**2.3 Agroclimatic indicators**

**Table 2** provides details on the 14 agroclimatic indicators calculated in this study. These include seasonal rainfall characteristics (cumulative seasonal rainfall, number of rainy days, rainfall intensity, extreme rainfall, dry sequences, start, end and length of the rainy season). A dry sequence is obtained by counting the number of consecutive days without rainfall that lie between two rainy days (a day is rainy when


#### **Table 2.**

*Definitions of the agroclimatic indicators used.*

its cumulative rainfall is greater than or equal to 1 mm). These are intra-seasonal droughts, also called rainfall breaks. We have defined four types: DSs, DSm, DSl and DSxl (see **Table 2**). The dry sequences that determine the potential risks of reseeding are the longest rainfall breaks observed over the 30 days after sowing [29]. Here, DSl and DSxl present themselves as effective indicators for detecting false starts to the season [18]. Next, in order to assess the ability of global models to adequately represent the agronomic start of the rainy season, we used Sivakumar's [30] definition of local rainy season start. Based on an agro-nomic criterion, he considers the start of the rainy season in the Sahelian and Sudanian regions as the date, starting on May 1, when there is a rainfall of at least 20 mm on three consecutive days, with no dry spells of more than seven days in the following 30 days. The end of the agronomic season is defined by considering the date after September 1 when a long dry period of 20 days occurs. The duration of the agronomic season is defined for a given season as the difference between the start and end date of the rainy season according to the agronomic criterion. Finally, indicators illustrating the intensity (R95 and R99) and frequency (R10 and R20) of rainfall events are used in addition to the cumulative season (PRCTOT).
