iEMSs 2016 Conference · July 10-14, 2016, Toulouse, FRANCE. ... The organizers of the iEMSs 2016 Conference want to express thanks to the actively ... University of Exeter Medical

iEMSs 2016 Conference Environmental modelling and software

for supporting a sustainable future

Proceedings │ Volume 5 │ Pages 1275-1335 8th International Congress on Environmental

Modeling and Software (iEMSs) July 10–14, 2016 Toulouse, France

Proceedings of the 8th International Congress on Environmental Modelling and Software (iEMSs)

July 10-14, 2016, Toulouse, FRANCE.

How to cite the full proceedings:

Sauvage, S., Sánchez-Pérez, J.M., Rizzoli, A.E. (Eds.), 2016. Proceedings of the 8th International Congress on Environmental Modelling and Software, July 10-14, Toulouse,

FRANCE. ISBN: 978-88-9035-745-9

How to cite an individual paper:

Author, A., Author, B., Author, C…, 2016. This is the title of your paper. In: Sauvage, S., Sánchez-Pérez, J.M., Rizzoli, A.E. (Eds.), 2016. Proceedings of the 8th International

Congress on Environmental Modelling and Software, July 10-14, Toulouse, FRANCE. ISBN: 978-88-9035-745-9

Peer Review:

Each paper has been peer reviewed by at least two independent reviewers with possible outcomes of reject, revise, and accept.

Foreword

The organizers of the iEMSs 2016 Conference want to express thanks to the actively

involved organizations and individuals, especially for their preparation and dedication

in coordinating a successful conference. We would also like to thank the Scientific

Committee for their work in proposing sessions and coordinating the reviews. We

would like to thank the Local and the Scientific Committee for their support in

preparing the conference agenda and for allowing the scientists and researchers

around the globe to participate and exchange valuable scientific knowledge at this

conference.

We would like to give a special thanks to Polytechnical National Institute for their

efforts in hosting the iEMSs Community. This conference would not have been

possible without the generous support and contributions of our wonderful Sponsors.

The following Conference Proceedings contain papers covering a variety of topics in

Environmental Modelling and Software supporting sustainable future and related to:

- Integrated Environmental Modeling: a multidisciplinary approach for innovative

applications (Volume 1),

- Methods in Environmental Modelling including optimization, sensitivity,

uncertainty issues, model development and modelling frameworks (Volume

2),

- Processing environmental information including data mining, machine

learning, GIS, remote sensing (Volume 3),

- Environmental Modelling and Software supporting decision making and

engaging the public (Volume 4) and

- Poster presentations for all topics (Volume 5).

The Conference Organizers hope you enjoy the conference and continue to view

these iEMSs gatherings as a positive opportunity for our international research

community to share the latest innovations in Environmental Modelling and Software.

Convenors

Dr. Sabine SAUVAGE Research Engineer, CNRS, France

Sabine Siméoni-Sauvage holds a PhD from the Institut National Polytechnique, Toulouse University, FRANCE. She is currently a Research Engineer at the National Center for Scientifical Research (CNRS) in the National Institute of Ecology and Environment. She is currently assigned to ECOLAB laboratory at Toulouse, and has been working on modeling contaminant transfer in river water systems for the past 16 years.

Her research interests are focused on the adaptation and development of models that describe the bio-physical interactions between flows, biology and chemistry processes involved in biogenic elements and contaminant transfer in rivers at different time and space scale. More specifically, she aims to integrate modeling between interfaces zones (ex: water/land, water/sediment) and specific buffer zones (ex: wetlands) in the dynamic of element transfer at large scale.

Dr. José Miguel SANCHEZ PEREZ Research Director, CNRS, France

Jose Miguel Sánchez-Pérez, has a Ph. D. (1992) in Hydrogeochemistry from the University of Strasbourg (France); M.S. (1985) in Geoloy from the University of the Basque Country in Spain. He is currently a research director assigned to ECOLAB Laboratory (CNRS - Université Paul Sabatier - Institut National Polytechnique de Toulouse) in bio-geochemical functioning of buffers zones.

José Miguel Sánchez-Pérez studies pollutant transport in hydrologic systems, using extensive field data and modeling. He specializes in the functioning of wetlands, riparian zones, and groundwater systems, with particular interests in the modeling of catchment-scale pollutant transport, to predict how ecosystem functions will change under various climate change scenarios.

Local Organizing Committee

Sabine Sauvage, Co-Chair José Miguel Sanchez Perez, Co-Chair Daniel P. Ames, Organizer and Web-Site Karina Gibert, Organizer Andrea Rizzoli, Organizer Alexey Voinov, Organizer Gene Whelan, Rajbir Parmar, EPA Sponsorship Sarva Pulla, Michael Scott, Web-Site Estelle Henry, Chloe Palomba, Secretary and administratition Susan Cuddy, Jim Ascough, (Papers and presentation student Awards) Stefan Reis, (Poster Session) Ann van Griesven, Tatiana Filatova, Martin Volk (Co-convenors of Stream A) James Ascough, Ioannis Athanasiadis, Georgii Alexandrov (Co-convenors of Stream B), Miquel Sànchez-Marrè, Karina Gibert, Marina Erechtchoukova (Co-convenors of Stream C), Alexey Voinov, Suzanne Pierce, Olivier Barreteau (Co-convenors of Stream D), Foreign Student Volunteers: Andrea Kaim, Myriam Grillot, David Rivas, Gunseli Erdem, Francesca Recanati, Argyrios Samourkasidis, Emma Cutler, Benjamin Kayatz, Thibault Hallouin, Pau Gimeno, Mohammed Ebrahim Azari, Xavier Corrales, Jae-Young Lee, Han Su, Anastasia Lobanova, Sarva Pulla. Local Student Volunteers : Clement Fabre, Amine Zettam, Marie Lefranck, Sebastien Haunold, Youen Grusson.

Scientific Committee

Sabine Sauvage (CNRS, France)

José-Miguel Sanchez Perez (CNRS, France)

Alexey Voinov (ITC, Netherlands)

Andrea-Emilio Rizzoli (IDSIA-USI/SUPSI, Switzerland)

Ann van Griensven (UNESCO, Belgium)

Anthony J Jakeman, (Australian National University, Australia)

Ari Jolma (Finland)

Dan Ames (Brigham Young University, USA)

Gary Polhill (Surrey University, UK)

Gene Whelan (EPA, USA)

Georgii Alexandrov (Russian Academy of Sciences, Russia)

Giorgio Guariso (Politecnico di Milano, Italy)

Ioannis Athanasiadis (The Netherlands)

James C. Ascough (USA)

Joaquim Comas (University of Girona, Spain)

Karina Gibert (Polytechnic University of Catalonia, Spain)

Kristina Voigt (Germany)

Marcello Donatelli (CRA, Italy)

Marina Erechtchoukova (Canada)

Mark Borsuk (Darthmouth University, USA)

Martin Volk (UFZ-Helmholtz-Centre for Environmental Research Leipzig, Germany)

Miquel Sànchez-Marrè (Polytechnic University of Catalonia, Spain)

Mohammad Ebrahim Banihabib (Iran)

Olivier Barreteau (France)

Peter A. Khaiter (Canada)

Jiri Hrebicek (Masaryk University, Czech Republic)

Rajbir Parmar (USA)

Ralf Denzer (University of the Saarland, Germany)

Rüdiger Schaldach (Germany)

Schimak Gerald (AIT, Austria)

Sophie Ricci (France)

Stefan Reis (Centre for Ecology and Hydrology, UK)

Tatiana Filatova (ITC Twente, Netherlands)

i

Poster Session 1275

Chan Sung Oh, Junghoon Choi, Euisang Jeong and Jin‐Yong Choi. Numerical Simulation considered Agricultural Non‐Point Source Pollution in Saemangeum River Basin, Korea

1276

Kinga Mazurek. Research on the landscape stability (case study) 1277

Carmen Capilla. Generalized regression networks to predict air quality 1278

Carmen Capilla. Neural networks data mining in an air quality database 1279

Nabil Bougherira, Azzedine Hani, Dounia Nechem and Hicham Chaffai. Modelling of ground water contamination, plants case of El Hadjar, N East of Algeria

1287

Hicham Chaffai, Hakim Khelfaoui, Azzedine Hani, Nabil Bougherira and Larbi Djabri. HYDRODYNAMIC AND HYDRODISPERSIVE MODELING: IMPACT OF INDUSTRIAL ACTIVITIES ON THE WATERS OF THE PLAIN OF BERRAHAL (ANNABA)

1295

Martin Paegelow and María Teresa Camacho Olmedo. About the Use of Multiple Training Dates for Land Change Modeling

1303

Eva Carmina Serrano Balderas, Laure Berti‐Equille, Maria Aurora Armienta Hernandez and Jean‐Christophe Desconnets. Water Quality Data Analytics

1304

Myriam Grillot, Jonathan Vayssières, Benoit Gaudou, Alassane Bah and Dominique Masse. An agent‐based model to analyze the ecological functioning of agro‐silvo‐pastoral landscapes in West Africa

1305

Taher Osman, Hichem Omrani, Amin Tayyebi and Prasanna Divigalpitiya. Using Decision support system to identify the suitable land‐parcels for hosting solar energy modules in Cairo Region

1306

Ottar Tamm and Toomas Tamm. Comparison of load estimation methods for total nitrogen in Estonia

1307

Domenico Ventrella and Pasquale Garofalo. Impact of climate change on winter durum wheat cultivated in Southern Italy: effect of extreme weather events

1308

Thierry Agbanou, Simon Allagbe, Brice Tente, Vincent Orekan and Martin Paegelow. Analysis of the intrinsic relationship between the spatial radiometry of Vitellaria paradoxa (C.F. Gaertn.) and soil in Sudano‐Guinean zone in northwest Benin

1309

Kristinn Edvardsson and Cezara Pastrav. An agent based model of the effects of fishing policy on the Icelandic coastal communities

1310

Juan Durango, Mehdi Saqalli, Laurence Maurice, Nicolas Maestripieri and Arnaud Elger. Heterogeneous data mining for a semi‐quantitative risk assessment of oil contamination from multiple‐sources in the Ecuadorian Amazon

1311

Marc Jaxa‐Rozen and Jan Kwakkel. An evaluation of tree‐based ensemble methods for sensitivity analysis in environmental models

1312

Jean‐Francois Mas, Azucena Pérez Vega, Araceli Andablo Reyes, Miguel Angel Castillo Santiago and Alejandro Flamenco Sandoval. Assessing Modifiable Areal Unit Problem (MAUP) Effects in the Analysis of Deforestation Drivers Using Local Models

1313

Kamil Nešetřil and Jaroslav Nosek. Dealing with uncertainty: multiple simple groundwater models

1319

Emma Cutler, Mark Borsuk, David Lutz and Matthew Ayres. An agent based model of southern pine beetle forest damages

1320

ii

Jean‐Francois Mas. Combining Geographically Weighted and pattern‐based models to simulate deforestation processes

1321

William Chapotat and Lionel Houssou. An agent‐based model of the Amazonian forest colonisation and oil exploitation: the Oriente study case

1328

International Environmental Modelling and Software Society (iEMSs) 8th International Congress on Environmental Modelling and Software

Toulouse, France, Sabine Sauvage, José-Miguel Sánchez-Pérez, Andrea Rizzoli (Eds.) http://www.iemss.org/society/index.php/iemss-2016-proceedings

-

Poster session

Convenor : Stefan Reis(a,b)

(a) NERC Centre for Ecology & Hydrology, Bush Estate, Penicuik, EH26 0QB, United Kingdom (b) University of Exeter Medical School, Knowledge Spa, Truro, TR1 3HD, United Kingdom

1275



Numerical Simulation considered Agricultural Non-Point Source Pollution in Saemangeum River Basin,

Korea

Chansung Oh1* Junghoon Choi2 Euisang Jeong3 Jinyong Choi4 1*Korea Rural Community Corporation ([email protected])

2Korea Rural Community Corporation ([email protected]) 3HydroCore([email protected])

Kunsan National University([email protected]) Abstract: A huge area for agricultural land is distributed around Saemangeum River Basin, in which the Non-Point Source Pollution (NPS) discharged from the area can adverse impact on the water quality of Saemangeum Reservoir. Thus, in order to improve the water quality of Saemangeum Reservoir, the management of agricultural NPS is required. To evaluate the Saemangeum Water Quality Improvement Measures (SWIMs) considered agricultural NPS, we have examined the most suitable agricultural NPS reduction measures for the Saemangeum River Basin, and construct its scenarios for paddy field. A series of watershed modelling has also performed to assess its efficiency according to the scenarios. To achieve these objectives, a river basin model, STREAM (Spatio-Temporal River-basin Eco-hydrological Analysis Model), was calibrated and validated first, and managed dike high and managed fertilization as BMPs (Best management practices) were applied to the paddy fields in the control area of agricultural NPS and total Saemangeum River Basin. The simulation results considered managed dike high show that inflow water amount, inflow pollutant loads of TN and TP to Saemangeum Reservoir decreased 6.2%, 7.0% and 10.0%, respectively. While the results for managed fertilization show that they decreased 0.0%, 0.6% and 2.9%, respectively. Keywords: Saemangeum River Basin; Non-point source pollution; Best Management practices;STREAM; Reduction measures

1276



-

Localization of stable landscape areas in Upper Silesian Coal Basin (southern Poland)

Kinga Mazurek1

1 Department of Physical Geography, Faculty of Earth Sciences, University of Silesia, ul. Będzińska 60, 41-200 Sosnowiec, Poland ([email protected])

Abstract: Analytical approach of the local landscape stability is usually based on natural or partially transformed landscapes whereas anthropogenic landscapes are addressed less frequently. The aim of this study was to indicate areas characterised by the stability of landscape as well as to determine which natural or social-economic conditions influence the stability of environment. The research area of 72,40 sq km was located in the Upper Silesian Coal Basin (southern Poland). Analysis of changes in land cover for the period of first half of 19th century up to 21st century was based on archival topographic maps and an orthophotomap. Data was vectorized accordingly in ArcGIS 10.3 by Esri. IDRISI Andes by Clark Labs was used to determine the Cohen's kappa coefficient. The level of similarity among given maps fluctuated in most cases from 0.81 to 1.00. Selected landscape metrics were established by means of Fragstats v4. An intersect of vector layers of the land cover was performed. The verification of surface area units (37375) identified by the above mentioned means was followed by the analysis of combinations of land cover changes (3024). It identified 3524 surface area units that did not change its cover since 19th century up to modern times. Eventually 10 largest units were characterised in detail. The results were juxtaposed with the urban planning documents. The actual land covers were compared to the future investments – 40% of the analysed units are planned to be transformed by means of introducing new functions (housing, service and production). Achieved results are crucial in supporting decisions on spatial planning of urban and communal areas, for they accurately determine the predisposition of a given area, its value and the possibility of its use. Keywords: landscape, stability, landscape changes, Upper Silesia Coal Basin, southern Poland

1277



Generalized regression networks to predict air quality

Carmen Capilla

Polytechnic University of Valencia, Spain ([email protected])

Abstract: This paper studies the performance of generalized regression networks to predict air quality. These variants of radial basis functions are evaluated using a database of hourly observations of nitrogen dioxide and nitric oxide, monitored in two sites of an urban area. The sampling stations are traffic urban and background suburban. Traffic intensities are observed in locations close to the air quality sites. In the station where the air pollution average levels are higher, the monitoring device also registers meteorological parameters. In the other site, the climatological data are provided by the National Institute of Meteorology. The generalized regression networks outputs, are the one day ahead predictions of hourly pollutants levels. The study period is three years. The networks accuracy is measured using the root mean square error, the mean absolute error and the correlation coefficient. The MATLAB Neural Networks Toolbox is used. The input vector distance from the neuron’s weight vector has been studied with values from 0,1 to 0,45. The best correlations coefficients between predictions of nitric oxide and nitrogen dioxide and their observations, are 0,62 and 0,65 in one site. The lowest root mean square errors are 44,75 and 20,72, with mean absolute errors 25,77 and 15,89. The network inputs are meteorological parameters, traffic level, and seasonal components. In the other site, the predictions have a correlation coefficient 0,52 for nitric oxides, and a root mean square error that results 33,24, with the same inputs. The mean absolute error is 17,03 using nitrogen dioxide concentrations as one of the predictors of nitric oxides. The nitrogen dioxide forecasts give a root mean square 30,79 and a mean absolute error 25,01. Keywords: Urban air quality; neural networks; air pollution level prediction.

1278



Neural networks data mining in an air quality database

Carmen Capilla

Polytechnic University, Applied Statistics and Operations Research and Quality Department, Camino de Vera s/n, 46022-Valencia, Spain, email:[email protected]

Tel: +34 963877490, Fax: +34 963877499

Abstract: This work analyses a database of air pollutants and meteorological observations. The objective is to apply data mining procedures, in order to predict hourly ozone levels. Neural networks are used with three forecast intervals: one, eight and 24 hours in advance. The samples are monitored in an urban area of the east coast of the Iberian Peninsula. Five sampling points are considered. They are included in the air quality monitoring system of the city. The networks performance is evaluated using the root mean square error, the mean absolute error and the correlation coefficient. These criteria distributions are measured by randomly resampling the database 1000 times with replacement. Predictors are nitrogen oxides, ozone, meteorological variables and daily and weekly seasonal cycles. The predictions with the best accuracy are for the one hour ahead interval. The results show that the worst forecasts are for the eight hours in advance. Forecasts computed with a traffic urban site database, give the best correlations coefficients (95% interval [0,9604, 0,9662]) for one hour in advance. Predictions using a background suburban station database, have mean absolute errors (95% interval [5,9133, 6,5151]) and root mean square errors (95% interval [8,1954, 9,0475]), which compared with the other four study locations, indicate a best performance of the data mining techniques. Keywords: Data mining; neural networks; air quality; prediction; urban area

1 INTRODUCTION

This work analyses a database of environmental observations. They have been sampled in an urban area (Valencia) of the east coast of the Iberian Peninsula. Five locations are considered in the study. The objective is to forecast ozone (O3) observations. O3 has adverse effects on vegetation and human health because of its strong oxidant properties, and its relation with other factors. Therefore, legislation on air quality improvement provides guidelines and methods to monitor air quality (European Union Directive, 2008). The rationale for the application of forecasting methods on this occasion, is the prediction of pollution levels where the alert thresholds are exceeded, for the design of emission reduction strategies. Neural networks have been shown to be effective models to forecast temporal observations. Gardner and Dorling (1998) presented some bibliographical references on their use with atmospheric data. These models can be applied to analyse linear or non-linear interactions of atmospheric observations (Gardner, 1999). Gomez-Sanchis et al (2006) studied a small dataset and predicted daily concentrations of ozone in a small town close to Valencia. They indicated that good results were obtained using meteorological data and pollution levels as predictors in neural networks models. Daily and monthly seasonalities were descriptively analysed by Castell-Balaguer et al (2012) in two locations of the Valencia urban area. In this study, the data mining techniques to predict O3 levels are neural networks. Three prediction intervals are investigated: one, eight and 24 hours. The networks outputs are hourly O3 values. The inputs are nitrogen oxides (NOx), O3, meteorological parameters and daily and weekly cycles. The databases contain observations from three years. Their number of records range from 18184 to 22708. The forecast results are evaluated using three criteria: the root mean square error, the mean absolute error and the correlation coefficient. Prediction reliability and accuracy is often measured by the variance of the prediction method (Glymour et al. 1997). Tibshirani (1996) compared several methods to estimate the standard error of

1279

C. Capilla / Neural networks data mining in a air quality database

predicted values from neural networks applied to a small air quality dataset. Bootstrap gave the most accurate estimates. In this paper the network performance is investigated by bootstrap resampling, to estimate the criteria standard deviations, mean values, and statistical distributions. 2 THE DATABASE Table 1 shows the sampling locations. The monitoring network data quality control is done by the Mediterranean Centre for Environmental Studies Foundation (Castell-Balaguer et al. 2010).

Table 1. Sampling locations.

Site

Type

Variables

(hourly data)

Coordinates

Altitude (m)

1 Traffic urban Air quality, meteorological

0º20’34’’W 39º27’29’’N

7

2 Traffic urban Air quality 0º24’30’’W 39º28’52’’N

15

3 Traffic urban Air quality meteorological

0º22’36’’W 39º27’29’’N

11

4 Traffic urban Air quality 0º20’15’’W 39º28’47’’N

7

5 Background suburban

Air quality 0º22’10’’W 39º28’46’’N

11

The meteorological observations in site 3 are wind speed (WS), wind direction (WD), temperature (T), relative humidity (RH), pressure (P) and solar radiation (SR). WS and WD are available in site 1. The highest percentage of records with missing values in any variable, is in site 5, and ranges from 29,9%% to 30,88%. Only complete records are used. The resulting database sizes are in Table 2, for each prediction interval and location. The observation period is three years, and hourly averages are used.

Table 2. Database sizes.

Site One hour Eight hours 24 hours 1 20971 20902 29827 2 21130 21042 20985 3 21776 21646 21530 4 22708 22527 22377 5 18436 18322 18184

The highest annual mean ozone level was observed in site 2, in the third observation year. Figure 1 shows the box plots of the pollutant values in that year. Figure 2 represents the means and standard deviations for each week day. 3 NEURAL NETWORKS The predictors are nitrogen oxides and ozone concentrations, meteorological variables, and daily and weekly cycles in the form of sine and cosine components. The output is ozone forecast, and is estimated with 5 to 30 neurons in the hidden layer. The hyperbolic tangent is the transfer function of this layer. The network is trained with the Levenberg-Marquard algorithm. The databases are splitted into the training, validation and test sets. The three sets are randomly resampled 1000 times with replacement. The forecast performance is assessed using the 1000 values of the correlation coefficient between observations and predictions, and the prediction error that provides the other two criteria.

1280


Figure 1. Box plots of hourly ozone level (Site 2, third observation year).

Figure 2. Means and standard deviations of hourly ozone level (Site 2, third observation year).

4 RESULTS AND DISCUSSION The average correlation coefficient between observation and predictions is represented in Figures 3 to 5. The average root mean square errors and mean absolute errors are plotted in Figures 6 to 8. The data set results are shown. The averages have been computed with the 1000 resampling results for each prediction interval and site, with number of neurons in the hidden layer from 5 to 30. The MATLAB Neural Network Toolbox has been used. For the 1 hour forecast interval, the best mean correlation coefficient has been observed in site 2 (Figure 3). However, the average root mean square and mean absolute errors have been lower in site 5 (Figure 6). In sites 2 and 3 the results have highest average values of the correlations with the eight and 24 hour interval (Figures 4 and 5). The site 1 predictions eight and 24 hours ahead, have smaller average root mean square errors (Figures 7 and 8). On the other hand, the means absolute errors have been on average smaller in site 3. The three model assessment criteria have shown on average, a better agreement between observations and predictions computed with the shortest prediction interval. Table 3 gives the summary statistics of the correlations coefficients, with the number of

ozon

e

hour0 1 2 3 4 5 6 7 8 9 10 1112 13 14 15 16 17 18 19 20 21 22 23

0

40

80

120

160

ozon

e

Mon

day

Tues

day

Wed

nesd

ay

Thur

sday

Frid

ay

Sat

urda

y

Sun

day

49

51

53

55

57

59

61

1281


neurons in the hidden layer which had the best results. Table 4 and 5 contains the same information on the other performance parameters.

Figure 3. Average correlation coefficients (cc) of the ozone predictions in the five sites (prediction interval 1h).


num neurons

Variablescc site 1cc site 2cc site 3cc site 4cc site 5

0 5 10 15 20 25 300,93

0,94

0,95

0,96

0,97

num neurons


0 5 10 15 20 25 300,64

0,67

0,7

0,73

0,76

0,79

1282



Table 3. Mean, standard error, and first and third quartiles of the 1000 correlation coefficient values.

Site Mean Standard error First quartile Third quartile 2

(Interval 1 hour, num neurons=15)

0,9634 0,002028

0,963334

0,963462

2 (Interval 8 hours, num neurons=5)

0,7718

0,011962

0,76603

0,77968


0,7711

0,010184

0,76441

0,77833


0,8087

0,026641

0,80471

0,81473


0,8134

0,006457

0,8094

0,81761

Table 4. Mean, standard error, and first and third quartiles of the 1000 root mean square errors.



8,6123

0,319353

8,46165

8,75055


19,0069

1,31576

18,696

19,1696


18,3076

0,783062

18,0035

18,512

num neurons


0 5 10 15 20 25 300,63

0,67

0,71

0,75

0,79

0,83

1283


Table 5. Mean, standard error, and first and third quartiles of the 1000 mean absolute errors.



6,2078

0,152657

6,1022

6,30895


16,6153

0,457188

16,3525

16,814


15,2134

0,317289

14,997

15,417

Figure 6. Average root mean square errors (rmse) and mean absolute errors (mae). Prediction interval 1h.

5 CONCLUSIONS The main goal of this paper was to compare the performance of neural networks for data mining in databases of air pollutants and meteorological observations. The observations have been monitored in an urban area of the east coast of the Iberian Peninsula. Five locations are considered, and the databases associated to them, contain observations that have been hourly aggregated. The data mining models are neural networks, and allow to study the prediction power of pollutants levels, meteorological parameters and seasonal cycles. The models output is hourly ozone, and three forecast intervals are investigated: one, eight and 24 hours. The networks performance is evaluated using the root mean square error, the mean absolute error and the correlation coefficient. These criteria distributions are obtained by randomly resampling the database 1000 times with replacement. The best accuracy forecasts are for one hour, but the evaluation measurements are worse for eight hours. The computations with a traffic urban site database, give correlations coefficients with a 95% interval equal to [0,9604, 0,9662]). Predictions using a background suburban station database, have mean absolute errors (95% interval [5,9133, 6,5151]) and root mean square errors (95% interval

num neurons

Variablesrmse site 1rmse site 2rmse site 3rmse site 4rmse site 5mae site1mae site 2mae site 3mae site 4mae site 5

0 5 10 15 20 25 308,6

9,6

10,6

11,6

12,6

6,2

6,4

6,6

6,8

7

7,2

1284


[8,1954, 9,0475]). These values are associated to one hour ahead predictions. The data mining techniques had worse results in the other study locations.



REFERENCES Castell-Balaguer, N., Téllez, L., Mantilla, E., 2012. Daily, seasonal and monthly variations in ozone

levels recorded at the Turia river basin in Valencia (Eastern Spain). Environ. Sci. Poll. Res. 19, 3461-3480.

num neurons

Variablesrmse site 1rmse site 2rmse site 3rmse site 4rmse site 5mae site 1mae site 2mae site 3mae site 4mae site 5

0 5 10 15 20 25 3019

20

21

22

23

24

14

15

16

17

18

19

num neurons

Variablesrmse site 1rmse site 2rmse site 3rmse site 4rmse site 5mae site 1mae site 2mae site 3mae site 4mae site 5

0 5 10 15 20 25 3018

19

20

21

22

23

14

15

16

17

18

1285


Castell-Balaguer, N., Téllez, L., Luján, A., Mantilla, E., 2010. Informe Final Previozono 2010.

Programa Especial de Vigilancia de las Concentraciones de Ozono Troposférico en la Comunidad Valenciana. http://www.agricultura.gva.es/documents/20549779/161530536/informe10.pd (last accessed 28.01.16)

Gardner, M.W., 1999. The Advantages of Artificial Neural Networks and Regression Tree Based Air

Quality Models (Chapters 1 and 2). A dissertation submitted to the School of Environmental Sciences of the University of East Anglia (part of the requirements for the degree of Doctor of Philosophy), UK.

Gardner, M.W., Dorling, S.R., 1998. Artificial neural networks (the multilayer perceptron)- A review of

applications in the atmospheric sciences. Atmos. Environ. 32, 2627-2636. Glymour, C., Madigan, D., Pregibon, D., Smyth P., 1997. Statistical themes and lessons for data

mining. Data Mining and Knowledge Discovery 1, pp. 11-28. Kluwer Academic Publishers, The Netherlands.

Gómez-Sanchis, J., Martín-Guerrero, J.D., Soria-Olivas, E., Vila-Francés, J., Carrasco, J.L., del Valle-

Tascón, S., 2006. Neural networks for analysing the relevance of input variables in the prediction of tropospheric ozone concentration. Atmos. Environ. 40, 6173-6180.

Official Journal of the European Union, 2008. Directive 2008/50/EC of the European Parliament and

of the Council of 21 May2008 on Ambient Air Quality and Cleaner Air for Europe. http:// eur-lex.europa.eu/legal-content/AUTO/?uri=CELEX:32008L0050&qid=145371475043&rid=1 (last accessed 25.01.16)

Tibshirani, R., 1996. A comparison of some error estimates for neural networks models. Neural Comp.

8, 152-163.

1286



-

Modelling of ground water contamination, plants case of El Hadjar, N East of Algeria

Bougherira Nabil- Hani Azzedine - Nechem Dounia - Chaffai Hicham

(Water resources and sustainable development Laboratory. Geology department, Annaba University, BP 12, 23000 Annaba, Algeria.)

([email protected])

Abstract: In order to assess the quality of water and its evolution in the aquiferous system surrounding the Meboudja River, heavily contaminated by the rejections of steel industry (plants of El Hadjar, Annaba N East of Algeria) as well as domestic solid waste, monitoring of the latter was carried out during the period 1999-2011. Surface water as well as groundwater samples were analyzed. Showed three (03) main sources of contamination. a) - A natural (mineral) contamination of water caused by dissolution of evaporates and carbonated formations present in the area. b) - Agricultural contamination caused by excessive use of chemical fertilizers (high continent of nitrates) c) – Pollution caused by interaction of the heavily contamined Meboudja river and the aquiferous system Treated and untreated effluents are directly discharged in the river which acts as a diffuse source of contamination all along its path. High chloride content (up to 1400 mg.l-1), electric conductivity (exceeding 6000 micro S.cm-1) occurring mainly in wells situated downstream, in addition to important iron and manganese concentrations are observed. The aquifer parameters are estimated by using pumping test data at several wells, modeling of ground water flow and mass transport was carried out using visual MODFLOW software. It shows clearly that during the period under consideration the river the river steam and its interaction of the aquiferous system, accounts for the faster migration of contaminants in the over exploited area. Keywords: Meboudja River – contamined - electric conductivity

1. Introduction

The natural chemical quality of groundwater depends mainly on their origin, the nature of the alluvium and rocks that store water, as well as physic-chemical conditions of the environment. The considered zone, that is expected to be the third industrial node of Algeria, has experienced an intensification of demography (higher than 800000 inhabitants) (RGPH, 2001) and economy (higher than 150 industrial units). Environmental problems such as air and water pollution could seriously set back these economical and urban developments. Indeed, several hundreds of tons/day of solid and liquid wastes are dumped in environment lacking any treatment. This uncontrolled dumping has negative effects that are clearly identified such as nauseous smells, smoke generation, water and soil pollution (Debièche et al., 2003, Hani, 2003). These have been the subject of this study, given their important role in changing the chemical form of the elements. To characterize the environmental conditions, it is possible to emphasize the study of the dominant presence and / or absence of certain species, including: metals such as iron and manganese, insoluble in oxidizing conditions.

2. Study area

The studied area is situated in the lower Seybous plain (NE Algeria) upstream of Meboudja river (Figure 1). The northern limit of the zone is constituted by the metamorphic basement, whereas

1287

N. Bougherira et al. / Modelling of ground water contamination, plants case of El Hadjar, N East of Algeria

the other limits are open limits which are in continuity with the shallow aquifer of the low-lying Seybous plain.

Figure.1 Location map of study area: 1: Undifferentiated Quaternary, 2: Ancient alluvium, 3: Numidian

sandstone or clay, 4: Metamorphic formation. River Meboudja is characterized by a permanent flow in winter. Its alimentation comes from

rain waters, and drainage of the lake Fetzara (the discharge reaches 16 m3s-1). During the summer, the water inflows are mainly the lake outputs (the flow rate ranges between 1 and 5 m3s -1). The river receives also urban contributions upstream, such as domestic sewage waters. The aquifer reservoir is developed on a clayey substratum. The aquifer formations are represented by 70 loamy sand and 30 clays (Japanese Agency of International Office of Survey for the Countryside Development, in Zenati, 1999). The average permeability ranges between 10-3 and 10-4 m s-1. The Mediterranean climate is of type with an annual rainfall of 650 mm, a mean temperature of 18°C and high atmospheric humidity. The dominating wind direction (Northwest-Southeast) blows from the studied area towards the region of Drean. The effective infiltration is about 15% of the total rainfall that is 100mm per year, which infiltrates through waste, soil and finally to the ground water (Hani, 2003, Debièche, 2002). 3. Material and methods

Many monthly surveys of the piezometric level and geochemical analysis have been monitored. The analyses are carried out on a network of 30 wells in the plain of Meboudja. Some sampled wells are used by the neighbouring population for daily drinking, irrigation and animal alimentation. The grounds were taken according to a direction south-north.

Mediterranean S

ea

936 938 940 942 944 946 948 950 952 954390

392

394

396

398

400

402

404

406

408

410

412

EDOUGHALGERIA

0 5 km

ANNABA

: 1 : 2 : 3 : 4

WadiMeboudja

Study area

El-Hadjar

Sidi-Amar

Hadjar diss

Pont-Bouchet

Boukhadra

INDUSTRIES(MITTAL-STEEL)

Sey

bous

e R

iver

1288


The temperature (T), electrical conductivity (EC) and pH were measured in situ using a multiparameter WTW set (Multiline P3 PH/LF SET), an Ox meter (WTW) with an oxygen probe (Cell Ox 325) for the measurement of dissolved oxygen. The concentration of chloride (Cl-), sulphate (SO4

2-

), calcium (Ca2+), magnesium (Mg2+), carbonates (HCO3-), were determined using the volumetric

method (AFNOR, 1987). Nitrate (NO3-) and ammonium (NH4+) were analysed by colorimetric method using spectrophotometer (Spectronic 20 D). The heavy metals (Fe, Mn) were determined using atomic absorption spectrophotometer (Unicam 929 AA Spectrometer). In the present study, groundwater monitoring has been taken up for effective assessment through understanding of hydrogeology, geology and water-chemistry of the watershed. The collected basic data is used for the preparation of the groundwater flow and mass transport model for quantitative assessment of impact of contaminant migration in the watershed. To determine the groundwater velocity distribution is used to analyze advective and dispersive transport to determine contaminant migration in the area. 4 Results and discussion TDS values (Fig. 2) are between 2380 and 4150 mg / L in the waters of the river Meboudja. The highest concentrations are in the sectors of steel complex and Sidi-Amar. The highest iron content is mainly localized at the downstream of the industrial area and both sides of the wadi Meboudja (Fig. 2). The impact of pollution from discharges of all kinds on groundwater is very marked. Indeed, the high values recorded in the waters of wells near the wadi gradually decrease away from the banks of the Meboudja.

Figure.2 Salts total dissolved (TDS) and iron in surface water and groundwater in the plain of the river Meboudja (mg / L)

The chemical composition of water from the wadi Meboudja observed from 1999 to 2011 , located downstream of the river Meboudja (Table below) shows the dominance of five major chemical elements: Ca2+, HCO3-, Na + , Cl-, and SO4

2-. The origin of these elements is likely related to leaching metamorphic formations (marbles) and licks contributions Lake Fetzara (Kherici 1993 Debièche 2002).

Table.1 physico-chemical composition of water from the wadi Meboudja well year mon

th pH

Eh mV

O2 mg/L

Ca2+ mg/L

Mg2+

mg/L Na+

mg/L K+

mg/L Cl-

mg/L SO4-

mg/L NO3-

mg/L

HCO3-

mg/L

TDS mg/L

Fer (T) mg/L

Mn2+

mg/L

Meb wadi 2009

Aug 7,3 411 0,6 69 29 389 7,6 512 62 18 352 2475 4,69 1,10 Dec. 8,8 391 2,3 96 26 378 2 675 43 21 135 1229 2,77 0,30

2011 Aug. 7,2 400 0,8 78 96 301 10 298 114 0,46 495 2030 3,88 2,08 Dec. 8,9 390 4,2 100 29 245 5,6 232 165 37,5 212 1645 5,68 2,80

2700 2600

2900

2900

2870

2550

3200

3600

3710

3900

2800

32002680

4050

3980

41503940

3500

3250

3000

4100

31002400

2490

2380

4150S

eybo

use

Mebou

dja

Hadjar Ediss

SidiAmar

El-Hadjar

Pont-Bouchet

IndustrieSidérurgique

2600

2440

0 - 0,5 mg/L Fe0,5 - 1,0 mg/LFe1,0 - 1,5 mg/L Fe1,5 - 2,0 mg/L Fe2,0 - 2,5 mg/L Fe2,5 - 3,0 mg/L FeEau de surfaceTDS, mg/L2490Terrils

0 1 2 Km

Lentille d'argile

1289


4.1 Ground water flow and Mass Transport Modelling Groundwater contamination is limited with the mass transport processes. These processes determine maximum extent of plume spread and geometric character of the concentration distribution. Advection is by far the most dominant mass transport process in contaminant migration. Hydrodynamic dispersion is usually a second order process. The magnitude and direction of advective transport is controlled by the configuration of water table or piezometric surface, presence of sources or sinks, permeability distribution within the flow filed and flow domain. The hydraulic head is obtained from the solution of a three dimensional groundwater flow equation (Bear, 1972)

t

hSq

χjh

Kχi ssii

(1)

where Ss is the specific storage of the porous material. The transport equation is linked to the flow equation through groundwater velocity term given

by

χih

θK

v iii

(2)

Where Kii a principal component of the hydraulic conductivity tensor; h hydraulic head; the porosity of the porous medium.

MODFLOW software (McDonald and Harbaugh, 1988) is used for the hydraulic simulation. All

these parameters are important in controlling the groundwater velocity, which drives advective transport. Adding dispersion to advective transport can cause important changes in the shape of a plume. Another important process is sorption and irrespective of the model describing sorption, the process was of paramount importance in controlling contaminant transport. The partial differential equation describing three-dimensional transport of contaminants in groundwater (Javandel et al., 1984) can be written as

N

1Kks

siij RC

θq

Cvχiχj

CD

χit

C

(3)

where C the concentration of contaminants dissolved in groundwater; t time; i the distance along the respective Cartesian co-ordinate axis;

vi the hydrodynamic dispersion coefficient; vi the seepage or linear pore water velocity; qs the volumetric flux of water per unit volume of aquifer representing sources (positive) and

sinks (negative); Cs the concentration of the sources or sinks; the porosity of the porous medium;

Rk chemical reaction term. Assuming that only equilibrium controlled linear or non-linear sorption and first order irreversible rate reactions are involved in the chemical reactions, the chemical reaction term can be expressed as (Grove and Stollenwerk, 1984)

Cθρ

Cλt

C

θρ

R bN

1kk

(4)

where C The concentration of contaminants sorbed on the porous medium; the rate constant of the first-order rate reactions; b the bulk density of the porous medium.

Rewriting Equation (4) as

1290


C

C

t

C

θρb

t

C

θρ b

(5)

We can rewrite Equation (3) by substituting Equations (4) and (5) as

Cθρ

CλC

C

t

C

θbρ

Cθq

viCiχjχ

CD

iχt

C bs

sij

(6)

Rearranging the terms we get the governing equation of mass transport model

Cθρ

CλCθq

Cvχjjχ

CD

iχt

C bs

siijR

(7)

where R is the retardation factor, defined as

C

C

θρb

1R

(8)

4.2 Groundwater Flow Model

The numerical approaches for solving mass transport equations are approximate forms of the advection-dispersion Equation (7) as a system of algebraic equations or alternatively simulating transport through the spread of a large number of moving reference particles (particle tracking). The second step is to provide boundary condition and assign values of concentration or loading rates defining various boundaries for all nodes located along boundary of the domain. Initial conditions and transport parameters were specified for all nodes. The seepage from the Meboudja was simulated by giving additional recharge input to the model. Constant concentration was assigned in different parts of the Meboudja based on ambient surface was as well as groundwater concentrations measured during field investigations.

The simulated model domain consists of 40 rows and 40 columns and 1 layer covering an area of 8000 x 8000 m. The superficial aquifer mostly consists of a 10-15 m thick alluvium along the Meboudja. The simulated vertical section has a maximum thickness of 15 m. The groundwater recharge at the rate of 100 mm yr-1 has been simulated in the top layer. Continuous seepage from the Meboudja stream was simulated as additional recharge in the model. The first stage of modeling is flow simulation for computation of hydraulic head distribution. The distribution of hydraulic head and hence the velocity field is unaffected by migration of plume because density and viscosity of contaminated groundwater is nearly the same as uncontaminated water in the area. The flow equation was therefore, first solved independently of the mass transport equation. Further, water level observations in the area indicate that hydraulic gradients do not change significantly with time. Thus groundwater flow was assumed to be in a steady state and the groundwater heads were computed by visual MODFLOW (Guiger and Frantz, 1996) using Slice Successive Over Relaxation (SSOR) package (McDonald and Harbaugh, 1988). The flow model was calibrated by adjusting several parameters within a narrow range of values until a best fit was obtained between observed data and simulated results. The accuracy of the computed water levels (Fig.5) was judged by computing mean error, mean absolute error and root mean squared error computed for 15 observation wells. The calculated mean error, mean absolute error and root mean squared error under steady state condition is –0.14, 3.2 and 3.6 m, respectively.

1291


Figure 2. Distribution of hydraulic heads. Dashed lines, simulated; continuous lines, measured. Contour interval, 0.5 m.

4.3 Mass transport Model Mass transport in three dimensions (MT3D) is a computer model for simulation of advection, dispersion and chemical reactions of contaminants in three-dimensional groundwater flow systems (Zheng, 1990). The model is used in conjunction with a block-centred finite difference flow model MODFLOW. Dispersion was accounted for the particle in motion by adding to the deterministic motion a random component, which is a function of dispersivities. The mean concentration for each grid block was calculated as the sum of the mass carried by all the particles located in a given block divided by the total volume of water in the block. The values of dispersivity in longitudinal and two transverse directions (Y and Z) were assumed to be 10, 1 and 0.1 m, respectively, and the values were taken from the literature (Kimbrough et al., 1999; Domenico and Schwartz, 1990; Tevissen, 1993). The tendency for L to be about 10 times larger than TH and for TZ to be much smaller than either of them is in line with the concentrations observed in the area. The initial TDS concentration assigned in the rest of the area is about 2180 mg l-1.

9.0

8.58.0

7.5

7.0

6.5

9.5

1292


Figure 3. Computed TDS concentration (mg L-1) of groundwater. 5. Conclusions and Recommendations Here a case study of groundwater/surface water pollution due to uncontrolled industrial effluent discharges and its environmental impact on groundwater regime is presented. Groundwater pollution extends laterally 500-600 m from the Meboudja, in which initial pollutants load, in the alluvial areas covering villages Sidi-Amar and El-Hadjar. The extension of pollution is due to heavy pumping for irrigation, resulting in induced seepage from Meboudja due to stream aquifer interaction, which in turn carries surface water effluent to the groundwater regime. The contaminated groundwater is being exploited for agriculture and industrial purposes in the absence of major surface water sources in the area. The modeling study has helped to gain a better insight of the hydrogeologic set up and assessment of contaminant migration. The untreated effluents emerging from the industries must be monitored for maintaining the standards prescribed for TDS concentration by the environment inspection for various industries in the region. The present study provided a base line data for assessment of contamination in the El-Hadjar area. For reduction of the stream aquifer interaction, the pumping around the Meboudja wadi should be reduced. Periodical monitoring of the water quality has to continued to check the rise in TDS concentrations of groundwater.

396.

039

7.1

398.

239

9.3

400.

440

1.5

402.

640

4.0

943.0 944.9 946.0 947.1 948.2 949.3 951.0

vers

Ann

aba

3 Km

1

2

3

4

5

6

7

8

9

10

11: Valeurs de TDS observées, mg L-1: Valeurs de TDS calculées, mg L-1

Meboudja

1293


REFERENCES Bear, J., 1972. Dynamics of Fluids in Porous Media, New York, American Elsevier, pp. 764. Debièche, T.H., Mania, J., and Mudry, J., 2003. Pollution d’une nappe alluviale par le chrome et

l’étain à partir d’un stockage de résidus métallurgiques: Application à la basse plaine de la Seybouse, Nord-Est Algérien. Africa Geosciences Review, Vol. 8, N°4, pp. 425-435.

Djabri, L., 1996. Mécanismes de la pollution et vulnérabilité des eaux de la Seybouse. Origines géologiques, industrielles, agricoles et urbaines. Th. Doct. Es-Sciences, Univ. Annaba, 261p.

Domenico, P. A., Schwartz, F.W., 1990. Physical and Chemical Hydrogeology, John Wiley & Sons, U.S.A., pp. 824.

Grove, D.B., Stollenwerk, K.G., 1984. Computer Model of One-Dimensional Equilibrium Controlled Sorption Processes, U.S. Geol. Survey Ressources Investigations Report 84-4059, pp. 58.

Guiger, N., Frantz, T., 1996. Visual MODFLOW: Users Guide, Waterloo Hydrogeologic, Waterloo, Ontario, Canada.

Hani; A., 2003. Analyse méthodologique de la structure et des processus anthropiques : application aux ressources en eau d’un bassin côtier méditerranéen. Th. Doct. Es-Sciences, Univ. Annaba, 214p.

Javandel, I., Doughty, C, Tsang, C.F., 1984. Groundwater Transport: Handbook of Mathematical Models, American Geophysical Union, Water Ressources Monograph 10, pp. 228.

Kherici, N., 1993. Vulnérabilité à la pollution chimique des eaux souterraines d’un système de nappes superposées en milieu industriel et agricole (Annaba- la Maffragh, NE Algérien). Th. Doct. Es-Sciences, Univ. Annaba, 170p.

Kimbrough, D.E., Cohen, Y., Winer, A.M., Creelman, L., Mabuni, C., 1999. Acritical assessment of chromium in the environment. Critical Review in Environmental Science and Technology, 29, 1, 1-46.

Louhi, A., 1996. Pollution des eaux et sols. Cas de la région de Annaba. Etude des interférences et dosage Al, Fe, Zn, Ni, Cr, Pb, Sn et Hg par spectrométrie d’émission Plasma-ARC (DCP-AES), absorption atomique (SAA) et spectrophotométrie UV/VIS. Th. Doct. Es-Sciences, Univ. Annaba, 167p.

McDonald, J. M., Harbaugh, A.W., 1988. A Modular Three-Dimensional Finite-Difference Groundwater Flow Model, Techniques of Water Ressources Investigations of the U.S. Geological Survey Book, 6, pp. 586.

Tevissen, E., 1993. Méthodologie d’étude et modélisation du transport de solutés en milieux poreux naturels. Application à la migration du chrome dans la nappe alluviale du Drac (Isère). Th. Doct. INPL, Nancy, 150p.

Zenati, N., 1999. Relation nappe-lac, confirmation par l’hydrochimie. Cas de la nappe superficielle de la plaine Ouest d’El-Hadjar, lac Fetzara, N.E. Algérien. Mémoire Magister, Univ. Annaba, 148p.

Zheng, C., 1990. MT3D, a Modular three-dimensional transport model for simulation of advection, dispersion and chemical reactions of contaminants in groundwater system prepared for the U.S. Environmental Protection Agency.

1294



Hydrodynamic and hydrodispersive modeling: impact of activities industrial zone on the waters

area of Berrahal (Annaba)

CHAFFAI Hicham, KHELFAOUI Hakim, HANI Azzedine, BOUGHERIRA Nabil and DJABRI Larbi Laboratory of Water Resource and Sustainable Development, Department of Geology, Faculty of

Earth Sciences, University Badji Mokhtar, BP 12, 23000-Annaba, Algeria. E-mail: [email protected]

Abstract: Industrial development in the region of Berrahal resulted in increased discharges inducing environmental degradation and water quality of the aquifer. In this research we will try to delineate areas affected by discharges in the industrial area from the development of a model of groundwater flow in transitional regime and pollutant transport model. The objective is to improve the understanding of interactions liquid discharges from the industrial zone and groundwater from the water table fine sand and gravel in order to make a management support tool for the preservation and ecosystem protection. Calibration of transitional regime model has allowed fixing the permeability of the water table across the study area, where it is of the order of 10-2 m/s in the south and the east and lowest in centre (10-5 m/s). Thus piezometric maps calculated by the model almost coincide with that measured on the ground. The modeling of groundwater flow and mass transport has led to a better understanding of the effluent-layer interactions and to understand the contaminant migration. Indeed the modeling showed that the pollutant follows the streamlines (convective transport) while the transverse diffusion is negligible.

Keywords: Industrial area, hydrodynamic modeling, hydrodispersive, Berrahal.

1 INTRODUCTION

The 3D digital models are widely used in hydrogeology. They allow understanding and characterizing groundwater flow and transport of solutes, or water balances underground systems. Often used as a tool for prediction of the evolution of a system, they are also useful for identifying the predominant parameters of a hydrosystem and improve the characterization of system operation, its susceptibility or vulnerability. Modflow was developed by the U.S.. Geological Survey. This code was made public in 1984. It has become a standard in the modeling of groundwater flow (Anderson and Woessner, 1991; Fetter, 1994). It has been applied to many hydrogeological systems: porous aquifers, karst aquifers of alluvial, or complex multilayer aquifers; in connection with various issues: the exchange table - river, the influence of climate on water systems, the problems of transfers of pollutants and contamination of aquifers or pumping the impact on groundwater flow. Its robustness has been tested successfully for a variety of studies. This code consists of a main program with sub-routines or independent modules (McDonald and Harbaugh, 1988). The modules (Well package, River package, etc.) simulate specific hydrological aspects. This modular structure allows the use of hydrological functions independently of each other. It has facilitated the development of new modules without changing the code itself since the modules generally use the data outputs of the code. Many users have contributed to adding specific modules, thus increasing the possibilities and capabilities of Modflow. Restrepo and al. (1998) have developed a specific module to wetlands with the ability to manage runoff through vegetation and a drainage system. In this research, Modflow code is used to model groundwater flow and to understand the

1295

H. Chaffai et al. / Hydrodynamic and hydrodispersive modeling: impact of industrial activities on the waters of the plain of Berrahal (Annaba)

potential impacts of releases from the industrial area Berrahal on the quality of groundwater used for livestock feed and sometimes for drinking water needs populations (Khelfaoui, 2014).

2 GEOGRAPHICAL SITUATION

The city of Berrahal is located 30 Km to the west of Annaba; capital of the province in the far north-eastern Algeria. It covers an area of 180.15 km2, crossed by the National Road No. 44 (Annaba, Skikda and Constantine), limited to the south by the lake Fetzara depression (Figure 1). The industrial area is located to the southwest from Berrahal, capital of the municipality at distance of about 2 km and an area of 121 ha. It is located between three secondary cities (Kalitoussa in the South and the East, Guirch to the West and Tacha in North) and has 82 lots and 28 active businesses (Figure 2).

Figure 1. Geographical situation of the study area

Legend

No. Company 1 cookie Manufacturing 2 Manufacture of insulation products 3 tiles factory 4 Factory pasta Mahbouba 5 Savonnerie Nozha 6 Public Works ENMTP 7 NAFTAL 8 Production of olive oil 9 Factory Meraka tiles 10 semolina production Beldi 11 factory woven bags Azatis 12 KIA Motors Dealer 13 Laiterie Fetzara 14 Factory Saba pasta 15 Sarl Espap GST 16 Deposit Pharmaceuticals 17 Maintenance Sarl Gesibat 18 Laiterie El Meida 19 Factory Bona soft drinks 20 Protuil Hot dip galvanizing 21 Storage of scrap iron COPRAC

Figure 2. Implantation of companies in the industrial area of Berrahal

919.5 920 920.5 921 921.5 922 922.5

401.5

402

402.5

403

403.5

Vers Kalitoussa

Route Nationale N° 44

Vers Constantine

Coord. Lambert

Vers Annaba

0 0.5 1 km

123 4

5

6

7

89

1012

13

14

15

16

17

1819

20

21

11

1296


3 MATERIALS AND METHODS

3.1 Hydrogeological context of the area studied

The study area which is part of the depression of the Lake Fetzara is treated as two adjoining layers together by vertical leakance through a semipermeable gritty clay horizon (Figure 3). - The first aquifer is the end of gravel and sand with a thickness of 7-20 m (unconfined, shallow). - The second is the semi-confined aquifer of cipolin and cracked limestone with a thickness of 2 to 5 m, which becomes captive to the west of the study area. The two webs are limited by: - The massive Edough north (mica schist and gneiss) as a supply reservoir for these aquifers over 30 m thick (power limit). - Depression Lake Fetzara south and east, which is the outlet and therefore a potential limit imposed. - Depression of Lake Fetzara south and east, which is the outlet and thus a potential limit imposed. - Depression of the plain of Tacha northwest considered a potential limit imposed. - The plain of Guirche southwest considered a potential limit imposed. - The low hills west of Guirch constituting a power limit.

The area studied spans an area of approximately 26.4 km2, located by the Lambert coordinates X (km) from 918.3 to 924.71 and Y (km) from 399 to 404.65.

The study area is discredited horizontally into 652 square mesh side and 200 m vertically into three layers; the superficial layer of fine sand and gravel, the semi-permeable layer of gravelly clay and deep layer of marbles and limestone cracked. The model allows the simulation of flows in permanent and transient in an aquifer system.

3.2 Groundwater Flow and Mass Transport Modelling

The hydraulic head is obtained from the solution of a three dimensional groundwater flow equation (Bear, 1972):

t

hSq

j

hK

i ssii ∂∂=+

∂∂

∂∂

(1)

Where SS is the specific storage of the porous material. The transport equation is linked to the flow equation through groundwater velocity term given by:

i

hKv ii

i ∂∂=

(2)

Where: Kii a principal component of the hydraulic conductivity tensor; H hydraulic head; θ the porosity of the porous medium.

MODFLOW software (McDonald and Harbaugh, 1988) is used for the hydraulic simulation. The partial differential equation describing three-dimentional transport of contaminants in groundwater (Javandel et al., 1984) can be written as:

( ) ++∂

∂−∂∂

∂∂=

∂∂

=

N

1Kks

siij RC

qCv

ij

CD

it

C

(3)

Where: C the concentration of contaminants dissolved in groundwater; t time; χi: the distance along the respective Cartesian co-ordinate axis; Di: the hydrodynamic dispersion coefficient; vi: the seepage or linear pore water velocity;qs: the volumetric flux of water per unit volume of aquifer representing sources (positive) and sinks (negative); Cs: the concentration of the sources of sinks;

1297


θ : the porosity of the porous medium; Rk: chemical reaction term. Assuming that only equilibrium controlled linear or non-linear sorption and first order irreversible rate reactions are involved in the chemical reactions, the chemical reaction term can be expressed as (Grove et Stollenwerk, 1984):

+−∂∂−=

=CC

t

CR bN

1kk

(4)

Where: C the concentration of contaminants sorbed on the porous medium; λ : the rate constant of the first-order rate reactions; ρb: the bulk density of the porous medium. Rewriting Equation (4) as:

C

C

t

Cb

t

Cb∂∂

∂∂=

∂∂

(5)

We can rewrite Equation (3) by substituting Equations (4) and (5) as:

( ) +−∂∂

∂∂−+

∂∂−

∂∂

∂∂=

∂∂

CCC

C

t

CbC

qviC

ij

CD

it

C bs

sij (6)

Rearranging the terms we get the governing equation of mass transport model:

( ) +−+∂∂−

∂∂

∂∂=

∂∂

CCCq

Cvjj

CD

it

C bs

siijR

(7)

Where R is the retardation factor, defined as:

C

Cb1R

∂∂+=

(8)

4 RESULTS AND DISCUSSION

4.1 Transient modeling

The simulation phase is seen from November 2006 to May 2007, this phase is divided into four periods with a pitch of 60 days time (November 2006, January 2007, March 2007 and May 2007).

4.1.1 The conceptual model

After identifying the different layers and the boundary marker, the location of three boreholes BH1, BH2 and Guirch (Figure 3) with their depths and their speed (of the order of 150 m3/day) for four periods the piezometric monitoring, introduction of values of the permeability, the storing and the porosity of each layer coefficient (in the absence of measurement of these parameters in the study area, we referred to the measurements in the plain of Annaba which has similar characteristics especially in the plain of El-Hadjar and data taken in reports for identical formations. The data of these parameters are summarized in Table 1.

Table 1. Input data of hydrodynamic characteristics for the model realization

Permeability (m/s) Storage coefficient Porosity Superficial layer (fine

sand and gravel) 10-2 à 10-5 10-1 à 10-2 0.3 (Todd, D.K., 1980)

Clay gritty semi-permeable layer 10-6 (Daly, C., 1982) 10-2 (Morris, D.A., et A.I.

Johnson, 1967) 0.45 (Todd, D.K., 1980)

Cipolin deep layer of limestone and cracked 10-4 10-4 0.01 (Rabinowitz, D.D.,

et G.W. Gross, 1972)

1298


Figure 3. Diagram showing the aquifer horizons in eastern part of the study area

4.1.2 Wedging the hydrodynamic model

The piezometric maps calculated by the model are confronted with field results. These cards have generally the same shape and overlay of two piezometric maps of the model and that measured on the ground in the month of March 2007 (Figure 4), the level of isopiézométriques curves can depend on: - The pumping rates at certain meshes (existence of some feed borings in drinking water). - Potential imposed at the boundaries bordering the plain and the values of the effective recharge from precipitation. - Permeability values characterizing each of the cracks.

4.1.3 Adjusting values of permeability and storage coefficient

Values from field results led in a first phase to establish homogeneous permeability ranges. Selected Piezometric statements are those of November 2006, January 2007, March 2007 and May 2007. The timing consecutive tests helped change the permeability values in order to reach a maximum distance of 1 m between the measured and simulated piezometric coast. Thus, a map of the distribution of the permeability values which can result in a zonation from 10-2 to 10-5 m/s for the surface layer of fine sand and gravel, and 10-4 m/s semi-confined aquifer of limestones and marbles cracked. For the semi-permeable layer of gravel clays, the permeability is of the order of 10-6 m/s (Figure 5). The most permeable zones are located on the edges, especially in the eastern part and in contact with the South Lake Fetzara, which can promote the lake-layer interactions. The storage coefficient varies from 0.1 (drilling level Guirche West) to 0.01 for the surface layer; it is of the order of 10-4 at the semi-confined aquifer, and at 0.01 of the semi-permeable layer (Table 2).

1299


Table 2. Water balance calculated in transitional regime

PeriodsParameters

Nov. 2006 Jan. 2007 Mar. 2007 May 2007Entries

Emmagasinement (m3/d) 0 0,7 10E-5 7,54 10E-3 0,3 10E-5Constant potential (m3/d) 22428 24989 20820 20820Drilling (m3/d) 0 0 0 0Refill (m3/d) 0,63 10E-3 1,26 10E-3 37,81 10E-4 22,06 10E-4Total inputs (m3/d) 22428 24989 20820 20820

ExitsEmmagasinement (m3/d) 143,18 10E-3 122,89 10E-4 0,1 10E-5 0Constant potential (m3/d) 22279 24840 20670 20670Drilling (m3/d) 149.26 149.26 149.26 149.26Refill (m3/d) 0 0 0 0Total output (m3/d) 22428 24989 20820 20820DifferenceEntries – Exits (m3/d)

61,62 10E-4 42,86 10E-4 -6,77 10E-4 -69,75 10E-4

Error % 0

4.2 Mass transport model

The MT3D model (Zheng, 1990) was used for mass transport simulation in the web (convection, dispersion and chemical reactions). This model is in tune with the MODFLOW software the United States Geological Survey (USGS) that supports the modeling of groundwater flow finite difference.

Figure 4. Piezometry measured and calculated by the model of the month of March 2007

The dispersity values in the x, y and z are respectively taken equal to 10, 1 and 0.1 m (Kimbrough et al., 1999; Domenico and Schwartz, 1990; Tevissen, 1993). The tendency for αL about 10 times larger as αT and 100 times as αZ is consistent with the observed concentrations in the sector. The initial value Dissolved Solids Total (TDS) assigned in the rest of the sector is about 500 mg / l.

1300


The gradual reduction of TDS values away from the industrial area reflects a relatively constant recharging of the pollutant from the main flow axis. Thus, a constant value in TDS has been assigned to the various nodes of the industrial zone, with a value of 700 mg / l at the source (at the industrial zone) and 500 mg/l in the rest of the study area. A study on a longer period and some profiles from the industrial area to the lake would identify the risk associated with industrial discharges on the ecosystem of Lake Fetzara (Figure 6).

Figure 5: Distribution card of the permeabilities of the surface table cloth simulated in transitory mode

Figure 6. Map of propagation of the pollutants starting from the industrial area of Berrahal, example of the Total Dissolved Solid TDS (in mg/l)

1301


5 CONCLUSION

The objectives of this modeling work were to improve understanding of hydrogeological interactions of liquid discharges from the industrial area Berrahal with groundwater from the water table fine gravel and sand, in order to make it a tool management support for the preservation and protection of this ecosystem. From the conceptual model, a hydrogeological model with the digital Modflow code was built. Calibration transient helped secure the permeability of the water table across the study area, where it is of the order of 10-2 m/s in the south and the east and lower in the center (10-5 m/s). Thus piezometric maps calculated by the model almost coincide with that measured on the ground. The modeling of groundwater flow and mass transport has led to a better understanding of the effluent-layer interactions and to understand the contaminant migration. Indeed the modeling showed that the pollutant follows the streamlines (convective transport) while the transverse diffusion is negligible.

ACKNOWLEDGMENTS

We thank the sector of the water resources for information at the time of the realization of this work.

REFERENCES

Anderson Mary P. and Woessner William W., 1991. Applied Groundwater Modeling: Simulation of Flow and Advective Transport. Academic Press; 1 edition: 381 pp.

Bear J.: 1972, Dynamics of Fluids in Porous Media, New York, American Elsevier, pp. 764. Domenico Patrick A., Schwartz Franklin W., 1990. Physical and chemical hydrogeology. Wiley, -

Technology & Engineering - 824 pp. Fetter, C.W., 1994. Applied Hydrogeology, 3rd ed.: Macmillan College Publishing, Inc., New York, 616

pp. Grove D.B. and Stollenwerk K.G. (1984) Computer Model of One-Dimensional Equilibrium Controlled

Sorption Processes, U.S. Geol. Survey Resources Investigations Report 84-4059, pp. 58. Javandel I., Doughty C., Tsang C.F. (1984) Groundwater Transport: Handbook of Mathematical

Models, American Geophysical Union, Water Resources Monograph 10, pp. 228. Journal Officiel de la République Algérienne Démocratique et Populaire, 2006. Valeurs limites des

paramètres de rejets d’effluents liquides industriels. N°26, Annexe I. Hani A., Djorfi S., Djabri L., Lamouroux C. and Lallahem S., 2007. Impact of the industrial rejections

on water of Annaba aquifer (Algéria). European Water 19/20: 3-14. Khelfaoui Hakim., 2014. Modélisation des écoulements et de transport de masse dans une région

industrielle : Exemple de la nappe de Berrahal, Nort-Est Algérien. Thèse de Doctorat en Sciences, Faculté des Sciences de la Terre, Dpt. de Géologie, UBM-Annaba, 170 p.

Kimbrough, D.E., Cohen, Y., Winer, A.M., Creelman, L. and Mabuni, C., 1999. A critical assessment of chromium in the environment. CRC Crit Rev Environ Sci Technol 29, 1–46.

McDonald M.D. and A.W. Harbaugh. 1988. A modular three-dimensional finite-difference flow model. Techniques of Water Resources Investigations of the U.S. Geological Survey, Book 6. 586 pp.

Restrepo, V. R., Thompson, G. G., Mace, P.M., Gabriel, W. L., Low, L. L., MacCall, A. D., Methot, R. D., Powers, J. E., Taylor, B. L., Wade, P. R., and Witzig, J. F. 1998. Technical guidance on the use of precautionary approaches to implementing National Standard 1 of the Magnuson– Stevens Fishery Conservation and Management Act. National Oceanic and Atmospheric Administration (US) Technical Memorandum NMFS-F/SPO-31. 54 pp.

Tevissen E., 1993. Methodologie d'étude et modélisation du transport de solutés en milieux poreux naturels. Soil Sci Soc Am 2: 213–269.

World Health Organization, 2011. Guidelines for Drinking-water Quality, fourth edition, ISBN 978 92 4 154815 1, 564p.

Yves Le Cofrec, 2011. Sites et sols pollués. Edition Dunod, Paris, 407p. Zheng, C., 1990. MT3D, A Modular Three-Dimensional Transport Model for Simulation of Advection,

Dispersion and Chemical Reactions of Contaminants in Groundwater Systems, Report to the U.S. Environmental Protection Agency Robert S. Kerr Environmental Research Laboratory, Ada, Oklahoma.

1302

About the Use of Multiple Training Dates for Land Change Modeling

Martin Paegelow 1 and María Teresa Camacho 2

1 Université Toulouse Jean Jaurès, GEODE UMR 5602, France – [email protected] Dpto. de Análisis Geográfico Regional y Geografía Física, Universidad de Granada, Spain –

[email protected] Abstract: Most spatial-temporal geomatic simulation tools for land change use two land use / cover (LUC) maps as training dates, particularly popular prediction models based on Markov chains (MC). This study first focusses on potential errors when considering only two past dates. Then, authors illustrate its consequences on quantitative model calibration by a dataset englobing six LUC maps offering a wide range of training date pairs. The spectrum of so computed Markovian probability transitions is even larger by varying the confidence level in input data. The comparison of Land Use / Cover Change (LUCC) budgets and possible Markov chains show a broad spectrum of results and the arbitrariness by choosing only two dates. These techniques are illustrated with data of Garrotxes catchment (8 500 ha) in French Pyrenees. Currently crops has disappeared and he near future depends on the intensity of pastoral activity and spontaneous reforestation. Authors suggest two techniques to integrate the information of more than two training dates to perform the quantitative aspect of forecasting scenarios. Primary they compute a time distance weighted annual rate of change to fix the expected total change in the simulation step as well as the categorical one. Also they produce alternatives to Markov chains on the transitional level. Therefore, authors integrate all available LUCC-budgets and propose different techniques of weighting observed transitions with the aim to produce transition matrices englobing the whole available information as alternatives to only 2 dates based Markov chains. Keywords: Land Change Modeling; Model Validation; Multiple Training Dates.

1303

1304



An agent-based model to analyze the ecological functioning of agro-silvo-pastoral landscapes in

West Africa

Myriam Grillota,b, Jonathan Vayssièresa,b, Benoit Gaudouc, Alassane Bahb,d, Dominique Massee,f

a CIRAD Umr SELMET “Mediterranean and Tropical Livestock Systems”, Montpellier, France b Dp PPZS “Pastoral Systems and Dry Lands”, Dakar, Sénégal c IRIT CNRS, Université de Toulouse, Toulouse, France d UCAD - Ummisco, “Dakar Cheikh Anta Diop University - Unit for Mathematical and Computer Modeling of Complex Systems”, Dakar Fann, Senegal e IRD, UMR Eco&Sols “Functional Ecology and Biochemistry of Soils and Agroecosystems”, Montpellier, France f LMI IESOL “Ecological Intensification of cultivated Soils in west Africa”, Centre ISRA IRD Bel Air, Dakar, Sénégal

Abstract

Biomass recycling plays a major role in the functioning and sustainability of agro-silvo-pastoral ecosystems in West Africa. However, farming systems must adjust rapidly to changing landscapes. This abstract describes an Agent-Based Model (ABM) that analyzes nutrient (N) cycles within the village “terroir” agro-ecosystem to assess how changes in land use and farming practices affect crop-livestock integration and spatial heterogeneity. The ABM is implemented on the GAMA platform (Generic Agent-based Modelling Architecture). Each simulated household is classified according to a farm type with a pre-defined biomass management strategy (use of crop products, manure, etc.). The model is spatially explicit, with biomass being moved across landscape units and through different ecosystem components: soil, plants, animals, humans, etc. Biomass flows are converted into N flows which are then used to calculate indicators such as N balance, N use efficiency, and the diversity and recycling indexes that characterize N cycles. The model was parameterized through on-farm measurements of crop and livestock productions (for the biophysical sub-system) and participatory workshops with farmers (for the decision sub-system). The model was validated on the basis of biomass flows observed in two villages in the Senegalese groundnut basin. The model is original in its multi-scale analysis of N cycles occurring at different organizational levels: plot, herd, household and landscape. For instance this highlights the consequences of households’ choices at the plot scale for spatial heterogeneity at the landscape scale. This ABM is potentially generalizable for designing ecological intensification pathways in West African village “terroirs”. Keywords: Agent-based-model, agro-silvo-pastoral ecosystem, biomass flows, multi-scale analysis

1305

Using Decision support system to identify the suitable land-parcels for hosting solar energy

modules in Cairo Region

Taher Osmana, Hichem Omranib, Amin Tayyebic and Prasanna Divigalpitiyad

a Ph. D. Candidate, Department of Architecture and Urban Design, Kyushu University, Japan. [email protected] Researcher, UDM dept., LISER, Luxembourg. [email protected] [email protected] Associate Professor, Graduate School of Human-Environment Studies, Kyushu University, Japan. [email protected]

Abstract: Cairo witnessed highly accelerated rates of unplanned urban growth during last three decades which converted 65% of Cairo citizens to live in slums currently. Although government planned for urban growth rates less than 2.5% annually between 1984-2013, unplanned growth exceeds 6.5% annually and jumped to more than 10% annually after 2011 creating more unplanned urban areas need to be provided with infrastructure particularly electricity. The wide gap between the authorized urban growth rate and the real one creates similar wide gap in the ability of providing infrastructure and the real needs, which creates severe shortage particularly in electricity power supply. The government prepares energy supply plans based on the authorized urban growth plans to build fossil fuel plants which need 5-10 years to be accomplished. Government then build new fossil plants to meet the shortage in electricity power. However, new unplanned growth areas had been added to the previous planned demand which increase the gap of energy supply before the new plants being accomplished and still this gap increases in an infinite-cycle. Electricity supply shortage made Cairo deny a huge power outage (2-6 hours /day) regularly in summer season and discrete in the rest of year causing severe disruption to more than 25 million Cairenes. The government plans became a late response to meet electricity demand which creates an increased dynamic shortage in energy supply in an infinite cycle with highly severe influences. On the other hand, Cairo is located in the top hotspots zone of potential solar energy all over the world, which was described as the ideal future electricity supply source for similar cases worldwide. The need to smart solar energy modules (SSEM) can expand automatically with the immediate demand of rapid urban growth to meet their local particularly needs of energy away from the complexed authorized governmental plans. SSEM is flexible with randomly urban occurrence, compatible with the nature of local urban characteristics, doesn’t need to a huge infrastructure to provide electricity, and flexible to work in various urban patterns. Therefore, the goal of this paper is to identify the various set of locations in Cairo that can be appropriateto host SSEM modules, focusing on the potential of compatible areas within the built environment and the additional lands that can combine high-quality solar potentials with closeness to existing major roads, infrastructure, and electricity transmission lines. That can offer multiple choices for minimizing the environmental influences related to the large-scale transition to a renewable energy mix where solar energy modules serve as a growing source to optimize the transmission incorporation of potential energy sources. The result of this study will represent in classifying the whole developed area of Cairo region to compatible, potentially compatible, incompatible land parcels for SSEM modules which will be applied ultimately as a part of the prospected decision support system for solar energy.

Keywords: Solar Energy modules, sensitive spatial analysis, Compatibility matrices, built environment.

1306

Comparison of load estimation methods for total nitrogen in Estonia

Ottar Tamm1 and Toomas Tamm2

1Institute of Forestry and Rural Engineering, Estonian University of Life Sciences, Kreutzwaldi 5, Tartu 51014, Estonia e-mail: [email protected]

2Institute of Forestry and Rural Engineering, Estonian University of Life Sciences, Kreutzwaldi 5, Tartu 51014, Estonia e-mail: [email protected]

Abstract: To mitigate eutrophication in the Baltic Sea, one must have overview of the nutrient loads entering the Baltic Sea. In Estonia, water quality sampling and analysis is usually done monthly or quarterly. Because of the scarcity of water quality data in Estonian basins, simple methods for nutrient estimation are used. Currently, total yearly nutrient loads are calculated by using the flow-weighted mean concentration method. The aim of this study is to investigate how currently used method estimates total nitrogen loads compared to the LOADEST (Load Estimator), which uses a regression model to estimate loads for days on which water quality data were not measured. The correlation of coefficient for the LOADEST regression model averaged 0.94 (35 water quality monitoring sites) for total nitrogen (Ntot). No bias in load estimates was found, compared to the measured water quality data. The calculated load for the total period (1992-2014) of Ntot was not significantly different from the total load when using the current method. However, differences in the yearly load estimates were found. For example, years 1994 and 1995 showed a strong underpreditcion in load, 20.2% and 22.2% respectively. 2011 revealed a possible overprediction of 25.7%. In one of the rivers, the maximum yearly load over- and under estimation was, 96.4% and 62.3%, respectively. The results from this study show, that LOADEST regression model can be successfully used to estimate total nitrogen loads in Estonia and should be a preferable choice for estimating yearly nitrogen loads or when trend analysis is needed. Keywords: LOADEST; flow-weighted concentration; total nitrogen load; Estonia; Baltic Sea

1307



Impact of climate change on winter durum wheat

cultivated in Southern Italy: effect of extreme weather events

Domenico Ventrella, Pasquale Garofalo

Consiglio per la ricerca in agricoltura e l'analisi dell'economia agraria - Unità di ricerca per i sistemi colturali degli ambienti caldo aridi (CREA-SCA), [email protected]

Abstract: The productive performances of the cropping systems are increasingly affected by climate change (CC). A particular aspect of CC is the forecasted increase in the frequency of extreme weather events such as frosts, droughts, heat waves or heavy precipitation events. This work is within FP7 project “MODelling vegetation response to EXTREMe Events” (MODEXTREME, KBBE.2013.1.4-09, Grant Agreement No. 613817, 2013-2016). The main goals are: i) to assess the capability of process-based modelling solutions (MS) modified by the inclusion of impact algorithms; ii) to compare the existing and modified MS under scenarios of climate change. The reference area considered in this study is strategically important for cultivation of winter durum wheat and it is located in Southern Italy, in the northern part of Puglia region. CropSyst and WOFOST were the crop simulation models considered as MS and therefore calibrated in their original and MODEXTREME modified versions against measured productive and phenological variables included in a dataset coming from a long-term experiment running in Foggia. The modified versions of the two MS include algorithms that regulate the effects of stresses due to extreme events of high temperature, water deficit and cold temperature on change in harvest index and leaf area index (only frost). Both calibration and the study of vulnerability to CC were carried out under BioMA (Biophysical Model Applications), a public domain software framework designed and implemented for developing, parameterizing and running modelling solutions based on biophysical models in the domains of agriculture and environment. The impact of CC was evaluated comparing a 20-year (2000/20) reference period against two future scenarios referred to 2020/40 and 2040/60 periods. Compared to the original MS, the modified-MS simulated a reduction of grain yield improving the simulation capacity of the two models decreasing the differences with the measured values in the calibration/validation steps. In particular, heat stress during flowering and reproductive phases was detected in the reference and future periods as simulated by climate models. Keywords: extreme weather events - crop simulation - vulnerability to climate change - winter wheat – CropSyst – Wofost - Bioma

1308

Analysis of the intrinsic relationship between the spatial radiometry of Vitellaria paradoxa (C.F. Gaertn.) and soil in Sudano-Guinean zone in

northwest Benin

AGBANOU Thierry (a), ALLAGBE Simon (a), TENTE Brice (a), OREKAN Vincent (a) and PAEGELOW Martin (b)

(a) Laboratory of Biogeography and Environmental Expertise (LABEE), University of Abomey-Calavi, Bénin, [[email protected]]

(b) GEODE UMR 5602 CNRS University of Toulouse Jean Jaurès, France, [[email protected]]

AbstractSeveral tree species such as Vitellaria paradoxaare used as energy wood / timber in Natitingoumunicipality. This wood is used for example for manufacturing: mortars, coal, and used as firewood. These forms of human’s cumulative uses to soil degradation influence the spatial distribution of these species. This study aims to assess the relationship between remote sensing soil conditions and spatial distribution of the species. For data collection, a sample of 131 plots of 30 m by 30 m ofVitellariaparadoxa (C.F. Gaertn.) population was geo-referenced and counted. Spatial interpolation by kriging empirical Bayes densities of the species was carried out under ArcGis. The Landsat satellite image of December 2013 was processed with ENVI and ArcGIS for extraction of ground radiometric indices and those of the vegetation. The Factorial Correspondence Analysis (AFC) was performed with these indices and species density under Sphinx software The results of the AFC show that the color index and the index of soil redness are inversely correlated to the spatial density of the species. The more the soil is degraded, the less Vitellaria paradoxa is present. The degradation of the species is mainly found in tropical ferruginous soils of quartzite and concretion in thenortheast part of the study area. It is therefore imperative for decision makers to define sustainable management plans for the Vitellariaparadoxa parks species.

Keywords : Benin, spatial correlation, radiometry, Vitellaria paradoxa.

1309

-

An agent based model of the effects of fishing policy on the Icelandic coastal communities

Cezara Păstrăva,b, Kristinn Nikulás Edvardssonc

aMatís Vinladsleid 12, 113 Reykjavík, Iceland bUtrecht University, Domplein 29, 3512 JE Utrecht, Netherlands cUniversity of Iceland, Sæmundargata 2, 101 Reykjavík, Iceland

[email protected], [email protected]

Abstract: In Iceland, over the past quarter century, since the implementation of the individual transferable quota system (ITQ), the efficiency of the fishing industry increased, but the coastal communities have seen significant changes in their socio-economic structure, due to the aggregation of quota in the hands of a few major companies. There is considerable debate going on over whether the ITQ system is the best option, or a different quota system would be more desirable for maintaining high levels of efficiency without negatively impacting the coastal communities.

In order to study the effects of different policies, we are developing an agent based model of the Icelandic coastal communities and the Icelandic fishing fleet. We model individual boats, companies, and communities as agents who interact according to the policies in effect. The boats are modelled after vessels active within the Icelandic fishing fleet, based on data from the Directorate of Fisheries and the Icelandic Ship Registry. Companies trade quota in order to maximize profits, and can choose to relocate their boats to a different harbour as a result. They can be influenced by community feedback in their decisions. Communities prosper based on the revenue they receive from the fishing activities taking place in their territory. Their initial state is based on social and economic indicators, which we derive from data obtained from Statistics Iceland.

Different quota systems (including the current ITQ one) are implemented as different sets of rulesgoverning the behaviour of boat and company agents. We compare the effects of these different scenarios have on community agents by measuring socio-economic indicators such as total population, population structure, employment, revenue from fishing, and revenue from other local industries. The model is implemented in the Unity3D game engine because it emphasizes spatial and geographical components, as well as user interaction - which allows us to later use it for science outreach activities involving policy makers.

Keywords: agent based modelling, fisheries management, policy, Unity3D, science outreach

1310

Paper Title: Heterogeneous Data Mining for a Semi-Quantitative Risk Assessment of Oil Contamination from Multiple-Sources in The Ecuadorian Amazon

Juan Durango ᵃᵇ; Mehdi Saqalliᵇ; Laurence Mauriceᶜ; Nicolas Maestripierib & Arnaud Elgerᵃ. ᵃ Ecolab, Université Toulouse III, France

([email protected], [email protected]) ᵇ Geode, Université Toulouse II

([email protected], [email protected]) ᶜ OMP-GET, Université Toulouse III

([email protected])

Abstract: The North Ecuadorian Amazon is one of the most biodiverse places per surface area on Earth. However, its richness and importance is threatened by complex human activities driven by oil exploration and extraction, indirectly responsible for forest clearing. Furthermore this extractive industry has long been regarded as polluting the environment and causing cancer and other diseases to the local population. Two periods of extractive practices and technology implementation can be well differentiated (before environmental regulation 1965-2000 vs. after regulations 2001-to present). GIS data and production inventories of potential polluting infrastructures due to spilling and infiltration (i.e. drill cuttings pits, oil wells, flare stacks and pipes) for more than 45 years, have been collected. These data are not geographically homogeneous and cover only 10-20% of the total number of infrastructures that represent a risk for the environment. Our aim is to estimate toxic emissions from these sources, considering the input of chemicals used in extraction and effluents from oil wells which can be expressed as different polluting fractions (i.e. carbon, hydrocarbons and associated heavy metals such as Ni and V). The spatial statistical analysis of these data will allow identifying patterns for a spatially-explicit model of risk. First results show that contamination differs geographically, while oil infrastructure density and population overlap. Oil practices in absence of environmental regulations led to higher contamination than permitted by present Ecuadorian regulations’ standards. A linear regression between cumulative oil contamination and the density of oil infrastructures is observed, notably concerning the number of oil wells (r²=0.95). These findings could help defining key parameters to facilitate future decision-making processes for environment and population protection in highly biodiverse biomes.

Keywords: oil infrastructure; intensity based risk; missing data estimation; environmental contamination

1311



An evaluation of tree-based ensemble methods for

sensitivity analysis in environmental models

Marc Jaxa-Rozena, Jan H. Kwakkela

a Faculty of Technology, Policy and Management, Delft University of Technology

([email protected], [email protected]) Abstract: The sensitivity analysis of uncertain model parameters is a key element of good modelling practice in the environmental sciences. However, models for complex environmental applications typically require global analysis techniques (e.g. with the variance-based approach of Sobol , 2001) to account for non-linearities and interactions between parameters; most of these techniques are highly computationally expensive, making them difficult to apply in practice. In this context, different screening methods - such as Morris elementary effects (Campolongo et al., 2007; Morris, 1991) - can approximate the output of a variance-based global sensitivity analysis, at a smaller computational cost. However, these techniques typically impose certain restrictions, such as specific sampling methods or the use of strictly continuous, rather than categorical, input parameters. As an alternative, this work evaluates tree-based methods for the estimation of global sensitivity analysis results, in the form of the random forests and Extra-Trees (ET) algorithms (Breiman, 2001; Geurts et al., 2006). These algorithms both rely on an ensemble of decision or regression trees generated from a learning sample, which are then aggregated to estimate the importance of input parameters on model output. These techniques can be used with conventional Monte Carlo or Latin Hypercube sampling, and with mixed continuous and categorical inputs; they are also available in existing libraries such as the Python scikit-learn package. The tree-based methods are compared to the Sobol and Morris techniques as implemented in the Python SALib library, for three cases: an Ishigami-Homma test function, a H1N1 influenza pandemic model (Pruyt and Hamarat, 2010) and the WorldWater water resource model (Simonovic, 2002). The ET algorithm is found to perform favorably compared to Morris elementary effects, approximating Sobol total effects indices with a lower mean absolute error than the Morris u* parameter. Furthermore, the ET feature importance estimates stabilize at a lower total sample size, and perform comparably on a measure of ranking error. As such, this approach could offer a useful and user-friendly alternative for the screening of uncertain parameters in models with inputs of mixed types. Keywords: sensitivity analysis; parameter screening; machine learning REFERENCES Breiman, L., 2001. Random Forests. Machine Learning 45, 5–32. doi:10.1023/A:1010933404324 Campolongo, F., Cariboni, J., Saltelli, A., 2007. An effective screening design for sensitivity analysis of

large models. Environmental Modelling & Software, Modelling, computer-assisted simulations, and mapping of dangerous phenomena for hazard assessment 22, 1509–1518.

Geurts, P., Ernst, D., Wehenkel, L., 2006. Extremely randomized trees. Mach Learn 63, 3–42. doi:10.1007/s10994-006-6226-1

Morris, M.D., 1991. Factorial Sampling Plans for Preliminary Computational Experiments. Technometrics 33, 161–174. doi:10.1080/00401706.1991.10484804

Pruyt, E., Hamarat, C., 2010. The influenza A (H1N1) v pandemic: an exploratory system dynamics approach, in: Proceedings of the 28th International Conference of the System Dynamics Society, Seoul, Korea, 25-29 July 2010. System Dynamics Society.

Simonovic, S.P., 2002. World water dynamics: global modeling of water resources. Journal of Environmental Management 66, 249–267. doi:10.1006/jema.2002.0585

Sobol , I.M., 2001. Global sensitivity indices for nonlinear mathematical models and their Monte Carlo estimates. Mathematics and Computers in Simulation, The Second IMACS Seminar on Monte Carlo Methods 55, 271–280. doi:10.1016/S0378-4754(00)00270-6

1312

1313

1314

1315

1316

1317

1318



Dealing with Uncertainty– Multiple Simple Groundwater Models

Kamil Nešetřila and Jaroslav Nosekb

aTechnical University of Liberec, Czechia, http://msp.mti.tul.cz, [email protected] bTechnical University of Liberec, Czechia, https://lcs.cxi.tul.cz/lide/jaroslav-nosek

Abstract: Heterogeneous subsurface is usually characterized by scarce data – therefore it is reasonable to fully exploit available data and develop multiple simple models. Strategy of multiple simple models and especially alternative (equifinal) simple models was demonstrated on three case studies. Those models have diverse structure, are based on diverse assumptions or test diverse hypotheses. Tenability of the models was semi-quantitatively evaluated by the pedigree matrix of Refsgaard et al. (2006). In the first case study four alternative models represent four alternative hypotheses explaining loss of water in a river. One hypothesis quantifies a natural process – the effluence to the subglacial channel. The water loss can be explained this way therefore the hypothesis was not rejected. Other hypotheses explain the loss by seepage to the surface coal mine through diverse geological layers. Closed-form solution with simplified assumptions (e.g. geometry of the modelling domain is simplified to a single rectangle) was used. The second case study combines different simple models for different parts of a contaminated site. Maximal extent of the contaminant plume (downstream from the contamination source) is estimated by computing steady-state plume length by three methods implemented in the CoronaScreen (Wilson et al. 2005) code. The maximal permitted pumping rate upstream is calculated by a simple spatial computation based on the distance from the plume's fringe. Pumping rate is limited by recharge from precipitation and lateral influx of the capture zone of the exploited well. The third case study demonstrates usefulness of using simple 1D advection-dispersion model with merged scaling parameters (1st order transformation rate and retardation coefficient) to simulate nanomaterial transport in groundwater. Keywords: ground water; multiple model ensemble; simplicity; equifinality ACKNOWLEDGMENTS The contribution was prepared with the financial support of FP7 project GUIDEnano "Assessment and mitigation of NM-enabled product risks on human and environmental health: Development of new strategies and creation of a web-based guidance tool for nanotech industries", no. 604387. REFERENCES Refsgaard, J.C., Van Der Sluijs J.P., Brown J., van der Keur, P., 2006. A framework for dealing with

uncertainty due to model structure error. Advances in Water Resources. 29 (11), 1586–1597, doi:10.1016/j.advwatres.2005.11.013

Wilson, R.D., Thornton, S.F., Huettmann, A., Gutierrez-Neri, M., Slenders, H., 2005. CoronaScreen : Process-based models for natural attenuation assessment: guidance for the application of na assessment screening models. University of Sheffield, UK; TNO, The Netherlands. https://www.sheffield.ac.uk/polopoly_fs/1.521443!/file/CORONA-Guidance-Document-v1.0.pdf

1319



-

An Agent Based Model of Southern Pine Beetle Forest Damages

Emma Cutlera, Mark Borsuka, David Lutzb, and Matthew Ayresc

aThayer School of Engineering, Dartmouth College, Hanover, NH 03755, USA ([email protected], [email protected]), bEnvironmental Studies Program,

Dartmouth College, Hanover, NH 03755, USA ([email protected]), cDepartment of Biological Sciences, Dartmouth College, Hanover, NH 03755, USA ([email protected]) Abstract: Southern pine beetle (Dendroctonus frontalis Zimm.) has historically been confined to the southeastern United States, where periodic outbreaks have caused widespread damage, including the loss of harvestable timber and non-market forest ecosystem services. As the climate becomes warmer and drier, the pest’s range is expanding northward along the Eastern Seaboard, raising concerns about future environmental and economic impacts. This research presents a probabilistic agent based model that simulates tree mortality in response to bark beetle outbreaks. A forest is represented as a grid of trees, in which the probability of an outbreak in a given year depends on average diameter at breast height, total basal area, percent stand composition of host trees, and minimum winter temperatures. In years when an outbreak occurs, the location of the initial infestation within the grid is chosen randomly, and the infestation spreads to adjacent trees at a rate based upon stand features and beetle population dynamics. In this way, patches of dead trees form throughout the landscape, thus reproducing the characteristic “spots” that result from southern pine beetle outbreaks. The model is designed to be applied to the US Forest Service Forest Inventory and Analysis sites to simulate the damages from southern pine beetle under variable climate conditions and forest stand composition. Ultimately, this can inform management decisions so as to maximize the net present value of market and non-market forest ecosystem services, with particular focus on timber, carbon storage, and surface albedo. Keywords: Southern pine beetle; forest damages; agent based model

1320

1321

1322

1323

1324

1325

ı́ı́

1326

1327



-

New Spacing Algorithm for Level Bench Terraces

Yihune D. Tadesse College of Agriculture and Environmental Sciences,

Bahir Dar University, Bahir Dar 5501, Ethiopia E-mail: [email protected]

Abstract: Due to social and natural factors, many slope lands are being brought into cultivation which is aggravating the global and local soil erosion rates. Steep lands are very difficult to till and cultivate unless bench terraces are constructed. Similar to other soil conservation structures, spacing determination is so decisive in bench terraces. Despite the presence of a widely used formula for the vertical interval determination of bench terraces; it lacks comprehensiveness, simplicity, modification and logical proof. There is much difficulty to find its derivation in published documents. Moreover, studies specifically oriented to the spacing of bench terraces are not common. This paper is focused on mathematical solutions which justify for the maximum vertical interval (VImax) between consecutive structures and width of the bench. It employs a sequential approach integrating preliminary levels algebraic, geometric and trigonometric applications as well as figurative and arithmetic tools to develop a comprehensive, precise and simplest set of two spacing formulae for level bench terraces. The common factors of consideration assumptions have been avoided and expressed in the VImax formula. Keywords: vertical interval; bench terrace spacing; terrace design; soil conservation design

1 INTRODUCTION

Terraces constitute a crucial landscape engineering measure to control soil erosion, raise crop output, and maintain sustainable agribusiness development under hillside agriculture in mountainous regions (Cao et al., 2007). On sloping lands, terraces are primarily installed to grow crops without causing excessive soil erosion. In soils prone to erosion, however, bench terraces or traditional terraces are appropriate (Blanco and Lal, 2008). Bench terraces are a series of level or nearly level platforms built along contours at suitable intervals (FFTC, 1995). Bench terraces had a beneficial effect on crop yield through the levelling of the slope and the water conserving effect, allowing an increase in crop density (Posthumus and Stroosnijder, 2010). In many instances in the past; bench terraces were built without proper design, resulting in either high construction/maintenance costs or limited use (Sheng, 2002). Based on measurements of sediment yield from terrace risers and beds and from terrace units made during three consecutive rainy seasons, Van Dijk and Bruijnzeel (2004) reported that despite widespread bench terracing of rainfed hillsides in upland West Java, soil loss rates remain high.

Figure 1. A schematic representation of a land slope transformed into a level bench terrace

CA FA

Wfb C

B

A

D E

F r S

GH

Riser Bench

Original land slope

CA = Cut Area FA = Fill Area

1328

D.T. Yihune/ New spacing algorithm for level bench terraces

Based on the type of formation, bench terraces are classified into:

Progressive/Developed bench terraces - are formed by soil, stone or stone faced soil bund upgrading. The vertical interval (VI) of such terraces is determined by the VI of the (fanya juu) bund/terrace prior to the bench terrace. Width of the shelf (Wb) determination is not essential.

Excavated bench terraces - are constructed, at a time, by digging and filling. Vertical interval (VI)

and width of the shelf (Wb) are the most essential design specifications for the construction of excavated bench terraces.

Considering the runoff amount, land slope and VI have inverse relations. Due to their relative suitability on slope lands, bench terraces are closely spaced than other soil conservation structures to consider runoff amount going out from a (nearly) flat bench. In other words, they are peculiar from other types of soil conservation structures in that they are affected by the cut and fill (engineering) rather than runoff generating characteristics or peak runoff amount. Even if the factors such as soil and crop root depth are known to be factors affecting spacing, soil conservation structures are not being built with a suitable and scientific spacing due to incapability to incorporate a comprehensive list of factors in various VI formulae. Many researchers have not addressed the soil depth constraint more than holding the assumption that soil depth is not a constraint; Sheng (2002) recommends designing terraces according to the needs of farmers, crops, climate, and tools to be used for farming. Sheng (1977) states it as the minimum soil depth for the bench terraces can be obtained by dividing the height of the riser by two. In other words, the VI between conservation structures should not be more than twice the soil depth. Therefore, only soil depth and land slope (Hurni, 1986) were considered previously. 1.1 Objectives

To develop the simplest formula for width of a flat shelf (Wfb) and a new comprehensive formula for maximum Vertical interval (VI) computation of excavated level bench terraces

To understand the relationship between the angle of repose and the riser slope gradient To remove the common factors of consideration assumptions which are described qualitatively

in VI calculation and to express them quantitatively of excavated level bench terraces 2 METHODOLOGY

Important points in a full cut-fill cross-sectional view diagram are arbitrarily named as alphabetical letters that helped in understanding the geometry of cutting and filling in bench terraces. Then, a theorematic argument integrating preliminary levels of algebraic and trigonometric applications as well as figurative and arithmetic tools was employed to develop a comprehensive and rational maximum vertical interval and width of a level bench terrace formulae. 3 RESULTS 3.1 The width of a flat shelf (Wfb)3.1.1 Theorem 1 - The width of a flat shelf (Wfb) In bench terraces, Wfb and VI are related by the formula. Where Wfb = Width of a flat shelf (in which its unit is similar to the VI); VI = Vertical interval; S% = Slope gradient in percent (before the bench terrace construction); r% = Slope of the designed riser in percent (after the bench terrace construction).

Wfb = 100VI (

1329


3.1.2. Preliminaries Vertical interval (VI) is the vertical distance between similar points/parts of successive conservation structures. In this demonstration, VI is shown from center to center of the succeeding risers or alternate non-cut and non-fill points (as points A and C in figure 1). 3.1.3. Proof of Theorem 1 -The width of a flat shelf (Wfb) It is shown in the figure that the original slope of the land before the construction of the bench terraces is S° and thus tan S° can be calculated as: tan S° = (1) In bench terrace formation, the cutting of ∆ ADB is designed to fill ∆ CEB. Moreover, ABD = CBE [Equality of vertical(ly opposite) angles] Thus, ∆ ADB = ∆ CEB (2) Accordingly, the triangles have the same height which implies, Side opposite to S° = (3) If the riser slope gradient is r, tan r° = = ; = (4) Wfb = From (2), it can be deducted = (5) Side adjacent to S° = Side adjacent to S° = [Due to (4) and (5), respectively] (6) Inserting (3) and (6) in (1) shows that tan S° =

Multiplying the numerator and the denominator by 2 changes the equation into: tan S° =

tan S° ( ) + Wfb tan S° = VI Wfb tan S° = VI tan S° ( ) Wfb = Because S% = 100 tan S° and r% = 100 tan r°, Wfb = Wfb = Therefore, it has been proved that

Wfb = 100VI (

1330


3.2. The Maximum VI (VImax) for a flat shelf 3.2.1. Theorem 2 - The Maximum VI (VImax) for a flat shelf

Unless a permanent catchment of stony nature is needed, the VImax for LBTs (level bench terraces) is computed as: Where Dso = the depth of soil (before the bench terrace construction); Dcr = the maximum crop root depth (after the bench terrace construction); S% = Slope gradient in percent (before the bench terrace construction);

r% = Slope of the designed riser in percent (after the bench terrace construction) 3.2.2. Preliminaries Soil depth bears the greatest constraint at points just below the riser such as D (which is point H before the bench terrace construction) in Figure 1 or 2. This leads to the proof of VI max at D. 3.2.3. Proof of Theorem 2 - Maximum VI for a flat shelf

ABD = AHG (Equality of angles created by parallel lines intersecting the same straight line) In other words, AHG = ABD = S° Thus, tan S° when considered at AHG equals tan S° = = ................................................ = due to equality of parallel lines This implies = tan S° Due to (4) [ = ], = (7) After the bench terrace construction, the soil depth at point H should be sufficient for crop root depths to be planted.

= – = Dso –– Dcr (8) Where Dso and Dcr = the depth of soil and the depth of the maximum crop root depth required, respectively (before and after the bench terrace construction, respectively). Dcr is slightly/significantly greater than the maximum root depth to be planted. Due to equality of parallel lines and = – can be clearly seen in Figure 1 or 2,

= = –

= – [Due to (3) and (7), respectively] (9) Inserting (9) into (8) changes the equation to – = Dso – Dcr

= Dso – Dcr

= Dso – Dcr Multiplying both sides by gives VI =

VImax =

1331


Figure 2. A level bench terrace diagram with the maximum crop root depth to be planted and a parent

material VI =

Hence, for a level bench terrace, the VImax = It is designated as the VImax because the soil depth will not be sufficient for the root zone if the VI is increased as it can be clearly seen in Figure 2. In fact, the VImax can be increased when a permanent catchment of stony nature is needed.

4 Discussion It is for the flow of the proofs that Theorem 1 - The width of a flat shelf (Wfb) preceded Theorem 2 - VImax. However, the sequence should be reversed in practice. As it can be observed in Figure 1 or 2, bench terraces cannot be imagined where the riser slope gradient (r) is equal to or less than the land slope gradient (S). They always have r > S and the fill will have a riser gradient equal to the angle of repose. Therefore, the angle of repose should be considered in the construction of similar to any other conservation structure and can only be used when the slope is less than the angle of repose. 4.1. Discussion on Theorem 1 -The width of a flat shelf (Wfb) In previous spacing outlines, there was a tendency to derive out VI rather than the Wb in a probably similar mathematical analysis and was implied that Wb should be determined first (FFTC, 1995; Sheng, 1977; Sheng, 1988; Sheng, 2002). However, it should be argued that the maximum interval should be decided not only on the soil depth but also on land slope (Hurni, 1986) and crops, specified in 3.2 as ‘the depth of the maximum crop root depth’; other known factors which affect spacing but not incorporated in a formula. This research has also discovered that riser gradient also affect the VImax.

4.2. Discussion on Theorem 2 - Maximum VI (VImax) for a flat shelf All conservation structures should be checked for the Theorem on VImax since the structures are proposed to have the ultimate formation of bench terraces progressively described in Section 1. However, the peak runoff amount determines the VI to a great extent than the soil depth in gentler slopes of other structures which have a far spacing. Thus, the assumption that soil depth is a constraint in limiting the VI (in VI equations that do not consider soil depth); factors of subjective consideration such as land slope and crop root depth has

A

D F r G

H

I

B

Original land slope Parent material

1332


been avoided since they and a new attribute, riser slope, have been incorporated in the Theorem on VImax. Therefore, the Theorem on VImax shows that bench terraces should not be used in a shallow soil which is a common feature of tropical steep slopes or else the VI will be too small. Sheng (1977) states the minimum soil depth for the bench terraces can be obtained by dividing the height of the riser by two. In other words, the VI between conservation structures should not be more than twice the soil depth. It can be generalized that previous equations mostly focus on only the soil depth. Therefore, BTs (bench terraces) are not conservation structures alone but part of a long-term agricultural and conservation plan and investment which require specification of crops to be planted and cropping pattern/system to follow, whether and what to grow on shoulder bund/lips, and farming tools and implements to be used.

5 Assumptions, Conditions and Practical Procedures

In the derivation of the spacing equations, it is assumed that the bench fill is sufficiently compacted and compressed so that its volume equals the cut part and are applicable on a common condition that land configuration is a straight slope.

It is for the flow of the proofs that Theorem 1 - The width of a flat shelf (Wfb) preceded Theorem 2 - VImax. However, the sequence should be reversed in practice and the procedure in Section 1 should be followed.

6 Conclusions

A simpler formula for width of a flat shelf (Wfb) and a comprehensive formula for the determination of maximum Vertical interval (VImax) in level bench terraces have been developed. Factors such as the soil depth and the maximum crop root depth as well as riser and slope gradient are seamed into the maximum VI theorem.

Bench terrace riser gradient equals the angle of repose only on the fill portion. The common factors of consideration assumptions have been avoided and expressed in the

maximum VI formula. The maximum VI formula is assumed to be incomplete and needs computation of exact VI

values less than the maximum vertical interval (VImax) when overtopping is expected or observed in cases such as a humid area with a higher maximum VI.

7 Recommendations The maximum VI and width of a flat shelf formula can be used for level bench terraces. In cases where the depth of the deepest /sub-soiling/ tillage is greater than the depth of the maximum crop root depth, the former should be used in the maximum VI computation. When overtopping is expected, inward sloping bench terraces should be preferred; when overtopping is observed, lip/shoulder bund can be added. Equations for excavated BTs must be developed for both VI and Wb on which at least one is independent of the other. However, equation for either VI or Wb is developed; and the other, mostly Wb (which is so important in the design, layout and construction), is overlooked. Being determined by soil depth, land slope and crop root depth, VI is of prime importance. The width of the shelf (Wb) should be derived from the VI. If Wb developed so is incompatible with traditional farming tools or mechanized farming, decreasing the Wb to be compatible and calculating back exact VI values less than the maximum vertical interval (VImax) would be the best solution. ACKNOWLEDGMENTS The author would like to thank Merkuz Abera (PhD) for his valuable comment at the presentation of the proof at the College of Agriculture and Environmental Sciences Weekly Seminar at Zenzelma Campus of Bahir Dar University that initiated the author for discussion, output articulation and Theorem 2 - VImax.

1333


REFERENCES Blanco, H., Lal, R., 2008. Principles of Soil Conservation and Management. Springer

Science+Business Media B.V. 617p. Cao, S., Chen, L., Feng, Q., Liu, Z., 2007. Soft-riser bench terrace design for the hilly loess region of

Shaanxi Province, China. Landscape and Urban Planning 80: 184–191. FFTC (Food and Fertilizer Technology Center for the Asian and Pacific Region), 1995. Soil

Conservation Handbook (English Edition): pp. I-14 – I-18. Taibei, Taiwan. Hurni, H., 1986. Guidelines for Development Agents on Soil Conservation in Ethiopia, Community

Forests and Soil Conservation Development, Addis Ababa, Ethiopia. 100 pp. Posthumus, H., Stroosnijder, L., 2010. To terrace or not: the short-term impact of bench terraces on

soil properties and crop response in the Peruvian Andes. Environment Development and Sustainability 12:263–276.

Sheng, T.C., 1977. Protection of cultivated slopes – terracing steep slopes in humid regions. In: Kunkle, S.H., Thames, J.L., (Eds.), Guidelines for Watershed Management. FAO Conservation Guide 1, FAO, Rome, pp. 147-179.

Sheng, T.C., 1988. Part II - Terraces and ditches. In: Dembner, S., (Ed.), Watershed management field manual: Slope treatment measures and practices. FAO Conservation Guide 13/3, FAO, Rome, pp. 33-142.

Sheng, T.C., 2002. Bench Terrace Design Made Simple. In: Yuren, J., editor. Proceedings of 12th International Soil Conservation Organization Conference, Beijing, China, Vol. IV, pp. 500-504.

Van Dijk, A.I.J.M., Bruijnzeel, L.A., 2004. Runoff and soil loss from bench terraces. 2. An event-based erosion process model. European Journal of Soil Science 55, 317–334.

1334

An agent-based model of the Amazonian forest colonisation and oil exploitation: the Oriente study

case

William CHAPOTAT(a), Lionel HOUSSOU(a), Audren BOUADJIO - BOULIC(b),Nicolas MAESTRIPIERI(a), Emilie LERIGOLEUR(a), Benoit GAUDOU(b), Mehdi

SAQALLI(a).

(a) UMR 5602 GEODE, CNRS, University of Toulouse 2 Jean Jaurès, Toulouse 31058, France: [email protected], [email protected], [email protected],

[email protected], [email protected](b) UMR 5505 IRIT, CNRS, University of Toulouse 1 Capitole, Toulouse 31000, France:

[email protected], [email protected]

Abstract : Since the 1970s, the northern part of the Ecuadorian Amazonia (the Oriente) has been colonized with migrants coming mainly from the Andes, with the support of the oil extraction industry which opened up roads. To evaluate the impacts of this colonization and oil extraction over environment, society and population health, we initiated a retro-prospective study through the construction of a spatially explicit agent-based model reconstituting human activities and public policies over this territory. The present poster describes this model, built using the KIDS (Keep It Descriptively Stupid) approach and implemented under the GAMA platform. It integrates three main modules:

The biophysical module includes rainfall, watershed, forest and pollution dynamics.The social module is implemented with many agents, who communicate among them, produce

livestock, food and cash crops in their farms but also work in oil companies. Moreover, they interact with their spatial and economic environment.

The regulation module defines oil and crop prices, demographics, public policies related to agriculture and health. It is implemented with experts based scenarios.

The poster presents the processes common to these three modules. It gives details about conceptual choices and the calibrations performed. It exposes the test and validation procedures done by confrontation with external data. Finally, we present a set of prospective scenarios, which will be used to explore potential futures of the Oriente and as decision-support tools for authorities.

Key Words: Multi-agent model, Ecuadorian Amazonia, oil pollution issues, socio-economic behaviorand vulnerability.

1335

Documents

iEMSs 2016 Conference · July 10-14, 2016, Toulouse, FRANCE. ... The organizers of the iEMSs 2016 Conference want to express thanks to the actively ... University of Exeter Medical