Performance evaluation of CHIRPS satellite precipitation estimates over Turkey

Satellite-based precipitation data can be a valuable source for performing hydro-climatological analyses: such as drought and water balance estimations with long and temporally consistent data. In this study, the Climate Hazards Group InfraRed Precipitation with Stations (CHIRPS) precipitation estimations were compared with station-based precipitation measurements at daily, dekadal, and monthly time scales using statistical and categorical validation measures. The assessment study includes 77 ground-based meteorological stations distributed throughout Turkey, and the study period corresponds to years 2008 through 2018. Overall evaluation indicates that CHIRPS estimates exhibit high correlation on dekadal and monthly time scales; however correlation decreases for daily estimates. A positive bias is found over Turkey, and a particular overestimation was detected between the precipitation amounts of 0-25 mm/dekad and 0-80 mm/month. CHIRPS tends to underestimate high precipitation amounts of 25-80 mm/dekad and 150-300 mm/month. CHIRPS estimations show the best performance in winter when Turkey's precipitation regime is dominated by cyclones and the lowest performance in the spring season. The study area was delineated into six regions according to their precipitation climatology and hydrological basins borders. The best performance for monthly CHIRPS estimates was in western Anatolia (r=0.88). For all regions, precipitation detection capability is significantly high not only for monthly but also for dekadal CHIRPS estimates. Probability of detection was found between 0.83 and 0.98, and the false alarm rate changes between 0.30 and 0.13 for the thresholds of 50 mm/month and 5 mm/month, respectively. In rainfall categories of 30, 40, and 50 mm / month the capacity to determine rain and no rain was significantly high. CHIRPS estimations for 54 meteorological stations were reasonably biased (between +/- 20%) with ground-based observations. Results show that CHIRPS estimations over Western Anatolia, Southern Anatolia, and Marmara regions are more consistent than the mountainous eastern and northeastern parts of the country. CHIRPS estimates can be used for hydro-climatological studies such as drought and water balance modeling over Turkey considering the error characteristics presented in this study.