EVALUATION OF CHIRPS AND GPM-IMERG PRECIPITATION ESTIMATES AGAINST GROUND RAIN GAUGE OBSERVATIONS OVER LEBANON

Abstract

The objective of this paper is to assess two satellite precipitation products, Climate Hazards Center’s Infrared Precipitation with Stations (CHIRPS) and the product of the Global Precipitation Mission; the Integrated Multi-satellitE Retrievals for GPM (GPM IMERG v4) against ground rain gauges over Lebanon. A network of around 84 rain gauge stations distributed over the area was used after extensive data quality control. This evaluation focuses on annual precipitation from hydrologic year 2000-2001 till 2016-2017, and on a monthly basis from January 2000 till April 2018 since this period was the common data available from CHIRPS and Ground stations. As for the common period between Ground data and IMERG, it was from 2014 till 2017. On a yearly basis CHIRPS recorded better results than IMERG (R2CHIRPS= 0.353 whereas R2IMERG= 0.084). Monthly analysis showed good performance by both satellite products with general underestimation by IMERG and slight overestimation by CHIRPS during wet months. January is the rainiest month in Lebanon with an average monthly precipitation of 163 mm. Moreover, the effects of longitude, latitude, elevation and location (topography) on precipitation readings were examined. Longitude values were categorized into 3 classes whereas Latitudes were divided into four due to the longitudinal geographical borders of Lebanon and based on a precipitation map. Results revealed that annually CHIRPS performed better at “Long2” (from 35.5 to 36) and “Long3” (from 36 to 36.6) than “Long1” (from 35.1 to 35.5). Also it overestimated rainfall at “Long1” and “Long2” when Ground-observed rainfall was below 800 mm and underestimated it when it reached above 800 mm. IMERG did not record a specific trend. As for the effect of Latitude, CHIRPS and IMERG performed better over the South at “Lat 4” (from 34.3 to 34.7) with R2CHIRPS= 0.515, RECHIRPS= 22% and R2IMERG= 0.688, REIMERG= 13.3%. On a monthly basis, neither latitude nor longitude had a noteworthy effect on the bias between remotely sensed and ground-observed precipitation. As for the analysis of elevation influence, this aspect was divided into 3 categories C1 from 0 to 800m, C2 from 800m to 1500m and C3 > 1500m. Annual records showed underestimation by CHIRPS (around 10%) at C1 and C2 and for IMERG by 23% at C1 and 36% at C2 areas. However, at high elevations (C3) both products overestimated precipitation by 6% for CHIRPS and 2% for IMERG. On a monthly basis, January recorded as the rainiest month in Lebanon over C1 and C3 (162 mm). CHIRPS overestimated over all categories with better results during rainy months whereas IMERG underestimated over C1 and C2. CHIRPS showed slightly better results at low elevations (R2CHIRPS =0.669, RMSE CHIRPS = 43 and R2IMERG =0.657, RMSEIMERG = 32), however IMERG performed better at higher ones (R2CHIRPS =0.565, RMSECHIRPS =53 and R2IMERG =0.761, RMSEIMERG =32). Also with respect to location, average precipitation over the coast observed by ground stations was around 834 mm whereas that detected by CHIRPS was 766 mm. CHIRPS underestimated rainfall by around 10%. IMERG showed a better correlation with rain gauges than CHIRPS did (R2CHIRPS =0.196, RE CHIRPS = 23.5% and R2IMERG =0.678, REIMERG = 10.4%). Over the inland area, annual rainfall ranged between 373 mm and 1201 mm. CHIRPS and IMERG slightly overestimated precipitation with CHIRPS performing better (R2CHIRPS =0.621, RE CHIRPS = 20.5%. Lastly, over the mountains both products yielded lower estimates than ground observations that ranged between 410 mm and 1603 mm. On a monthly scale, CHIRPS and IMERG performed better over the coastal area (R2CHIRPS = 0.701 and R2IMERG = 0.756) than they did over the inland and mountainous ones. IMERG tends to overestimate over inland and underestimate rainfall over the mountains (higher % difference). This can be explained by the topography of Lebanon that is constituted of narrow plains followed by a complex terrain of relatively high elevation mountains. This type of landscape can hinder the proper functioning of both satellite products. These results highlight the ability of satellite products to record precipitation as well as the need for improving their predictions to detect light rain and snowfall. This work emphasizes the importance of accurate and good quality ground rainfall measurements in a small country like Lebanon known for its complicated topography and problematic geographical location.