Probabilistic linkages of propagation from meteorological to agricultural drought in the North African semi-arid region

Younes Dahhane ^1* Victor Ongoma ¹ Abdessamad Hadri ¹ Mohamed Hakim Kharrou ¹ Oualid Hakam ¹ ABDELGHANI CHEHBOUNI ²

• ¹ international water research insitute, Mohammmed VI polytechnic university, Benguerir, Morocco
• ² Center for Remote Sensing Applications, Mohammed VI Polytechnic University, Ben Guerir, Morocco

Understanding the probability of drought occurrence in agricultural areas is important for designing effective adaptation strategies to drought impacts on agriculture and food security. This knowledge is critical, especially in arid and semi-arid areas of Morocco, which are prone and vulnerable to droughts.This study examines the linkage between meteorological drought (MD) and agricultural drought (AD) in a critical agricultural region in Morocco. Different agricultural drought indexes (NDVI anomaly, vegetation condition index (VCI), temperature condition index (TCI), vegetation health index (VHI)), and a meteorological drought index (Standardized Precipitation Evapotranspiration Index (SPEI) in different time scales (3, 6, 9, 12 months)) are assessed for the period 2000-2022. Statistical measures such as Spearman correlation (R), root mean square error (RMSE), and mean absolute error (MAE), are utilized to assess the performance of the meteorological drought index to detect the agricultural drought. The propagation time from meteorological drought to agricultural drought was identified, and probabilistic linkages between the two types of droughts were investigated using the copula function and Bayesian network. Results show that a combination of SPEI3 as meteorological drought index and VHI as agricultural drought index has the highest correlation coefficient of 0.65 and the lowest RMSE and MAE of 1.5 and 1.5, respectively. The propagation time from meteorological to agricultural drought was 39 days on a scale of 12 months, and seasonally, it was 29 days, 32 days, and 82 days, for autumn, winter, and spring, respectively. Bayesian network results show that agricultural droughts have the high probability to occur whenever there is severe and extreme meteorological drought, with the highest probabilities for mild and moderate agricultural drought. The findings have significant applications in water resource management and agricultural planning, for water usage and food security based on likelihood of agricultural drought occurence.