366849 Forecasting Seasonal Rainfall Characteristics in Rwanda Using NextGen Python-Based Climate Predictability Tool

Monday, 13 January 2020
Hall B1 (Boston Convention and Exhibition Center)
Asher Siebert, IRI, Palisades, NY; and M. Mbati, N. Acharya, A. Gahigi, and Á. Muñoz

As Rwanda is a primarily agrarian society, the timing and quantity of seasonal rainfall is of critical importance for food and livelihood security. The nation has a highly spatially variable climate because of large topographic differences and the complexity of the regional meteorology. This being said, the two main rainy seasons are in March-May and September-December. While the predominant moisture source for Rwanda is the Indian Ocean, some moisture is also advected from DR Congo especially during January-May. These complexities have made forecasting Rwanda’s rainfall a challenging enterprise. An additional confounding factor has been the limited meteorological station observation record from the mid-1990s to around 2010. In recent years, the ENACTS (Enhancing National Climate Services) initiative has been implemented in Rwanda to address the above-mentioned data gap and create a temporally and spatially continuous rainfall and temperature dataset for the country from the early 1980s to the present.

Recently, scientists at the International Research Institute for Climate and Society have developed a Python-based version of the Climate Predictability Tool (PyCPT), as part of the Next Generation climate forecast system (#NextGen). Meteorologists at Rwanda’s Meteorological Agency, MeteoRwanda, are adopting and applying this innovation. This paper illustrates the #NextGen approach to forecast total rainfall and rainy day frequency in Rwanda for the March-May and September-December rainy seasons. PyCPT is here used to create both single model and multi-model forecasts (employing climate model output from the suite of models from the North American Multi-Model Ensemble, NMME). Gridded rainfall from both the Climate Hazard Group InfraRed Precipitation with Station Data (CHIRPS) and ENACTS datasets are used as high resolution predictands in this study. Skill metrics are calculated and gridpoint probability of exceedance graphs are generated.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner