Assessment of Satellite Datasets for Rainfall Predictions based on Canonical Correlations

Climate predictions are essential tools for reducing the vulnerability of climate variability risks. For this reason, National Meteorological and Hydrological Services (NMHSs), supported by the World Meteorological Organization (WMO), implemented the regular operation of 19 Regional Climate Outlook Forums (RCOFs) around the world since 1997. The RCOFs are operational platforms that provide consensual-regional climate predictions over climatologically homogenous zones. In Central America, South America, and the Caribbeans RCOFs, teleconnections modeling through Canonical Correlation Analysis (CCA) between weather station observations and Sea Surface Temperatures (SSTs) is the most widely applied method to generate seasonal rainfall forecasts. However, the RCOFs approach faces a decrease in weather station data availability and limitations in providing sub-seasonal forecasts. This research aims to assess the reliability of applying satellite data for sub-seasonal rainfall forecasts based on CCA. Satellite information can be advantageous to weather station observations as it provides a continuous high spatial resolution dataset with full coverage for at least 20 years. To investigate the use of satellite products, we used Panama Republic as the forecast domain and the Global SST as the CCA predictor. The forecast process was performed for 6 cases applying different predictand datasets: IMERG Early, IMERG Final, and CHIRPS by alternating the forecast length from 1 month to 3 months. Lastly, the CCAs prediction skill of each case were compared with the traditional forecast, which relies on weather station observations. Considering that the population and stakeholders need more information at intra-seasonal timescales with good spatial resolution, this study demonstrates the potential of satellite remote sensing data in RCOFs operational seasonal and sub-seasonal forecasts at 0.1 X 0.1 resolution.