PRODUCT FACT SHEET: NDWI EUROPE Versin 1 (Dec. 2011) NDWI: Nrmalized Difference Water Index Type Tempral scale Spatial scale Ge. cverage Vegetatin respnse 10-day prduct 1 km Eurpe DESERT Actin LMNH Unit - 1 -
During drught event, vegetatin canpy can be affected by water stress. This can have majr impact n the plant develpment in general and can cause crp failure r lwer crp prductin in agricultural areas. Early recgnitin f plant water stress can be critical t prevent such cnsequences. By prviding near-real time infrmatin n the plant water stress t the stakehlders, water and agricultural management can be much imprved, ntably by irrigating specifically areas where plant water needs are nt fulfilled anymre. The Nrmalized Difference Water Index (NDWI) is knwn t be strngly related t the plant water cntent. It is therefre a very gd prxy fr plant water stress. The Nrmalized Difference Water Index (NDWI) (Ga, 1996) is a satellite-derived index frm the Near-Infrared (NIR) and Shrt Wave Infrared (SWIR) channels. The SWIR reflectance reflects changes in bth the vegetatin water cntent and the spngy mesphyll structure in vegetatin canpies, while the NIR reflectance is affected by leaf internal structure and leaf dry matter cntent but nt by water cntent. The cmbinatin f the NIR with the SWIR remves variatins induced by leaf internal structure and leaf dry matter cntent, imprving the accuracy in retrieving the vegetatin water cntent (Ceccat et al. 2001). The amunt f water available in the internal leaf structure largely cntrls the spectral reflectance in the SWIR interval f the electrmagnetic spectrum. SWIR reflectance is therefre negatively related t leaf water cntent (Tucker 1980). Its usefulness fr drught mnitring and early warning has been demnstrated in different studies (e.g., Gu et al., 2007; Ceccat et al., 2002). It is cmputed using the near infrared (NIR) and the shrt wave infrared (SWIR) reflectance (Eq.1), which makes it sensitive t changes in liquid water cntent and in spngy mesphyll f vegetatin canpies (Ga, 1996 ; Ceccat et al., 2001). 1. Prduct The Nrmalized Difference Water Index (NDWI) is a remte sensing derived index estimating the leaf water cntent at canpy level. Gegraphic cverage: available fr Eurpe Spatial scale: 1.2km Tempral scale: every 10 days aligned n the first day f each mnth, which crrespnds t 3 images per mnth (day 1-10, day 11-20, day 21-last day f mnth). - 2 -
Data surce: MODIS spectral bands 2 and 6 are prvided by the German Aerspace Centre (DLR) and pre-prcessed by the FOREST Actin (IES, JRC). Frequency f data cllectin: every day 2. Methdlgy 2.1 Calculatin f NDWI The NDWI is a remte sensing based indicatr sensitive t the change in the water cntent f leaves (Ga, 1996). NDWI is cmputed using the near infrared (NIR MODIS band 2) and the shrt wave infrared (SWIR MODIS band 6) reflectance s. The EFFIS Delivery Data Prducts delivered by the German Aerspace Centre (DLR) are used t cmpute the NDWI. MODIS data are delivered per tile in ETRS_LAEA crdinate reference system (Figure 1). Figure 1- Subdivisin in tiles f the MODIS data. Fr each tile, a maximum value time cmpsite is prduced every 10-day using the daily MODIS cmpsite files that cmbine all TERRA r AQUA passes f the day. Each mnth is split in 3 perids: frm the 1st t the 10th, frm the 11th t the 20th, and frm the 21th t the end f the mnth. The 10-days cmpsite are then resampled frm 250 m t 1 km and the rati between band 2 and band 6 is then calculated fllwing the equatin here abve. The different tiles are then msaicked t prduce a unique map fr Eurpe. Finally, a mask derived frm GLC 2000 is applied t recde - 3 -
water bdies, cities, permanent snw and desert as n data. 2.2 Calculatin f NDWI anmaly Anmaly estimatin: NDWI anmalies are prduced fr every 10-day perid as fllwed: where is the NDWI f the 10-day perid t f the current year and, is the lng-term average NDWI and δ is the standard deviatin, bth calculated fr the same 10-day perid t using the available time series (see baseline statistics). NDWI anmalies are prduced nly fr pixels that have at least five years f data fr the given 10-day perid. Baseline statistics: The archive f NDWI cvers the perid frm January 2006 t current day. NDWI and NDWI anmalies can be presented in the frm f maps and graphs, prviding infrmatin bth n the spatial distributin f the vegetatin water stress and its tempral evlutin ver lnger time perids. Gridded data can easily be aggregated ver administrative r natural entities such as hydrlgical watersheds. This allws fr the qualitative and quantitative cmparisn f the intensity and duratin f the NDWI anmalies with recrded impacts such as yield reductins, lw flws, lwering f grundwater levels, t cite but a few. The NDWI prduct is dimensinless and varies between -1 t +1, depending n the leaf water cntent but als n the vegetatin type and cver (Figure 2). High values f NDWI (in blue) crrespnd t high vegetatin water cntent and t high vegetatin fractin cver. Lw NDWI values (in red) crrespnd t lw vegetatin water cntent and lw vegetatin fractin cver. In perid f water stress, NDWI will decrease. Figure 2 NDWI f the 3 rd 10-day perid f May 2011. - 4 -
The NDWI anmaly prduct is given in standard deviatin units. It is cmmnly ranging frm -4 t +4 and frm red t green, red shwing negative anmalies (Figure 3). Figure 3 NDWI anmaly f the 3 rd 10-day perid f May 2011. The EDO Map server displays the latest available NDWI 10-day cmpsite image and the NDWI anmaly image calculated by cmparing this image t the histrical series in the same 10-day perid. Bth prducts are easy t read. The interpretatin must take int accunt the fact that this indicatr is shwing a variatin in the vegetatin cnditins. 1. Strength & weaknesses [+] Every ten days, the NDWI gives a systematic and spatially cntinuus picture f the vegetatin water stress at a high spatial reslutin (1km) fr the entire Eurpe. [-] Sil backgrund effects can be imprtant in case f partial vegetatin cver (Ga, 1996). Mrever snw cver is nt masked ut in the current versin. During winter mnths NDWI > 0.5 may crrespnd t snw cver. [-] Drught and water stress are nt the nly factrs that can cause a decrease f NDWI values/anmalies. Change in land cvers r pests and diseases can als be respnsible fr such variatin f the signal. Therefre this indicatr must be used jintly with ther indicatrs giving infrmatin n the deficit f rainfall /sil misture in rder t determine if the variatin in the vegetatin respnse (signal) is linked with a drught event r nt. [-]Anmalies are dependent f the time series available t calculate the mean values and the standard deviatins. This perid shuld be lng enugh t characterize the area where the index is calculated. The length f current archive (6 years) is t shrt t ensure fully the accuracy f the NDWI anmalies. - 5 -
2. Perfrmance f the indicatr (reference t literature, reprts) Accrding t Ga (1996), NDWI is a gd indicatr fr vegetatin liquid water cntent and is less sensitive t atmspheric scattering effects than NDVI (Nrmalized Difference Vegetatin Index). MODIS NDWI has been used t detect and mnitr the misture cnditin f vegetatin canpies ver large areas (e.g. Delbart et al. 2005, Jacksn et al. 2004) and tested as a drught indicatr (Gu et al. 2008). Gu et al (2007) fund that NDWI values exhibited a quicker respnse t drught cnditins than NDVI. Cntrary t NDWI, NDVI has limited capability fr retrieving vegetatin water cntent infrmatin, since prvides infrmatin n vegetatin greenness (chlrphyll), which is nt directly and unifrmly related t the quantity f water in the vegetatin (Ceccat et al 2002). Ceccat, P., Flasse, S., Tarantla, S., Jacquemnd, S., and Gregire, J.M. 2001. Detecting vegetatin water cntent using reflectance in the ptical dmain. Remte Sensing f Envirnment 77: 22 33. Ceccat, P., Flasse, S., and Gregire, J.M. 2002. Designing a spectral index t estimate vegetatin water cntent frm remte sensing data: Part 2. Validatin and applicatins. Remte Sensing f Envirnment 82: 198-207. Delbart, N., L. Kergat, T. L. Tan, J. Lhermitte, and G. Picard (2005). Determinatin f phenlgical dates in breal regins using nrmalized difference water index. Remte Sensing f Envirnment 97: 26-38 Ga, B.-C. 1996. NDWI - A nrmalized difference water index fr remte sensing f vegetatin liquid water frm space. Remte Sensing f Envirnment 58: 257-266. Gu, Y., Brwn, J.F., Verdin, J.P., and Wardlw, B. 2007. A five-year analysis f MODIS NDVI and NDWI fr grassland drught assessment ver the central Great Plains f the United States. Gephysical Research Letters 34. Gu, Y., Hunt, E., Wardlw, B., Basara, J.B., Brwn, J.F., Verdin, J.P. 2008. Evaluatin f MODIS NDVI and NDWI fr vegetatin drught mnitring using Oklahma Mesnet sil misture data. Gephysical Research Letters 35. Jacksn, T.J., Chen, D., Csh, M., Li, F., Andersn, M., Walthall, C., Driaswamy, P., and Hunt, E.R. 2004. Vegetatin water cntent mapping using Landsat data derived nrmalized difference water index fr crn and sybeans. Remte Sensing f Envirnment 92: 475 482. Tucker, C. J. 1980. Remte sensing f leaf water cntent in the near infrared. Remte Sensing f Envirnment 10: 23-32. - 6 -