466 Evaluation of Long-Term, Satellite-Based Rainfall Estimates in a Climatic Transition Zone in Equatorial Africa

Tuesday, 8 January 2019
Hall 4 (Phoenix Convention Center - West and North Buildings)
Jeremy Diem, Georgia State Univ., Atlanta, GA; and B. Konecky, J. Salerno, and J. Hartter

The millions of rural households in tropical Africa are impacted greatly by rainfall variability. Unfortunately, ground-measured rainfall data in the region has been decreasing rapidly over the past several decades, which prevents a better understanding of rainfall variability. Station data are now sparse within the region, including large areas of central Africa. But there now exist multiple, satellite-based rainfall products that provide daily rainfall estimates from 1983 to the present at relatively fine spatial resolutions for the region. Therefore, the purpose of this paper is to evaluate the suitability of estimates from four satellite-based rainfall products for western Uganda, an area adjacent to station-poor central Africa and unique with a relatively dense network of rainfall stations that can be used for product validation. The four products are African Rainfall Climatology Version 2 (ARC2), Climate Hazards Group InfraRed Precipitation with Stations (CHIRPS), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks – Climate Data Record (PERSIANN-CDR), and TAMSAT African Rainfall Climatology And Timeseries (TARCAT). The bias and accuracy of ten-day, monthly, and seasonal rainfall totals were assessed using approximately ten years of data from eight rain gauges. The ability of the products to adequately capture intra-seasonal and inter-annual rainfall also was examined. CHIRPS was the most accurate product at capturing intra-seasonal rainfall variability and estimating ten-day, monthly, and seasonal rainfall totals; it can be considered sufficiently accurate at estimating seasonal rainfall totals throughout most of the region. TARCAT was the most reliable product when inter-annual variability was involved: it had the largest correlations with ground-measured rainfall and it did not have any significant negative change points in its time series. TARCAT did have an underestimation bias, but its lack of reliance on “real time” rain-gauge data makes it much more temporally stable than the other products. Consequently, inter-annual analyses involving TARCAT were the least biased. ARC2 was arguably the least reliable product due to the fact that it is unsuitable for assessing inter-annual variability, especially multi-decadal variability, in rainfall. ARC2 showed a significant negative decrease in rainfall from 1983-2016 in western Uganda; however, this trend was caused by artificially decreased rainfall totals after 1991. The introduction of a dry bias in ARC2, and to a lesser degree CHIRPS, was most likely caused by the major decrease in rain-gauge data over the past several decades. It is actually highly likely that western Uganda has experienced increasing annual rainfall totals, rather than a drying trend, since the early 1980s.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner