Abstrakt
Madagascar is among the countries whose agriculture is heavily dependent on rainfall. However, the country lacks accurate and reliable early warning systems for droughts and floods, partly due to insufficient station rainfall data.
The purpose of this study is to identify rainfall datasets that can complement observation data by appraising 15 datasets (gauge-based, reanalysis, and satellite estimates). The study compares the temporal and spatial performance of datasets at annual and seasonal scales during 1983-2015.
In all the analyses, CHIRPS presents lower biases, so it is chosen as the reference data in the Taylor diagram for the final evaluation analysis. Even though ranking datasets is neither possible nor appropriate since each dataset performs differently throughout each analysis, some datasets show reasonable consistency.
This is the case with MSWEP, ERA5, and UDEL. On the other hand, MERRA2, CMAP, and TAMSAT are least preferred for use due to their considerable biases (specifically TAMSAT during the dry season).
CRU, PRECL, ERAINT, CFSR, and JRA55 also present some degrees of deficiencies at either annual or seasonal scales. These findings are crucial for any future rainfall analysis over the country in order to minimize inaccuracy in monitoring rainfall.