Satellite-based remote sensing of rainfall in areas with sparse gauge networks and complex topography
Rainfall is an essential parameter in the analysis and research of water resource management. However, the complexity of rainfall combined with the uneven distribution of ground-based gauges and radar in developing countries’ mountainous and semi-arid areas limits its investigation. In this context,...
|Online Access:||PDF Full Text|
No Tags, Be the first to tag this record!
|Summary:||Rainfall is an essential parameter in the analysis and research of water resource management. However, the complexity of rainfall combined with the uneven distribution of ground-based gauges and radar in developing countries’ mountainous and semi-arid areas limits its investigation. In this context, satellite-based rainfall products provide area-wide precipitation observations with a high spatio-temporal resolution, engaging them in hydrological management in ungauged basins.
Therefore, in this study, I investigated method to establish a satellite-based rainfall algorithm for ungauged basins. The algorithm combines the new Integrated Multi-SatEllite Retrieval for the Global Precipitation Measurement (GPM) (IMERG) rainfall products and second-generation geostationary orbit (GEO) systems developing rainfall retrieval techniques with the high spatio-temporal resolution using machine learning algorithms.
For the first step, microwave satellite and Meteosat Second Generation Spinning Enhanced Visible and Infrared Imager (MSG SEVIRI) data for Iran were collected to develop a regionally based new rainfall retrieval technique. The method used geostationary multispectral infrared (IR) data to train Random forest (RF) models. I employed the microwave (MW) rainfall information from the IMERG as a reference for RF training. The rainfall area was delineated in the first step, followed by rainfall rate assignment. The validation results showed the new technique’s reliable performance in both rain area delineation and rain estimate, particularly when compared to IR-only IMERG. Multispectral IR data improves rainfall retrieval compared with one single band.
In the next step, I investigated the applicability of the developed algorithm in Ecuador with different orography and rainfall regimes compared to Iran. For this aim, I used the Geostationary Operational Environmental Satellite-16 (GOES-16) as the GEO satellite, which covers Ecuador at a suitable angle. The feature selection and algorithm tuning were performed to regionalize the models for Ecuador. The validation results show the reliable performance of the method in both rain area delineation and rain estimation in Ecuador. The results proved the suitability of the developed algorithm with different GEO systems and in different regions. Some inaccuracies at the Andes’ high elevation were evident after the spatial analysis of the validation indices. Evaluating the validation results against a high spatio-temporal radar network showed that the developed algorithm has difficulty capturing drizzles and extreme events dominant in the Andes’ high elevations and needs improvement.
In summary, this research presents a new satellite-based technique for rainfall retrieval in a high spatio-temporal resolution for ungauged regions, which can be applied in parts of the world with different rainfall regimes. This findings could be used by planners and water managers regardless of the availability of rain gauges at ground. Furthermore, the research showed, for the very first time, the advantage of using the new generation of GEO satellite combined with microwave satellites integrated in GPM IMERG for estimating rainfall.|
|Physical Description:||186 Pages|