On the Temporal Variability of Precipitation in Iraq: Arid‐Wet Years and Extreme Events

ABSTRACT Here we examine the variability of precipitation in Iraq, focusing on the differences between arid and wet years as well as on the intensity of extreme events. This is done analysing several long precipitation records—84 years of monthly precipitation in four stations (1938–2023, except 2003 and 2004) and 19 years of daily precipitation data (2005–2023) in the Baghdad station—together with reanalysis weather data. The long time series gives annual (mean ± standard deviation) precipitation values of 367 ± 140, 107 ± 59, 138 ± 65 and 136 ± 60 mm in Mosul, Rutba, Baghdad and Basrah, respectively, with a weighted mean of 182 ± 57 mm for the entire country. The time series are dominated by seasonality, with maximum winter and negligible summer precipitation, and one‐order‐of‐magnitude interannual variations. An annual‐precipitation probability density function distinguishes wet and arid years—above and below the 25 and 75 percentiles, respectively—with substantial differences in the seasonal cycle. The monthly precipitation is significantly correlated with the Indian Ocean Dipole (IOD) index, suggesting that the autumn‐winter arrival of moisture from the Red Sea, Arabian Sea and Persian Gulf enhances precipitation. Combining the Baghdad daily record with weather reanalysis data allows identifying 36 extreme precipitation events that are classified as jet streak, cut‐off low, upper‐air trough or Rex block events. These episodes account for 1452 mm of precipitation, or 56% of the total for the 2005–2023 period, but do not increase their frequency during wet years, hence suggesting that the ultimate cause for the arid‐wet year variability is the seasonal IOD conditioning. Finally, an extreme value analysis for both the daily and monthly data—with both peak over threshold (POT) and block maxima (BM) methods—provides return values for periods up to 100 years. For daily data, both methods give similar mean values but POT provides much narrower high‐confidence bands; for monthly data, both methods give similar mean and high‐confidence values.