A Composite Approach to Assessing Similarity in the Risk Level of Agricultural Drought: An Example of the Tensift and Moulouya Watershed in Morocco

Abstract

In this chapter, a per-pixel trend analysis approach and the temporal variance of drought risk were used to compare the seasonal dynamics of the multivariate risk level of agricultural drought in two hydrological systems (Tensift and Moulouya watersheds in Morocco). The multivariate risk level is calculated from the time series of the enhanced composite model for agricultural drought monitoring constructed from anomalies in precipitation, evapotranspiration, soil moisture, NDVI and land surface temperature (LST). These five factors of agricultural drought are obtained from multi-sensor remote sensing data (MODIS, CHIRPS) for the period 2004 to 2022. In both watersheds, the results show a significant upward trend in the level of extreme and moderate risk and a very contrasting seasonal variance. At the same time, over the past two decades, risk levels (normal and low) have shown a marked downward trend. The multivariate frequency of generalized extreme risk is identical in both hydrological systems. Over the data period, the seasonal occurrence of multivariate risk is 6/19 (extreme risk), 7/19 (moderate risk) and 6/19 (low to normal risk). The recent spatio-temporal dynamics of the level of extreme risk have significant negative correlations (−0.6 at Tensift and −0.62 at Moulouya) with cereal yield anomalies. Multivariate risk is positively correlated with the SPI index at the seven-month scale of the growing season. The maximum correlation is 0.86 with a p-value of 0.0000 at Tensift and 0.74 with a p-value of 0.0004 at Moulouya. Depending on the year, the multivariate extreme risk can vary from 0 to 100% in terms of exposed areas in both the Tensift and the Moulouya watersheds. It is generalized over all the spatial extents of these watersheds for the years 2022, 2020, and 2016 and non-existent for the years 2004, 2006, 2009, 2010, 2013, and 2015.