Analysis of rainfall trend and the impact of teleconnection patterns on its variability: case of Medjerda Transboundary Watershed

Medjerda is a key transboundary watershed (Algeria and Tunisia) in the Maghreb region.  The analysis of rainfall regime in this basin is of paramount importance for water resources management and for regional economic development notably in relation with agriculture. This study examines the rainfall trends over the Medjerda watershed on a multi-temporal scale (monthly, seasonally and annually), taking also the spatial scale into consideration. A database of monthly rainfall observed in 60 stations evenly spread over the watershed is used for this purpose. After filling gaps and homogenisation of data, the Mann-Kendall (MK) test for trend detection was applied to rainfall time series. Prior to the trend detection, a Trend-Free Pre-Whitening (TFPW) procedure has been applied to ensure the independence of the data. Moreover, the Sen’s slope method was adopted to estimate the magnitude of rainfall trends over the catchment. Rainfall values in the different stations were then used to generate a map of interpolated Sen’s slope in order to explore the spatial distribution of rainfall trend within the watershed. The results showed that there is a decreasing trend in annual rainfall over the whole basin, yet non statistically significant. Nevertheless, at smaller scales, there is more rainfall trends detected (increasing or decreasing) for different seasons or months, showing the importance of analyzing also monthly or seasonal trends that may differ from annual trends and have some important consequences in terms of water management . Finally, the influence of different teleconnection patterns (the North Atlantic Oscillation, NAOi, the South Oscillation, SOIi and the Mediterranean Oscillation, MOi) on rainfall trend were also investigated. The analysis showed weak correlations at the seasonal scale. Indeed, NAOi and MOi have a similar effect on winter rainfall demonstrated by a positive correlation, while SOi is negatively correlated with spring rainfall. These results highlight the potential small-scale impacts of large-scale climate oscillations on rainfall trends.