QUALITY INFORMATION DOCUMENT For the Global Ocean Wind ...resources.marine.copernicus.eu/documents/QUID/CMEMS-WIND-QUI… · associated standard deviations do not exceed 1.30m/s and

QUALITY INFORMATION DOCUMENT

For the Global Ocean Wind Products

WIND_GLO_WIND_L4_NRT_OBSERVATIONS_012_004

Issue: 1.2

Contributors: Abderrahim Bentamy

CMEMS version scope : April 2019 Release

Approval Date by Quality Assurance Review Group : 29/04/2019

QUID for Global Ocean Wind Product

WIND_GLO_WIND_L4_NRT-OBSERVATIONS_012_004

Ref: CMEMS-WIND-QUID-012-004

Date : January 2019

Issue : 1.2

© EU Copernicus Marine Service – Public Page 2/ 26

CHANGE RECORD

Issue Date § Description of Change Author Validated By

1.0

2018-04-27

All

Creation of the document, Wind QUID split up from the SIW QUID.

Abderrahim Bentamy

Jean François Piollé

Cedric Prevost

Abderrahim Bentamy

1.1 2018-08-24 Rebranded to Wind TAC M. Belmonte Mercator Ocean

1.2 2019-01-10 Adapted to April 2019 Release

M. Belmonte Mercator Ocean




Date : January 2019

Issue : 1.2


Table of contents

Executive summary ................................................................................................................................................ 4

Products covered by this document ................................................................................................................ 4

Summary of the results .................................................................................................................................... 4

Production Subsystem description ................................................................................. Erreur ! Signet non défini.

L4 Global blended ocean wind ........................................................................................................................ 7

ValidatIon framework .......................................................................................................................................... 10

Procedure ........................................................................................................................................................ 10

Result summary .............................................................................................................................................. 13




Date : January 2019

Issue : 1.2


I. EXECUTIVE SUMMARY

Products covered by this document

This document describes the quality of the newest global ocean near-real-time 6-hourly

L4 wind product (WIND_GLO_WIND_L4_NRT_OBSERVATIONS_012_004), which

contains 6-hourly blended wind fields estimated from scatterometer wind vector observations;

SSMIS wind speeds, and NWP wind forecasts. The wind variables include 6-hourly averaged

wind speed, zonal and meridional wind components and wind stress amplitude, curl and

divergence of both wind and wind stress vectors. The input (scatterometer and retrievals) L2

scatterometer observations are made available by the operational Ocean and Sea Ice Satellite

Application Facility (OSI SAF) of EUMETSAT, and SSMIS are from Remote Sensing

System (RSS). In this version NWP data are equivalent wind stress winds calculated by

KNMI. The latter are available for IFREMER. The new L4 product involves the following

new variables: wind vector curl and divergence, and wind stress curl and divergence. The new

L4 product is produced by IFREMER and will be distributed by CMEMS.

Table 1: Wind TAC Wind products and partner roles.

Product Product description Production unit, PU Dissemination unit DU

WIND_GLO_WIND_L4_NRT_OBSERVATIONS

_012_004

Near-real-time global ocean L4 blended

winds with ECMWF operational forecast winds as background

OSI-IFREMER-BREST-FR

CNR-ISAC-GOS (Roma)

Summary of the results

The investigations of the quality of the objective method, used to estimate the gridded

wind fields, the resulting blended wind estimates, and of the operational procedure are

checked as follows:

Characterization of the error relied on the objective method used to estimate global

gridded wind fields from swath observations. It is performed using synthetic data

derived from numerical model (ERA Interim) interpolated onto satellite swaths. No

systematic departure is found between ERA Interim analysis and 6-hourly averaged

winds derived from the synthetic data based on the use of the objective method. The

correlations between the two sources exceed 0.97.




Date : January 2019

Issue : 1.2


Determination of L4 blended wind product accuracy through comprehensive

comparisons with 6-hourly winds from available moored buoy data. The latter are

derived from various buoy networks: NDBC/NOAA (Atlantic, Pacific oceans), UK

Met Office and Météo-France (Atlantic and Mediterranean Sea), TAO (Tropical

Pacific), PIRATA (Tropical Atlantic), and RAMA (Indian Ocean). Comparisons are

performed for April 2017 – April 2018. For instance, results drawn from NDBC

comparisons lead to quite low biases for wind speed as well as for wind direction. The

associated standard deviations do not exceed 1.30m/s and 20°, respectively. The

scatter and vector correlation coefficients for wind speed and direction are higher than

0.90 and 1.83 (vector correlation), respectively.

Assessment of L4 blended wind product quality based on comparisons with spatially

and temporally collocated ASCAT retrievals. The former are performed over global

ocean to characterize L4 and ASCAT wind speed and direction agreements. The

comparisons indicate that blended wind speed and wind components are in good

agreement with the remotely sensed ones. The main aim of such comparisons is to

highlight how the blended analysis retrieves the remotely sensed wind observations

derived mainly from ASCAT. Collocated data are used for comparisons purposes. The

overall statistics characterizing ASCAT and blended collocated data comparisons

indicate that the biases are close to zero and the standard deviation (std) values are less

than 1 m/s. The correlation coefficients exceed 0.95. It is noticeable that zonal and

meridional components have similar behaviours. Furthermore, ASCAT and blended

exhibit better agreements than those drawn from ASCAT and ECMWF equivalent

stress wind comparisons.

Quality control of each netcdf file geophysical content is performed based on the

calculation of the statistical parameters characterizing the difference between ASCAT

retrievals and L4 wind speed and direction data, and between ECMWF equivalent

stress and L4 winds. Only files such as differences indicate significant discrepancies

are checked. The statistical parameter time variabilities drawn from ASCAT and L4

comparisons are quite steady. For instance, wind speed bias, standard deviation, and

correlation are of 0.02 m/s, 1.0 m/s, and 0.97, respectively, along the study period.




Date : January 2019

Issue : 1.2


Estimated Accuracy Numbers

Table 2: Statistical parameters characterizing differences between 6-hourly averaged NDBC buoy and L4 blended wind speed and direction estimates for April 2017 - April 2018. Length for number of samples, Std and Cor stand for standard deviation and correlation, respectively.

Wind Speed Wind Direction

Length Bias

(m/s)

Std

(m/s)

Cor

(scalar)

Bias

(deg)

Std

(deg)

Cor (Vector)

OffShore(>=50km) L4 58962 0.07 1.28 0.92 -2 18 1.80

Coastal(<50km) L4 35638 0.85 2.11 0.85 -4 25 1.57




Date : January 2019

Issue : 1.2


II. PRODUCTION SUBSYSTEM DESCRIPTION

L4 Global blended ocean wind

The regular (in space and time) wind fields are estimated from near real time

scatterometer and radiometer data in combination with equivalent stress wind estimated from

ECMWF forecasts. The remotely sensed winds are derived from available ASCAT

scatterometers onboard Metop-A and Metop-B satellites. The radiometer winds are from the

special sensor microwave imager sensor (SSMIS) onboard defense meteorological polar

satellites (DMSP) F16, F17, F18 and F19. Wind speed and direction from WindSat

radiometer onboard the Department of Defense Coriolis satellite are also used. The

scatterometers as well as radiometers are provided in near real time by SAF OSI (KNMI) and

RSS, respectively. They are extracted based on ftp procedures. The scatterometer data are

provided as L2b products including backscatter coefficients measurements and the associated

radar parameters as well as wind retrievals. SSMIS and WindSat winds are provided as L2b

products. Both L2b products are quality controlled prior any analysis. L4 winds are calculated

from L2b products in combination with equivalent stress wind from ECMWF forecasts using

the objective method. The latter are processed and provided by KNMI. The analysis is

performed for each synoptic time (00h:00; 06h:00; 12h:00; 18h:00 UTC) and with a spatial

resolution of 0.25° in longitude and latitude over global ocean. The objective method details

may be found in (Bentamy and Croizé-Fillon, 2012)1. Figure 1 shows zooms of L4 blended

wind products estimated over the Northwest Atlantic Ocean and the Western Mediterranean

area and occurring during October 21st

2017.

1Bentamy A., D. Croizé. Fillon, 2012: Gridded Surface Wind Fields from Metop/ASCAT

Measurements. Inter. Journal of Remote Sensing, 33, pp 1729-1754.




Date : January 2019

Issue : 1.2


Table 3 L4 6-hourly blended wind Product Specification

Product Line WIND_GLO_WIND_L4_NRT_OBSERVATIONS_012_004 Geographical coverage Global

Variables wind speed

wind zonal component

wind meridional component

wind stress amplitude

wind stress zonal component

wind stress meridional component

wind vector curl

wind vector divergence

wind stress curl

wind stress divergence

root mean square (rms)

sampling length

Analysis yes

Available time series from Jan 2018 to present

Temporal resolution 6-hourly (00h:00, 06h:00, 12h:00, 18h:00 UTC) averaged field

Target delivery time daily

Delivery mechanism CMEMS Information System: SUBSETTER, FTP

Horizontal resolution 1/4°

Number of vertical levels 1 (surface)

Format Netcdf4




Date : January 2019

Issue : 1.2


Figure 1: Example of the four synoptic wind fields (CMEMS L4 product) estimated from

remotely sensed data occurring on 21 October 2017.




Date : January 2019

Issue : 1.2


III. VALIDATION FRAMEWORK

Validation is a continuous on going activity to characterize accuracy and quality of the

delivered sea ice and wind products. It is mainly be based on operational data, but can be

supported by campaign data

Each PU is responsible for validation of their products. The Wind TAC Validation activities

are for the most based on what is already implemented at the partners’ institutes and has

shown to be useful.

Description of validation data and procedures and link to validation results for each product

are given in the next sections.

Procedure

The subsystem performance and associated product quality are scientifically assessed in

the following way:

Determination of L4 blended wind products accuracy through comprehensive

comparisons with 6-hourly winds from available and valid moored buoy data. The

latter are derived from various buoy networks: NDBC/NOAA (Atlantic, Pacific

oceans), UK Met Office and Météo-France (Atlantic and Mediterranean Sea), TAO

(Tropical Pacific), PIRATA (Tropical Atlantic), and RAMA (Indian Ocean). Figure 2

shows buoy locations.. More than 196 buoy raw data are routinely collected,

investigated, and collocated in space and time with monthly satellite estimates. The

main statistical parameters, including the first four conventional moments and the

linear regression parameters, are estimated and provided (Table1). The differences

between buoy and blended wind products are investigated according to geographical

locations (e.g. off-shore, coastal, high-latitudes, mid-latitudes and tropical areas).

Scatterplots and the related statistical parameters (bias, rms, correlation, linear

regression coefficients) illustrating the comparison between L4 blended global ocean

wind and available and validated buoy wind speeds and directions (Figure 3)..

At global scale the quality of each L4 blended wind product is monthly assessed based

on comparisons with spatially and temporally collocated ASCAT retrievals (when

available). The former are performed over global ocean to characterize L4 blended and

scatterometer wind speed and direction agreements.

Determination of global maps of differences between 6-hourly satellite wind product

and the associated ECMWF analysis. Maps illustrate the bias, rms difference and

correlation coefficient spatial patterns for wind speed, zonal, and meridional

components.




Date : January 2019

Issue : 1.2


Time series of differences between 6-hourly satellite wind product and the associated

ECMWF analysis. They deal with bias, rms difference and correlation coefficient for

wind speed, zonal, and meridional components.

Comparison of the results characterising L4 blended wind product (wind speed, zonal

and meridional components) and ECMWF forecasts (equivalent stress winds) (see

Figures 6, 7, and 8)

Quality control of each netcdf file geophysical content is performed based on the

calculation of the statistical parameters characterizing the difference between L4

blended wind speed and direction data and the related ECMWF forecasts2. Only files

such as the difference frequency exceeding three times of standard deviations is higher

than 10% will be checked. According to the finding, the blended wind field could be

reprocessed.

2 MYO-SIWTAC-Cal/val Plan

https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=11&cad=rja&uact=8&ved=2ahUKEwi78Iq3q_XhAhUmxoUKHcF3BUAQFjAKegQIAhAC&url=http%3A%2F%2Fmyocean.met.no%2FSIW-TAC%2Fdoc%2Fcalibration-report-Ifremer_Wind_Clim.doc&usg=AOvVaw3BYMRzNCXW3Tcpmnbs6CdN




Date : January 2019

Issue : 1.2


Figure 2: Locations of buoys used for the determination of the L4 wind product.




Date : January 2019

Issue : 1.2


IV. RESULT SUMMARY

L4 blended and NDBC buoy wind Comparisons

The quality of the resulting wind fields is investigated trough comprehensive

comparisons with 6-hourly averaged wind estimated derived from buoy measurements.

Comparisons are performed for coastal buoy (located less than 50 km from coastline) and for

offshore buoys. Figure 3 shows an example of comparison results. The latter are illustrated

through scatterplots between buoy and satellite wind speeds (left panels), and between buoy

and satellite wind directions (right panels). and Table 2 illustrate the statistical results

obtained on one hand from collocated offshore buoy and L4 blended wind data, and on other

hand from collocated coastal (buoy moored less than 50 km off coastlines), occurring during

January – May 2017 period.

Table 2 indicates that for offshore comparisons, no systematic departures are depicted

and for most wind variable bins, the collocated data are close to the perfect line. The

associated errors are lower than 1 m/s and 20° for wind speed and direction, respectively. The

significant bias is found for buoy wind speeds less than 4 m/s. Indeed, L4 blended wind

speeds tend to be slightly overestimated compared to in-situ data due to differences in spatial

representation and binning. Table 2 summarizes the results characterizing the comparisons

between NDBC and L4 blended 6-hourly wind speeds and directions. The related statistical

parameters are mean (Bias) and standard deviation (Std) of buoy minus blended data, scalar

and vector correlation coefficients (Cor) for wind speed and direction, respectively. The

overall statistics indicate that the 6-hourly satellite wind fields compare well to averaged buoy

data. The rms differences do not exceed the scatterometer specifications, for wind speed and

direction, respectively. For in-situ and scatterometer winds higher than 3 m/s no significant

bias trend is found. The wind direction bias is relatively small. The results obtained from

comparisons performed based on the use of collocated coastal buoys are lower than those

found for offshore buoys. Therefore, L4 blended winds should be used with caution in near

shore areas.




Date : January 2019

Issue : 1.2


L4 blended and ASCAT L2b wind Comparisons

Surface winds retrieved from scatterometer measurements represent the state of art in

global wind vector estimations. Therefore, ASCAT retrievals are used to assess and evaluate

the blended wind vector estimates at global scales. The main aim of such comparisons is to

highlight how the blended analysis retrieves the remotely sensed wind observations derived

mainly from ASCAT. Collocated data are used for comparisons purposes. The bias and rms

differences between collocated ASCAT wind observations and the blended wind analyses of

wind speed, zonal and meridional components, estimated during May 2017, are shown in

Figure 4. The overall statistics characterizing ASCAT and blended collocated data

comparisons indicate that the biases are close to zero and the standard deviation (std) values

are less than 1 m/s. The correlation coefficients exceed 0.95. It is noticeable that zonal and

meridional components have similar behaviours. More specifically, the std differences of the

three variables (wind speed, zonal, and meridional components) exhibit low values (lower

than 1m/s) over the Atlantic, Pacific and Indian trade wind regions, where the wind is quite

Figure 3: Comparisons of NDBC and L4 wind speed (left panels) and wind

direction (right panels)




Date : January 2019

Issue : 1.2


steady. The highest std values are found at high latitude and especially north and south 40°

where surface wind is more variable. At these mid and high latitudes the std values are about

1-2m/s for wind speed, while for zonal and meridional components they are about 1.5–3 m/s.

Through the three std patterns, we can notice a band of high rms values located north of the

equator in the Atlantic and Pacific oceans and in the Gulf of Guinea. This may be related to a

misplacement of the intertropical convergence zone (ITCZ) in the blended wind fields.

However, the sampling length of collocated data from ASCAT and blended winds is

minimum over this band. Indeed, between the equator and 10°N, the sampling length is about

30 whereas is more than 70 elsewhere. Some high std value regions are depicted too and

especially for both wind components. For instance, over the northwest Atlantic Ocean (Gulf

Stream current) the std values may exceed 2m/s for both wind components. This is mainly

related to storm tracks characterising surface wind condition during north hemisphere winter.

To further assess the quality of the new L4 product, statistical parameters characterizing

ASCAT wind retrievals and ECMWF equivalent stress winds are calculated from spatial and

temporal collocated data. The spatial distributions of the resulting parameters are shown in

Figure 5. The latter indicates that L4 winds exhibit better results (Figure 4) that those obtained

for ECMWF.




Date : January 2019

Issue : 1.2


Figure 4: Mean (left) and standard deviation (right) wind differences between collocated

ASCAT retrievals and L4 blended estimates during May 2017. The left panels indicate the

wind speed (a), zonal wind component (b), and meridional wind component (c) biases. The

corresponding std distributions are shown in d), e), and f) panels, respectively.




Date : January 2019

Issue : 1.2


L4 blended and ECMWF wind Comparisons (Long term quality control)

Figure 5: Mean (left) and standard deviation (right) wind differences between collocated

ASCAT retrievals and ECMWF equivalent stress wind estimates during May 2017. The left

panels indicate the wind speed (a), zonal wind component (b), and meridional wind

component (c) biases. The corresponding std distributions are shown in d), e), and f) panels,

respectively.

The quality control of geophysical content of each L4 blended wind file is performed

based on the calculation of the statistical parameters characterizing the difference between L4

and ECMWF (equivalent stress wind) wind speed, zonal, and meridional wind components.

Figures 6-8 show the time series of mean differences (biases) (top) and the related standard

deviations (middle), and of correlation coefficients (bottom), of wind speed, zonal, and




Date : January 2019

Issue : 1.2


meridional wind components, respectively. Times series are shown for the period September,

15, 2017 through May, 05 2018. The time variabilities of the statistical parameters do not lead

any significant trend in difference between L4 and ECMWF winds.

To assess the quality control of geophysical content of each L4 blended wind estimate, the

statistical parameters characterizing the difference between on one hand L4 and ASCAT

retrievals (L2b), and L4 and ECMWF (equivalent stress wind) on other hand are calculated.

Figures 9-11 show time series of mean differences (biases) (top) and the related standard

deviations (middle), and of correlation coefficients (bottom), respectively. The time

variabilities of the statistical parameters drawn from ASCAT and L4 comparisons are quite

steady. For instance, wind speed bias, standard deviation, and correlation are of 0.02 m/s,

1.0 m/s, and 0.96, respectively, along the study period (January 2018).




Date : January 2019

Issue : 1.2


Figure 6: Time series (September 2017 – May 2018) of statistical parameters characterizing

the comparisons between 6-houly L4 blended and ECMWF analyses estimated from global

data. Top through bottom panels show time series of biases (L4 minus ECMWF), standard

deviations (m/s), and correlation coefficients, respectively.




Date : January 2019

Issue : 1.2


Figure 7: As Figure 4 for zonal wind component.




Date : January 2019

Issue : 1.2


Figure 8: As Figure 4 for meridional wind component.




Date : January 2019

Issue : 1.2


Figure 9: Time series (January 2018) of statistical parameters (Bias (top panel), STD

(middle), and correlation (bottom)) characterizing the wind speed comparisons between on

one hand 6-hourly L4 blended and ASCAT L2b (in red color), and on other hand 6-houly L4

blended and ECMWF analysis (in blue color) estimated over free ice and land global ocean




Date : January 2019

Issue : 1.2


Figure 10: As Figure 7 for zonal wind component




Date : January 2019

Issue : 1.2


Figure 11: As Figure 7 for meridional component




Date : January 2019

Issue : 1.2


V SYSTEM’S NOTICEABLE EVENTS, OUTAGES OR CHANGES

Please include in section "SYSTEM’S NOTICEABLE EVENTS, OUTAGES OR CHANGES" any time where you know there was/is lack in satellite observation.

Date Change/Event description System version other

11/2018 The background used in this version is the

stress-equivalent wind vector (U10S)

estimated from ECMWF 10m wind vector

operational forecasts. The latter are provided

by KNMI. Moreover, SSMIS F18 and F19

are used, in addition to SSMIS F16 and F17,

as ancillary data aiming at the enhancement

of the remotely sensed wind observation

spatial and temporal sampling. The third

main change is the introduction of the new

variables wind vector curl and divergence,

and wind stress curl and divergence.

V5




Date : January 2019

Issue : 1.2


VI QUALITY CHANGES SINCE PREVIOUS VERSION

Please include a paragraph regarding quality changes since the previous version. It can be as paragraph or in the form of a table.

As expected, using new observations derived mainly from radiometers, improves the quality of V5 product compared to V4. The improvement is mostly depicted for high variable (in space and/or time) wind conditions.

Documents

QUALITY INFORMATION DOCUMENT For the Global Ocean Wind ...resources.marine.copernicus.eu/documents/QUID/CMEMS-WIND-QUI… · associated standard deviations do not exceed 1.30m/s and