How certain is desiccation in west African Sahel rainfall (1930–1990)?

Abstract

Hypotheses for the late 1960s to 1990 period of desiccation (secular decrease in rainfall) in the west African Sahel (WAS) are typically tested by comparing empirical evidence or model predictions against “observations” of Sahelian rainfall. The outcomes of those comparisons can have considerable influence on the understanding of regional and global environmental systems. Inverse‐distance squared area‐weighted (IDW) estimates of WAS rainfall observations are commonly aggregated over space to provide temporal patterns without uncertainty. Spatial uncertainty of WAS rainfall was determined using the median approximation sequential indicator simulation. Every year (1930–1990) 300 equally probable realizations of annual summer rainfall were produced to honor station observations, match percentiles of the observed cumulative distributions and indicator variograms and perform adequately during cross validation. More than 49% of the IDW mean annual rainfall fell outside the 5th and 95th percentiles for annual rainfall realization means. The IDW means represented an extreme realization. Uncertainty in desiccation was determined by repeatedly (100,000) sampling the annual distribution of rainfall realization means and by applying Mann‐Kendall nonparametric slope detection and significance testing. All of the negative gradients for the entire period were statistically significant. None of the negative gradients for the expected desiccation period were statistically significant. The results support the presence of a long‐term decline in annual rainfall but demonstrate that short‐term desiccation (1965–1990) cannot be detected. Estimates of uncertainty for precipitation and other climate variables in this or other regions, or across the globe, are essential for the rigorous detection of spatial patterns and time series trends.