Statistical modeling and mapping of rainfall in the endorheic basins of Northern Algeria: a comparison of spatial interpolation methods

Abstract

The creation of rainfall maps often relies on preliminary spatial interpolation. Significant deviations from the real distribution of rainfall are likely to occur due to the wide variety of interpolation methods, the effectiveness of which may vary. The main objective of this study is to find the best interpolation method to estimate the spatial distribution of rainfall in the Chott El Hodna basin (25,834 km²), typical of endorheic basins in Algeria. The rainfall database consisted of 42 years of monthly observations from 52 stations (1975–2017). Eight spatial interpolation models were compared, six of which were deterministic and two stochastic. Deterministic models include nearest neighbor, inverse distance weighting, local polynomial, minimum curvature, thin plate spline, and natural neighbor. The stochastic models are ordinary kriging and regression-kriging (RK). RK uniquely incorporates additional information about the geotopographical environment of the basin. The forecasting performance of each method was evaluated using statistical cross-validation indicators, as well as visual analysis and comparison with previously published isohyet maps. The evaluation concluded that the RK model is the most appropriate for producing a map of annual mean rainfall in the Chott El Hodna basin. In addition, this map covering the period from 1975 to 2017 revealed a significant average drop in rainfall: 31% compared to the periods 1913–1938 and 1913–1963, and 24% relative to the period 1922–1960/1969–1989. Further research is needed to determine the causes of these trends, assess their long-term impact, and develop effective adaptation strategies.