensing Support to
Crop Yield Forecast and Area Estimates (pp. 19-24). ISPRS Archives XXXVI-
8/W48, 2006. Workshop held in Stresa, Italy, 30 November-1 December 2006.
Perry, C.R., Chhikara, R.S. & Spears, F.M. Modeling of Stratum Variance
for Use in Sample Allocation in Agricultural Area Frame Surveys. ASA Section
on Survey Research Methods.
Cherrill, A. 1997. A comparison of three landscape classifications and
investigation of the potential for using remotely sensed land cover data for
landscape classification. Journal of Rural Studies, 10(3): 275-289.
66
Chuvieco, E., & Huete, A. 2009. Fundamentals of satellite remote sensing.
CRC Press Inc.: Boca Raton, FL, USA.
Clevers, J.G. & van Leeuwen, H.J. 1996. Combined use of optical and
microwave remote sensing data for crop growth monitoring. Remote Sensing of
Environment, 56(1): 42-51.
Comber, A.J., Fisher, P.F., & Wadsworth, R.A. 2005. What is land cover?
Environment and Planning B Planning and Design, 32(2): 199-209.
Comber, A., Fisher, P., Brunsdon, C., & Khmag, A. 2012. Spatial analysis of
remote sensing image classification accuracy. Remote Sensing of Environment,
127: 237-246.
Cotter, J.J., & Tomczak, C. 1994. An image analysis system to develop area
sampling frames for agricultural surveys. Photogrammetric engineering and
remote sensing, 60(3): 299-306.
Central Statistics Authority of Ethiopia. 2011. Comparative Study between
Area Frame and List Frame for Agricultural Survey. Paper prepared for the
African Commission on Agricultural Statistics, Twenty-Second Session, 30
November – 3 December 2011, Addis Ababa.
Csornai, G., Wirnhardt, C., Suba, Z., Nádor, G., Tikász, L., Martinovich,
L., Kocsis, A., Zelei, G., László, I. & Bognár, E. CROPMON: Hungarian crop
production forecast by remote sensing. ISPRS Archives, 36(8): W48.
Dadhwal, V.K., Singh, R.P., Dutta, S., & Parihar, J.S. 2002. Remote sensing
based crop inventory: A review of Indian experience. Tropical Ecology, 43(1):
107-122.
Davies, C. 2009. Area Frame Design for Agricultural Surveys. RDD Research
Report, Research and Development Division, USDA-NASS, Fairfax, VA.
Unpublished internal document.
Defourny, P., Blaes, X., & Bogaert, P. 2007. Respective contribution of yield
and area estimates to the error in crop production forecasting. In Baruth, B.,
Royer, A. & Genovese, G., ISPRS WG VIII/10 Workshop 2006: Remote Sensing
Support to Crop Yield Forecast and Area Estimates (pp. 39-44). ISPRS
Archives XXXVI-8/W48, 2006. Workshop held in Stresa, Italy, 30 November-
1 December 2006.
67
Delincé, J. 2001. A European approach to Area Frame survey. Proceedin
Dheeravath, V., Thenkabail, P.S., Chandrakantha, G., Noojipady gs of the
Conference on Agricultural and Environmental Statistical Applications in Rome
(CAESAR). 5-7 June, Vol. III. Joint Research Centre of the Euro pean
Communities (JRC), 5-7 June 2001., P., Reddy, G.P.O., Biradar, C.M.,
Gumma, K. & Velpuri, M. 2010. Irrigated areas of India derived using
MODIS 500 m time series for the years 2001–2003. ISPRS Journal of
Photogrammetry and Remote Sensing, 65(1), 42-59.
Di Gregorio, A. 2013. Use of LCCS approach in agricultural survey.
Unpublished internal report, FAO.
Di Gregorio, A. & Leonardi, U. 2011. Formalization of the meaning of Corine
legend with an object oriented" language using LCCS3. FAO-GLCN Internal
unpublished report.
Di Gregorio, A. & Leonardi, U. 2010. LCCS Tutorial. FAO-GLCN
Publication. Available at: http://www.glcn.org/index_en.jsp.
do Prado, H.A., Barreto Luiz, A.J. & Filho, H.C. (eds). 2010. Computational
Methods for Agricultural Research: Advances and Applications: Advances and
Applications. IGI Global: Hershey, PA, USA.
Doraiswamy, P.C., Akhmedov, B., Beard, L., Stern, A., & Mueller, R. 2007.
Operational prediction of crop yields using MODIS data and products. In
Baruth, B., Royer, A. & Genovese, G., ISPRS WG VIII/10 Workshop 2006:
Remote Sensing Support to Crop Yield Forecast and Area Estimates (pp. 45-
50). ISPRS Archives – Volume XXXVI-8/W48, 2006. Workshop held in
Stresa, Italy, 30 November-1 December 2006.
Doraiswamy, P.C., Moulin, S., Cook, P.W., & Stern, A. 2003. Crop yield
assessment from remote sensing. Photogrammetric Engineering & Remote
Sensing, 69(6): 665-674.
Dutta, S., Parihar, J.S. & Justice, C. 2008. Workshop on Best Practices for
Crop Area Estimation with RS Data: Summary of Country Inputs. Presented at
the GEOSS Workshop on Crop Area Estimation with remote sensing, 5-6 June
2008. Barza Italy.
68
Available at
https://www.earthobservations.org/documents/cop/ag_gams/200707_01/Summa
ry_countries_AG.pdf.
Dutta, S., Sharma, S.A., Khera, A.P., Ajai, Yadav, M., Hooda, R.S.,
Mothikumar, K.E. & Manchanda, M.L. 1994. Accuracy assessment in cotton
acreage estimation using Indian remote sensing satellite data. ISPRS Journal of
Photogrammetry and Remote Sensing, 49(6): 21-26.
Duveiller, G. & Defourny, P. 2010. A conceptual framework to define the
spatial resolution requirements for agricultural monitoring using remote
sensing. Remote Sensing of Environment, 114(11): 2637-2650.
Dymond, J.R., Begue, A., & Loseen, D. 2001. Monitoring land at regional and
national scales and the role of remote sensing. International Journal of Applied
Earth Observation and Geoinformation, 3(2): 162-175.
Earl, R., Thomas, G., & Blackmore, B.S. 2000. The potential role of GIS in
autonomous field operations. Computers and electronics in agriculture, 25(1):
107-120.
Eiden, G. 2008. “Land-Cover and Land-Use Mapping.” Encyclopedia of Life
and Support System – EOLSS: Land Cover, Land Use and Soil Sciences.
UNESCO Publication: Paris.
El Nahry, A.H., Ali, R.R., & El Baroudy, A.A. 2011. An approach for
precision farming under pivot irrigation system using remote sensing and GIS
techniques. Agricultural Water Management, 98(4): 517-531.
Ene, L.T., Næsset, E., Gobakken, T., Gregoire, T.G., Ståhl, G., & Holm, S.
2013. A simulation approach for accuracy assessment of two-phase post-
stratified estimation in large-area LiDAR biomass surveys. Remote Sensing of
Environment, 133: 210-224.
Erto, P. (ed.). 2009. Statistics for Innovation: Statistical Design of
"Continuous" Product Innovation. Springer-Verlag Italia: Milan, Italy.
EU-JRC. 2015. Monitoring Agricultural Resources (MARS). Available at:
http://mars.jrc.ec.europa.eu/. Last accessed on 19 January 2016.
69
EU-AGINFRA. 2015. AGINFRA Homepage. Available at: http://aginfra.eu/.
Last accessed on 19 January 2016.
FAO. 1995. FAO Statistical Development Series No. 6: Conducting
Agricultural Censuses and Surveys. FAO Publication: Rome (also available at:
http://www.fao.org/economic/la-direccion-de-estadistica/censo-agropecuario-
mondial/conducting-of-agricultural-censuses-and-surveys/en/; last accessed 19
January 2016).
FAO. 1996. Multiple Frame Agricultural Surveys - Current surveys based on
area and list sampling method (Vol. 1). FAO Publication: Rome.
FAO. Statistics Division Women & Population Division. 1996. Guidelines for
the improvement of statistics on women: obtaining statistics from national
agricultural surveys in selected countries of the Near East. FAO Publication:
Rome.
Available at: http://www.fao.org/docrep/w0022e/w0022e00.htm#TopOfPage.
Accessed 20 January 2016.
FAO. 1998. Multiple Frame Agricultural Survey. FAO Publication: Rome.
FAO. 2010. Sampling Frames for Agricultural Censuses and Surveys. In A
system of integrated agricultural censuses and surveys. Vol. 1. FAO
Publication: Rome (also available at:
http://www.fao.org/docrep/009/a0135e/A0135E06.htm#ch10; last accessed on
19 January 2016).
FAO-RAP. 2002. General Status of the Food and Agriculture Statistics in Iran.
Unpublished internal report, FAO.
FAO and Central Statistics Authority of Ethiopia (CSA). 2012 Technical
cooperation programme - Terminal Report. Prepared for FAO by the
Government of Ethiopia. Unpublished document.
FAO & SUPARCO. 2012. Advanced training on Monitoring of crops through
Satellite technology. Joint FAO & SUPARCO Publication: Rome.
FAO & SUPARCO. 2012. How SUPARCO makes crop forecasts and
estimates based on integral use of RS data. Joint FAO & SUPARCO
Publication: Rome.
70
Farooque, A.A., Chang, Y.K., Zaman, Q.U., Groulx, D., Schumann, A.W.,
& Esau, T.J. 2013. Performance evaluation of multiple ground based sensors
mounted on a commercial wild blueberry harvester to sense plant height, fruit
yield and topographic features in real-time. Computers and electronics in
agriculture, 91: 135-144.
Fecso, R., Gardner, W., Hale, B., Johnson, V. & Pavlasek, S. 1982.
Construction of a Remotely Sensed Area Sampling Frame for Southern Brazil.
United States Department of Agriculture Statistical Reporting Service Staff
Report No. AGES 820526. USDA: Washington, D.C.
Ferreira, S.L., Newby, T. & du Preez, E. 2006. Use of Remote Sensing in
support of crop area estimates in South Africa. In Baruth, B., Royer, A. &
Genovese, G., ISPRS WG VIII/10 Workshop 2006: Remote Sensing Support to
Crop Yield Forecast and Area Estimates (pp. 51-52). ISPRS Archives –
Volume XXXVI-8/W48, 2006. Workshop held in Stresa, Italy, 30 November-1
December 2006.
Firouzabadi, P. 2001. Performance evaluation of supervised classification of
remotely sensed data for crop acreage estimation. In Geoscience and Remote
Sensing Symposium, 2001. IGARSS'01. IEEE 2001 International. Vol. 6. IEEE
Publication: Piscataway, NJ, USA.
Flores, L.A., & Martınez, L.I. 2000. Land cover estimation in small areas
using ground survey and remote sensing. Remote Sensing of Environment,
74(2): 240-248.
Fontana, D.C., Melo, R.W., Wagner, A.P.L., Weberb, E. & Gusso, A. 2006.
Use of remote sensing for crop yield and area estimates in the southern of
Brazil. International Archives Of the Photogrammetry, Remote Sensing and
Spatial Information Sciences, 36: 53-58.
Fritz, S., Massart, M., Savin, I. & Leo, O. The use of MODIS data in
southern Russia for crop acreage Estimations and inter-comparison of results
from various crop Acreage estimation methods In Baruth, B., Royer, A. &
Genovese, G., ISPRS WG VIII/10 Workshop 2006: Remote Sensing Support to
Crop Yield Forecast and Area Estimates (pp. 59-64). ISPRS Archives –
Volume XXXVI-8/W48, 2006. Workshop held in Stresa, Italy, 30 November-1
December 2006.
71
Fritz, S., Massart, M., Savin, I., Gallego, J. & Rembold, F. 2008. The use of
MODIS data to derive acreage estimations for larger fields: A case study in the
south-western Rostov region of Russia. International Journal of Applied Earth
Observation and Geoinformation 10(4): 453-466.
Gallego, F.J. 1995. Sampling frames of square segments. Report EUR 16317
EN. Office for Official Publications of the European Communities Publication:
Luxembourg.
Gallego, F.J. 1999. Crop area estimation in the MARS project. In Conference
on ten years of the MARS Project. Available at http://www.ec-
gis.org/docs/F18163/AREAESTMARS10Y.PDF
Gallego, F.J. 2005. Stratified sampling of satellite images with a systematic
grid of points. ISPRS Journal of Photogrammetry and Remote Sensing 59(6):
369-376.
Gallego, J. 2004. Remote sensing and land cover area estimation. Int. J Remote
Sensing Vol 25 No 15 pp 3019-3047 Teylor&Francis
Gallego, F. 2004. Review of the main remote sensing methods for crop area
estimates. ISPRS Archives XXXVI-8/W48 Workshop Proceedings: Remote
Sensing Support to Crop Yield Forecast and Area Estimates, Commissions
VIII, WG VIII/10.
Gallego, J. 2010. Best practices for crop area estimation with remote sensing.
Edited by Gallego J., Michaelsen J., Bossyns B., Fritz S., Ispra , June 5-6 2008.
GEOSS Community of Practice Ag 0703a JRC Scientific and Technical
Reports ISSN 1018-5593
Gallego, J. & Bamps, C. 2008. Using CORINE land cover and the point
survey LUCAS for area estimation. International Journal of Applied Earth
Observation and Geoinformation, 10(4): 467-475.
Gallego, J., Craig, M., Michaelsen, J., Bossyns, B. & Fritz, S. (eds). 2008.
Best practices for crop area estimation with Remote Sensing. JRC Scientific
and Technical Reports. EU-JRC Publication: Ispra, Italy.
Gallego, F.J. & Stibig, H.J. 2013. Area estimation from a sample of satellite
images: the impact of stratification on the clustering efficiency. International
Journal of Applied Earth Observation and Geoinformation, 22: 139-146.
72
Gebizlioðlu, Ö. L., Dedeþ, H. & Ömer Toprak, A. 1996. Small Area
Estimation in Turkey. Paper prepared for the 1996 Annual Research
Conference and Technology Interchange, Arlington, VA, USA. 17-21 March
1996.
Genovese, G. P., Fritz, S. & Bettio, M. 2006. A comparison and evaluation of
performances among crop yield forecasting models based on remote sensing:
Results from the geoland observatory of food monitoring. International
Archives of the Photogrammetry, Remote Sensing And Spatial Information
Sciences, 36: 71-77.
Gentle, J., Perry, C. & Wigton, W. 2006. Modeling Nonsampling Errors in
agricultural surveys. In Proceedings of the Survey Research Methods Section,
ASA, Session 415: Sample Survey Quality V. Available at:
http://www.amstat.org/sections/srms/Proceedings/. Accessed on 19 January
2016.
GEO Group on Earth Observations. 2011. Global Agricultural Monitoring
System of Systems. Available at:
http://www.earthobservations.org/cop_ag_gams.shtml. Accessed on 19 January
2016.
Geo Sense Sdn. Bhd. 2010. Utilizing Hyperspectral Imaging System in
unmanned aerial vehicle (UAV) for Agricultural/Palm Oil Analysis. Available
at: http://www.slideshare.net/flixmill/geo-sense-uav-service-unmanned-remote-
sensing. Accessed on 20 January 2016.
Gommes, R. 2008. Non-parametric crop yield forecasting, a didactic case study
for Zimbabwe. In Baruth, B., Royer, A. & Genovese, G., ISPRS WG VIII/10
Workshop 2006: Remote Sensing Support to Crop Yield Forecast and Area
Estimates (pp. 79-84). ISPRS Archives – Volume XXXVI-8/W48, 2006.
Workshop held in Stresa, Italy, 30 November-1 December 2006.
Goodchild, M.F. & Gopal, S. (eds). 1989. Global Agricultural Monitoring.
CRC Press: Boca Raton, FL, USA.
Grace, K., Husak, G.J., Harrison, L., Pedreros, D. & Michaelsen, J. 2012.
Using high resolution satellite imagery to estimate cropped area in Guatemala
and Haiti. Applied Geography Science Direct Elsevier, 32(2): 433-440.
73
Groten, S.M.E. 2001. Monitoring food security in Burkina Faso 1995-2000:
final report. Netherlands Remote Sensing Board (BCRS) Publication: Delft,
The Netherlands.
Guo, W., Zhao, C., Gu, X., Huang, W., Ma, Z., Wang, H., & Wang, D.
2011. Remote sensing monitoring of maize planting area at town level.
Transactions of the Chinese Society of Agricultural Engineering, 27(9): 69-74.
Hague, T., Tillett, N.D. & Wheeler, H. 2006. Automated crop and weed
monitoring in widely spaced cereals. Precision Agriculture, 7(1): 21-32.
Harrison, A.R. 2006. National land use database: land use and land cover
classification. Office of the Deputy Prime Minister Publication: London.
Hengl, T., Nikolić, M. & MacMillan, R.A. 2013. Mapping efficiency and
information content. International Journal of Applied Earth Observation and
Geoinformation, 22: 127-138.
Holland, T.T. 2001. Building & Using Area Sampling Frames for Agricultural
Censuses and Surveys. Presentation prepared for the Regional Workshop on
Sampling for Agricultural Surveys, 21 June 2001. Montevideo.
Hollister, J.W., Gonzalez, M.L., Paul, J.F., August, P.V. & Copeland, J.L.
2004. Assessing the Accuracy of National Land Cover Dataset Area Estimates
at Multiple Spatial Extents. Photogrammetric Engineering & Remote Sensing,
4: 405-414.
Hooda, R., Yadav, M. & Kalubarme, M. 2008. Wheat production estimation
using remote sensing data: an Indian experience. In Baruth, B., Royer, A. &
Genovese, G., ISPRS WG VIII/10 Workshop 2006: Remote Sensing Support to
Crop Yield Forecast and Area Estimates (pp. 85-90). ISPRS Archives –
Volume XXXVI-8/W48, 2006. Workshop held in Stresa, Italy, 30 November-1
December 2006.
Houseman, E.E. 1975. Area Frame Sampling in Agriculture (Vol. SRS No.
20). Statistical Reporting Service Publication: Washington, D.C.
Huang, Q., Zhou, Q.B. & Zhang, L. 2009. An overview of crop growing
condition monitoring in China agriculture remote sensing monitoring system. In
International Symposium on Photoelectronic Detection and Imaging 2009:
74
Laser Sensing and Imaging (pp. 73840N-73840N). Society of Photo-Optical
Instrumentation Engineers Publication: Bellingham, WA, USA.
Hunt, F.K. 2013. Applications of the USDA NASS CDL Based Automated
Stratification. Presentation prepared for the ASPRS Annual Conference, 24-28
March 2013. Baltimore, MD, USA.
Husak, G.J., Marshall, M.T., Michaelsen, J., Pedreros, D., Funk, C., &
Galu, G. 2008. Crop area estimation using high and medium resolution satellite
imagery in areas with complex topography. Journal of Geophysical Research:
Atmospheres (1984–2012), 113(D14).
India. Planning Commission. 2001. Report of the Working Group on
Agricultural Statistics for the Tenth Five Year Plan. New Delhi.
De Jong, S.M. 2000. Spatial statistics for remote sensing. International Journal
of Applied Earth Observation and Geoformation Academic Publishers, 2(1):
62-63.
Kalkhan, M.A. 2011. Spatial statistics: geospatial information modelling and
thematic mapping. CRC Press: Boca Raton, FL, USA.
Kudrat, M., Sharma, K.P., Tiwari, A.K., Kumar, P., Prabhakaran, B., &
Manchanda, M.L. 2000. Discrimination of newly planted and ratoon crops of
sugar cane using multidate IRS-1C liss III data: A knowledge based approach.
Journal of the Indian Society of Remote Sensing, 28(2-3): 179-185.
Lamas, A.C., Abreu, D., Arroway, P., Lopiano, K.K., & Young, L.J. 2010.
Modeling misclassification in the June Area Survey. In Proceedings of the
section on survey research methods JSM 2010. American Statistical
Association Publication: Alexandria, VA, USA.
Latham, J.S. & Cumani, R. 2012. Remote Sensing application for agricultural
monitoring (Case studies: national assessments). Presentation prepared for
Space for Agriculture and Food Security meeting, WFP Headquarters, 8-9
March 2012. Rome.
Latham, J.S. 1981-83. Final report of a pilot project on the utilization of
Remote Sensing and Field data for the production of Agricultural statistics on
crop acreage. Unpublished internal report, FAO.
75
Leclerc, G. & Hall, C.A. 2000. Remote sensing and land use analysis for
agriculture in Costa Rica. In Hall, C.A.S., Perez, C.L & Leclerc, G. (eds),
Quantifying sustainable development: the future of tropical economies (pp.
295-346). Academic Press: San Diego, CA, USA.
Lesser, V.M. & Kalsbeek, W.D. 1997. A comparison of periodic survey
designs employing multi-stage sampling. Environmental and Ecological
Statistics 4(2): 117-130.
Levin, N. 1999. Fundamentals of remote sensing. 1st Hydrographic Data
Management Course, International Maritime Academy (IMO), Trieste, Italy.
Li, D., Liu, J., Zhou, Q., Wang, L. & Huang, Q. 2012. Influence of scale
variation on regional precision of rape acreage remote sensing monitoring. In
Agro-Geoinformatics (Agro-Geoinformatics), 2012 First International
Conference on (pp. 1-5). IEEE Publication.
Li, R., Li, C., Dong, Y., Feng, L., Wang, J., Yang, X. & Pan, Y. 2011.
Assimilation of Remote Sensing and Crop Model for LAI Estimation Based on
Ensemble Kaiman Filter. Agricultural Sciences in China, 10(10): 1595-1602.
Li, Q., Wu, B., Jia, K., Dong, Q., Eerens, H., & Zhang, M. 2011. Maize
acreage estimation using ENVISAT MERIS and CBERS-02B CCD data in the
North China Plain. Computers and electronics in agriculture, 78(2): 208-214.
Lillesand, T., Kiefer, R.W. & Chipman, J. 2014. Remote sensing and image
interpretation. John Wiley & Sons: Hoboken, NJ, USA.
Liu, J., Liu, M., Tian, H., Zhuang, D., Zhang, Z., Zhang, W. & Deng, X.
2005. Spatial and temporal patterns of China’s cropland during 1990–2000: an
analysis based on Landsat TM data. Remote Sensing of Environment, 98(4):
442-456.
Liu, X., Sun, R., Wu, F., Hu, B. & Wang, W. 2010. Land-cover classification
for Henan Province with time-series MODIS EVI data. Transactions of the
Chinese Society of Agricultural Engineering, 26(Supplement 1): 213-219.
Lunetta, R.S., Knight, J.F., Ediriwickrema, J., Lyon, J.G. & Worthy, L.D.
2006. Land-cover change detection using multi-temporal MODIS NDVI data.
Remote sensing of environment, 105(2): 142-154.
76
Lunetta, R.S., Shao, Y., Ediriwickrema, J. & Lyon, J.G. 2010. Monitoring
agricultural cropping patterns across the Laurentian Great Lakes Basin using
MODIS-NDVI data. International Journal of Applied Earth Observation and
Geoinformation, 12(2): 81-88.
Madana, M. 2002. Improving land use survey method using high resolution
satellite imagery. International Institute for Geo Information Science and Earth
Observation (ITC) (M.Sc. Thesis).
Marker, D.A. & Stevens Jr, D.L. 2009. Sampling and Inference in
Environmental Surveys. Handbook of Statistics 29A. Sample Surveys: Design,
Methods and Applications (Chapter 19). Elsevier: Oxford, UK.
Maxwell, S.K., Nuckols, J.R., Ward, M.H. & Hoffer, R.M. 2004. An
automated approach to mapping corn from Landsat imagery. Computers and
Electronics in Agriculture, 43(1): 43-54.
McBratney, A., Whelan, B., Ancev, T. & Bouma, J. 2005. Future directions
of precision agriculture. Precision Agriculture, 6(1): 7-23.
McCracken, S.D., Brondízio, E., Moran, E.F., Nelson, D., Siqueira, A. &
Rodriguez-Pedraza, C. 1998. The Use of Remote Sensing and GIS in the
Collection of Survey Data on Households and Land-Use: Example from the
Agricultural Frontier of the Brazilian Amazon. In Krug, T., Rudorff, B.F.T. &
Freitas, U.M., Anais IX Simpósio Brasileiro de Sensoriamento Remoto, Santos,
Brasil, 11-18 setembro 1998 (pp. 45-52). INPE Publication, Brazil.
Meentemeyer, V. & Box, E. 1987. Scale effects in landscape studies. In
Turner, M. (ed.), Landscape Heterogeneity and Disturbance, pp. 15-34.
Springer-Verlag: New York.
Meng, J., Wu, B., Li, Q., Niu, L., Zhang, F. & Du, X. 2011. Integrated
provincial crop monitoring system using remote sensing. Transactions of the
Chinese Society of Agricultural Engineering, 27(6): 169-175.
Michelson, D.B., Liljeberg, B.M. & Pilesjö, P. 2000. Comparison of
algorithms for classifying Swedish landcover using Landsat TM and ERS-1
SAR data. Remote Sensing of Environment, 71(1): 1-15.
Milne, B. 1992 Spatial aggregation and neutral models in fractal landscapes.
American Naturalist, 139: 32-57.
77
Mokgokolo, J. 2011. Challenges experienced with the demarcation of
enumeration areas for Census 2011. Unpublished manuscript. Pretoria:
Geography Division, Statistics South Africa.
Moody, A. 1998. Using landscape spatial relationships to improve estimates of
land-cover area from coarse resolution remote sensing. Remote Sensing of
Environment, 64(2): 202-220.
Moody, A. & Woodcock, C.E. 1995. The influence of scale and the spatial
characteristics of landscapes on land-cover mapping using remote sensing.
Landscape Ecology, 10(6): 363-379.
Moody, A. & Woodcock, C.E. 1996. Calibration-based models for correction
of area estimates derived from coarse resolution land-cover data. Remote
Sensing of Environment, 58(3): 225-241.
Moran, M.S., Inoue, Y. & Barnes, E.M. 1997. Opportunities and limitations
for image-based remote sensing in precision crop management. Remote Sensing
of Environment, 61(3): 319-346.
Muad, A.M., & Foody, G.M. 2010. Super-Resolution Mapping of Landscape
Objects from Coarse Spatial Resolution Imagery. Paper prepared for the
GEOgraphic Object-Based Image Analysis Conference 2010, Ghent, Belgium.
29 June-2 July 2010.
Mukherjee, J., Gebru, G., Sood, A., Mahey, R.K., Bal, S.K., Singh, H. &
Sidhu, P.K. 2010. Wheat yield and acreage prediction using LISS-III and
AWiFS sensors data of Indian remote sensing satellite of Rupnager district of
Punjab, India. Rivista Italiana Di Telerilevamento, 42(3): 115-127.
Mushtaq, K., & Dawson, P.J. 2002. Acreage response in Pakistan: a co-
integration approach. Agricultural Economics, 27(2): 111-121.
Narciso, G. 2008. Crop Area Estimates with Radarsat Feasibility Study in the
Toscana Region, Italy. In Baruth, B., Royer, A. & Genovese, G., ISPRS WG
VIII/10 Workshop 2006: Remote Sensing Support to Crop Yield Forecast and
Area Estimates (pp. 103-104). ISPRS Archives – Volume XXXVI-8/W48,
2006. Workshop held in Stresa, Italy, 30 November-1 December 2006.
78
NASA. Department of Agriculture.
Available at: http://history.nasa.gov/presrep00/pages/usda.html. Last accessed
on 20 January 2016.
Nematzadeh, M. 2001. Estimation of area with area frame sampling in
Hamadan province of Iran. In Geoscience and Remote Sensing Symposium,
2001. IGARSS'01. IEEE 2001 International, vol. 7. IEEE Publication:
Piscataway, NJ, USA.
Nusser, S. & House, C.O. 2009. Sampling, Data Collection, and Estimation in
Agricultural Statistics. In Handbook of Statistics Sample Surveys: Design,
Methods and Applications (Chapter 18). Elsevier: Oxford, UK.
Oza, M., Pandya, M. & Rajak, D.R. 2008. Evaluation and use of
Resourcesat-I data for agricultural applications. International Journal of
Applied Earth Observation and Geoinformation, 10(2): 194-205.
Ozdogan, M. 2010. The spatial distribution of crop types from MODIS data:
Temporal unmixing using Independent Component Analysis. Remote Sensing
of Environment, 114(6): 1190-1204.
Ozdogan, M. & Woodcock, C.E. 2006. Resolution dependent errors in remote
sensing of cultivated areas. Remote Sensing of Environment, 103(2): 203-217.
Ozdogan, M., Yang, Y., Allez, G. & Cervantes, C. 2010. Remote sensing of
irrigated agriculture: Opportunities and challenges. Remote Sensing, 2(9): 2274-
2304.
Pan, Y., Zhang, J. & Shen, K. 2011. Crop area estimation from UAV transect
and MSR image data using Spatial sampling method: A simulation experiment.
Procedia Environmental Sciences, 7: 110-115.
Pan, Y., Zhang, J., Zhou, W., Zhao, J. & Yu, X. 2011. Methodology of the
Crops Acreage Estimation Using Remote Sensing and Survey Sampling in
National Statistical System of China. In Agro-Geoinformatics (Agro-
Geoinformatics), 2012 First International Conference on. IEEE Publication.
Pandit, D.S., Murthy, Y.V.K. & Jayaraman, V. 2006. Identification of
sugarcane and onion crops using digital image processing of multidate,
multisensor high-resolution satellite data. In Kuligowski, R.J., Parihar, J.S &
Saito, G., SPIE Proceedings Vol. 6411: Agricluture and Hydrology
79
Applications of Remote Sensing (pp. 641109 ff.). International Society for
Optics and Photonics (SPIE).
Pax-Lenney, M., & Woodcock, C.E. 1997. The effect of spatial resolution on
the ability to monitor the status of agricultural lands. Remote Sensing of
Environment, 61(2): 210-220.
Pelorosso, R., Leone, A., & Boccia, L. 2009. Land cover and land use change
in the Italian central Apennines: A comparison of assessment methods. Applied
Geography, 29(1): 35-48.
Petersen, G.W., Bell, J. C., Mcsweeney, K., Nielsen, G.A. & Robert, P.C.
1995. Geographic Information Systems in Agronomy. Advances in Agronomy,
55: 67-111.
Picciotto, R. & Anderson, J.R. 1997. Reconsidering agricultural extension.
The World Bank Research Observer, 12(2): 249-259.
Piersimoni, F., Postiglione, P. & Benedetti, R. 2014. Spatial Sampling for
Agricultural Data. In Mecatti, F., Conti, P.L. & Ranalli, M.G. (eds),
Contributions to Sampling Statistics (pp. 179-198). Springer: Berlin.
Potgieter, A.B., Lawson, K., & Huete, A.R. 2013. Determining crop acreage
estimates for specific winter crops using shape attributes from sequential
MODIS imagery. International Journal of Applied Earth Observation and
Geoinformation, 23: 254-263.
Powell, R.L., Matzke, N., De Souza, C., Clark, M., Numata, I., Hess, L.L. &
Roberts, D.A. 2004. Sources of error in accuracy assessment of thematic land-
cover maps in the Brazilian Amazon. Remote Sensing of Environment, 90(2):
221-234.
Pradhan, S. 2001. Crop area estimation using GIS, remote sensing and area
frame sampling. International Journal of Applied Earth Observation and
Geoinformation, 3(1): 86-92.
Pratesi, M. & Salvati, N. Sampling Strategies and Multifunctionality in
Agricultural Surveys.In Atti del Convegno Intermedio SIS 2005 Statistica e
Ambiente, 21-23 September 2005. Messina, Italy.
80
Prishchepov, A.V., Radeloff, V.C., Dubinin, M. & Alcantara, C. 2012. The
effect of Landsat ETM/ETM+ image acquisition dates on the detection of
agricultural land abandonment in Eastern Europe. Remote Sensing of
Environment, 126: 195-209.
Priya, S. & Shibasaki, R. 2001. National spatial crop yield simulation using
GIS-based crop production model. Ecological Modelling, 136(2): 113-129.
RADI & CAS. Cropwatch website.
Available at: http://www.cropwatch.com.cn/htm/en/index.shtml.
Accessed on 18 January 2016.
Rai, A., Sahoo, P.M. & Ahmad, T. Geo-informatics in Agricultural Research
and Development: An IASRI Perspective. In IASRI…An Era of Excellence.
1959-2000. IASRI Publication: New Delhi.
Raju, P.V., Sai, M.S. & Roy, P.S. 2008. In-season time series analysis of
Resourcesat-1 AWiFS data for estimating irrigation water requirement.
International Journal of Applied Earth Observation and Geoinformation, 10(2):
220-228.
Rao, J. 2004. Small area estimation with applications to Agriculture.
Agricultural Survey Methods, 9: 139-147.
Ray, S.S. & Dadhwal, V.K. 2001. Estimation of crop evapotranspiration of
irrigation command area using remote sensing and GIS. Agricultural Water
Management, 49(3): 239-249.
Ren, J., Chen, Z., Zhou, Q. & Tang, H. 2008. Regional yield estimation for
winter wheat with MODIS-NDVI data in Shandong, China. International
Journal of Applied Earth Observation and Geoinformation, 10(4): 403-413.
Robinson, G.M., Gray, D.A., Healey, R.G. & Furley, P.A. 1984. Developing
a geographical information system (GIS) for agricultural development in
Belize, Central America. Applied Geography, 9(2): 81-94.
Rozenstein, O. & Karnieli, A. 2011. Comparison of methods for land-use
classification incorporating remote sensing and GIS inputs. Applied Geography,
31(2): 533-544.
81
Ruecker, G.R., Shi, Z., Mueller, M., Conrad, C., Ibragimov, N., Lamers,
J.P.A. & Dech, S.W. Cotton Yield Estimation in Uzbekistan Integrating
Modis, Landsat Etm+ and Field Data. In Baruth, B., Royer, A. & Genovese, G.,
ISPRS WG VIII/10 Workshop 2006: Remote Sensing Support to Crop Yield
Forecast and Area Estimates (pp. 123-128). ISPRS Archives – Volume
XXXVI-8/W48, 2006. Workshop held in Stresa, Italy, 30 November-1
December 2006.
Ruiz, L.A., Recio, J.A., Fernández-Sarría, A. & Hermosilla, T. 2011. A
feature extraction software tool for agricultural object-based image analysis.
Computers and Electronics in Agriculture, 76(2): 284-296.
Sabins, F. 1986. Principles of Remote Sensing. W.H. Freeman: New York,
USA.
Savin, I. 2008. Crop Yield Prediction with SPOT VGT in Mediterranean and
Central Asian Countries. In Baruth, B., Royer, A. & Genovese, G., ISPRS WG
VIII/10 Workshop 2006: Remote Sensing Support to Crop Yield Forecast and
Area Estimates (pp. 129-134). ISPRS Archives – Volume XXXVI-8/W48,
2006. Workshop held in Stresa, Italy, 30 November-1 December 2006.
Seelan, S.K., Laguette, S., Casady, G.M., & Seielstad, G.A. 2003. Remote
sensing applications for precision agriculture: A learning community approach.
Remote Sensing of Environment, 88(1): 157-169.
Sher, F. & Ahmad, E. 2008. Forecasting wheat production in Pakistan. Lahore
Journal Of Economics, 13(1): 57-85.
Shrivastava, R.J. & Gebelein, J.L. 2007. Land cover classification and
economic assessment of citrus groves using remote sensing. ISPRS journal of
photogrammetry and remote sensing, 61(5): 341-353.
Slaughter, D., Chen, P. & Curley, R. 1999. Vision guided precision
cultivation. Precision Agriculture 1(2): 199-217.
Smiatek, G. 1995. Sampling thematic mapper imagery for land use data.
Remote Sensing of Environment, 52(2): 116-121.
Srivastava, M. 2011. The Global Strategy to Improve Agricultural Statistics:
An Implementation Plan for Latin America the Caribbean. Paper prepared for
82
the 10th Meeting of the Executive Committee of the Statistical Conference of
Americas of ECLAC. 8 April 2011. Havana.
Stehman, S.V. 2004. A critical evaluation of the normalized error matrix in
map accuracy assessment. Photogrammetric Engineering & Remote Sensing,
70(6): 743-751.
Stehman, S.V. & Czaplewski, R.L. 1998. Design and analysis for thematic
map accuracy assessment: fundamental principles. Remote Sensing of
Environment, 64(3): 331-344.
Stehman, S., Olofsson, P., Woodcock, C., Friedl, M., Sibley, A., Newell, J.,
Sulla-Menashe, D. & Herold, M. 2010. Designing a reference validation
database for accuracy assessment of land cover. Paper prepared for the
Accuracy 2010 Symposium, 20-23 July 2010. Leicester, UK.
Steininger, M.K., Godoy, F. & Harper, G. 2009. Effects of systematic
sampling on satellite estimates of deforestation rates. Environmental Research
Letters, 4(3): 034015.
Strand, G.H. & Aune-Lundberg, L. 2012. Small-area estimation of land
cover statistics by post-stratification of a national area frame survey. Applied
Geography, 32(2): 546-555.
Strand, G. H. & Rekdal, Y. Area frame survey of land resources Rep. No.
03/2006. NIJOS Publication: Ås, Norway.
System for Environmental and Agricultural Modelling; Linking European
Sciences and Society. 2009. SEAMLESS Homepage. Available at:
http://www.seamless-ip.org/. Accessed 20 January 2016.
Thenkabail, P., Lyon, J.G., Turral, H. & Biradar, C. (eds). 2009. Remote
sensing of global croplands for food security. CRC Press: Boca Raton, FL,
USA.
Tian, Y., Wu, B., Zhang, L., Li, Q., Jia, K. & Wen, M. 2011. Opium poppy
monitoring with remote sensing in North Myanmar. International Journal of
Drug Policy, 22(4): 278-284.
83
Trant, M., Brunelin, S. & Nyekanyeka, M.J. 2010. FAO Mission Report
Support to the Malawi Agricultural Statistics System - Main Finding and
Recommendations. Unpublished internal document, FAO.
Treitz, P. & Rogan, J. 2004. Remote sensing for mapping and monitoring
land-cover and land-use change – an introduction. Progress in planning, 61(4):
269-279.
Trujillo, L., Merchan, O. & Ferraz, C. 2006. Dual Frame Implementation for
the Colombian National Agricultural Survey. Paper prepared for the 58th World
Statistical Congress, 2011, Dublin (Session CPS035). International Statistical
Institute.
Tsiligirides, T.A. Remote sensing as a tool for agricultural statistics: a case
study of AFS methodology in Hellas. Computers and Electronics in
Agriculture, 20(1): 45-77.
Ünal, E., Mermer, A., Urlaa, Ö., Ceylan, N., Yıldız, H., Aydoğdu, M. 2006.
Pilot Study on wheat area estimates in Turkey. In Baruth, B., Royer, A. &
Genovese, G., ISPRS WG VIII/10 Workshop 2006: Remote Sensing Support to
Crop Yield Forecast and Area Estimates (pp. 135-138). ISPRS Archives –
Volume XXXVI-8/W48, 2006. Workshop held in Stresa, Italy, 30 November-1
December 2006.
UNSC. 2010. Report of the Friends of the Chair on Agricultural Statistics.
Document for the Forty-first Session of the Statistical Commission, 23-26
February 2010. New York, USA. USA. National Research Council.
Committee on Remote Sensing for Agricultural Purposes. Agricultural Board.
1970. Remote Sensing: With Special Reference to Agriculture and Forestry.
National Academy of Sciences Publication: Washington, D.C.
Van Niel, T.G. & McVicar, T.R. 2004. Determining temporal windows for
crop discrimination with remote sensing: a case study in south-eastern
Australia. Computers and Electronics in Agriculture, 45(1): 91-108.
Van Velthuizen, H. & Hizsnyik, G.F.E. 2013. The African Farming Systems
Update Project Farming systems and food security: Priorities for science and
policy under global change. IIASA Publication: Laxenburg, Austria.
Verbeiren, S., Eerens, H., Piccard, I., Bauwens, I. & Van Orshoven, J.
2008. Sub-pixel classification of SPOT-VEGETATION time series for the
84
assessment of regional crop areas in Belgium. International Journal of Applied
Earth Observation and Geoinformation, 10(4): 486–497.
Vibhute, A.D. & Gawali, B.W. 2013Analysis and modeling of agricultural
land use using remote sensing and geographic information system: A review.
International Journal of Engineering Research and Applications (IJERA), 3(3):
81-91.
Villalobos, Á.G. 2006. Sample Survey Statistics Teaching: an Almost
Worldwide Problem on Teaching Agricultural Survey Methods. Paper prepared
for the 7th International Conference on Teaching Statistics, 2-7 July 2006.
Salvador, Bahia, Brazil.
Villalobos, Á.G. & Wigton, W.H. 2011. On Applying Area And Multiple
Frame Sampling Methods In A Wide Range Of Baseline Agricultural And Rural
Survey Programmes. Paper prepared for the 58th World Statistical Congress, 21-
26 August 2011. Dublin. International Statistical Institute.
Villalobos, A.G. & Wallace, M.A. 1996. Multiple frame agricultural surveys.
Vol. 1: Current surveys based on area and list sampling methods. FAO
Statistical Development Series 7. FAO Publication: Rome.
Villalobos, A.G. & Wallace, M.A. 1996. Multiple frame agricultural surveys.
Vol. 2: Agricultural surveys programmes based on area frame or dual frame
(area and list) sample design. FAO Statistical Development Series 10. FAO
Publication: Rome.
Walker, P.A. & Mallawaarachchi, T. 1998. Disaggregating agricultural
statistics using NOAA-AVHRR NDVI. Remote Sensing of Environment, 63(2):
112-125.
Wang, D., Zhou, Q., Chen, Z. & Liu, J. 2012. Optimal design of spatial
sampling schemes for winter wheat sown area estimation. In Agro-
Geoinformatics (Agro-Geoinformatics), 2012 First International Conference on
(pp. 1-6). IEEE Publication: Piscataway, NJ, USA.
Wang, J., Jiang, C., Hu, M., Cao, Z., Guo, Y., Li, L. & Meng, B. 2013.
Design-based spatial sampling: Theory and implementation. Environmental
Modelling & Software, 40: 280-288.
85
Wang, S. 2006. A statistic method of crop acreage based on image recognition.
In Intelligent Systems Design and Applications, 2006. ISDA'06. Sixth
International Conference on, vol. 2. IEEE Publication: Piscataway, NJ, USA.
Weis, M., Müller, S., Liedtke, C.E. & Pahl, M. 2005. A framework for GIS
and imagery data fusion in support of cartographic updating. Information
Fusion, 6(4): 311-317.
Wigton, W., Srivastava, A., Zekaria, S., Mudeir, Y., Mossa, Y., Tadesse,
G., Yigezu, B., Tariku, A., Beyene, H. & Abdi, J. 2009. Agricultural
Statistical Methodology in Ethiopia. Central Statistical Agency (Ethiopia)
Publication: Addis Abeba.
Winton, W. & Central Statistical Authority (CSA). 2009. Agriculture
Statistical Methodology in Ethiopia. FAO internal Report Unpublished
World Bank, FAO & UNSC. 2012. Action Plan of the Global Strategy to
improve Agricultural and Rural Statistics.
World Bank, FAO & UNSC. 2010. Global Strategy to Improve Agricultural
and Rural Statistics. Report No. 56719-GLB.
Wu, B. & Li, Q. 2004. Crop area estimation using remote sensing on two-
stage stratified sampling. Paper prepared for the XXth ISPRS, Congress, 12-23
July 2004, Istanbul, Turkey.
Wu, B. & Li, Q. 2012. Crop planting and type proportion method for crop
acreage estimation of complex agricultural landscapes. International Journal of
Applied Earth Observation and Geoinformation, 16: 101-112.
Wu, W., Shibasaki, R., Yang, P., Tan, G., Matsumura, K.I., & Sugimoto,
K. 2007. Global-scale modelling of future changes in sown areas of major
crops. Ecological Modelling, 208(2): 378-390.
Wu, Q., Sun, L. & Wang, F. 2010. The applications of 3S in operational
monitoring system of main crops acreage in China. In World Automation
Congress (WAC), 2010 (pp. 319-324). IEEE Publication: Piscataway, NJ, USA.
Xiao, X., Liu, J., Zhuang, D., Frolking, S., Boles, S., Xu, B. & Li, C. 2003.
Uncertainties in estimates of cropland area in China: A comparison between an
86
AVHRR-derived dataset and a Landsat TM-derived dataset. Global and
Planetary Change, 37(3): 297-306.
Xie, Y., Sha, Z. & Yu, M. 2008. Remote sensing imagery in vegetation
mapping: a review. Journal of Plant Ecology, 1(1): 9-23.
Xu, W., Tian, Y. & Huang, J. 2005. Use of artificial neural networks for
estimating crop acreage from MODIS data in large-scale. In Communications,
Circuits and Systems, 2005. Proceedings. 2005 International Conference on,
vol. 2. IEEE Publication.
Xu, W., Tian, Y., Qing, J., Huang, J. & Zhang, Y. 2005. A method of
estimating crop acreage in large-scale by unmixing of MODIS data. In
Proceedings of SPIE. Society of Photo-Optical Instrumentation Engineers
Publication.
Yadav, I.S., Rao, N.S., Rawal, R.D., Srinivasan, V.R., Sujatha, N.T.,
Bhattacharya, C. & Elango, S. 2002. Acreage and production estimation of
mango orchards using Indian Remote Sensing (IRS) satellite data. Scientia
Horticulturae, 93(2): 105-123.
Yang, C., Everitt, J.H., & Bradford, J.M. 2007. Airborne hyperspectral
imagery and linear spectral unmixing for mapping variation in crop yield.
Precision Agriculture, 8(6): 279-296.
Yang, C., Everitt, J.H., & Murden, D. 2011. Evaluating high resolution
SPOT 5 satellite imagery for crop identification. Computers and Electronics in
Agriculture, 75(2): 347-354.
Yang, C., Liu, T.X., & Everitt, J.H. 2008. Estimating cabbage physical
parameters using remote sensing technology. Crop Protection, 27(1): 25-35.
Zhang, F., Zhu, Z., Pan Y., Hu T. & Zhang J. 2010. Application of Remote
Sensing technology in Crop Acreage and Yield Statistical Survey in China.
Paper prepared for the Meeting on the Management of Statistical Information
Systems (MSIS 2010), 26-29 April 2010. Daejeon, Republic of Korea.
Zhang, N., Wang, M. & Wang, N. 2002. Precision agriculture –d a worldwide
overview. Computers and electronics in agriculture, 36(2): 113-132.
87
Zhang, R. & Zhu, D. 2011. Study of land cover classification based on
knowledge rules using high-resolution remote sensing images. Expert Systems
with Applications, 38(4): 3647-3652.
Zhang, X., Qin, Y. & Qin, F. 2013. Remote sensing estimation of planting
area for winter wheat by integrating seasonal rhythms and spectral
characteristics. Transactions of the Chinese Society of Agricultural
Engineering, 29(8): 154-163.
Zhao, J., Shi, K. & Wei, F. 2007. Research and Application of Remote Sensing
Technology in Chinese Agricultural Statistics. Beijing, Department of Rural
Surveys, National Bureau of Statistics of China.
Zhu, S., Zhou, W., Zhang, J. & Shuai, G. 2012. Wheat acreage detection by
extended support vector analysis with multi-temporal remote sensing images. In
Agro-Geoinformatics (Agro-Geoinformatics), 2012 First International
Conference on (pp. 1-4). IEEE Publication: Piscataway, NJ, USA.
