Abstract :
[en] Half of world’s food comes from irrigated area that uses about 72% of available water resources. In Morocco, water availability is the main limiting factor for crop growth and final yield and it is becoming a national priority for the agricultural sector. This situation leads the stakeholders to define most favorable strategies in planning and management of available water resources, on one hand, and to assess accurately vegetation water content status, on the other hand, in order to improve irrigation scheduling and prevent water stress adversely affecting yield.
Remotely sensed reflectance has been used to estimate vegetation water content for different crops and to monitor water irrigation per surface unit, considering its high temporal and spatial resolution. In this study, we used two spectral indices of vegetation water content indicator (the Normalized Difference Infrared Index (NDII) and the Moisture Stress Index (MSI)) developed using Near Infrared (NIR) and Short Wave Infra-Red (SWIR) bands. The
study area is the irrigated perimeter of Tadla in Morocco (35% dominated by irrigated wheat crop).
In a first step, we compared observed vegetation water content of 16 studied plots of wheat and derived spectral indices NDII and MSI at the end of cropping season. The two images used at this step were acquired on March 26, 2013 and on April 11, 2013 when soil was fully covered by vegetation. Statistical analyses showed that the two spectral indices, NDII and MSI, simulated accurately vegetation water content. The statistical indicators, r, R², RMSE, nRMSE and MAE were -0.81, 0.65, 3.26% of water content (≈0.13 kg/m²), 4.26% and 2.69% for the NDII and 0.81, 0.65, 3.27% of water content (≈0.14 kg/m²), 4.27% and 2.72% for the MSI, respectively. To validate these results, we compared observed vegetation water content values and those predicted using the k-fold CV method. The errors were minimal for NDII and MSI, and the indicators of model evaluation obtained for predicted vegetation water content from NDII were: RMSE = 3.17%, nRMSE = 4.13%, MAE = 2.52% and R²=0.64. For MSI, these indicator were RMSE = 3.28%, nRMSE = 4.29%, MAE = 2.68% and R²=0.61. In a second step, we delimited the cereal area in the studied perimeter using a supervised classification method. The classification has been validated and the overall accuracy and Kappa coefficient were estimated respectively at 96.7% and 0.9545. Based on the regression model resulting from the comparison between NDII and measured vegetation water content, we produced maps of vegetation water content of wheat over the whole Beni-Moussa East irrigated area (41,000 hectares).
The results of this work demonstrated the potential of spectral indices (NDII and MSI) derived from SPOT5 satellite images data to quantify and map vegetation water content of wheat. It showed also the potential of the SWIR band to improve the monitoring of irrigation by mapping water stress of wheat at field and regional level.
