Niederschlagsfernerkundung (RS)
Zusammenfassung der Projektergebnisse
Precipitation is a key variable in the global hydrological cycle. Obtaining area-wide precipitation information remains a challenge, especially in remote and high-altitude regions such as the Tibetan Plateau (TP), the study area of the prime bundle project. Satellite-based precipitation products can provide such information at high spatial and temporal resolution. Existing methods using geostationary (GEO) satellite systems that meet the requirements of high spatio-temporal resolution are mostly based on the use of one infrared channel at 10.8 µm, which leads to inaccuracies in terms of derived precipitation information. Therefore, the main objective of the subproject "Remote sensing of precipitation" (prime-RS) was to develop a new method for reliable precipitation derivation over the TP by combining multispectral GEO data with microwave (MW) based precipitation information from the GPM IMERG product using machine learning techniques. In addition to the derivation of precipitation, the discrimination of precipitation phase is of great importance for the TP. Therefore, another objective was to investigate the possibility of satellite-based discrimination into solid and liquid precipitation. First, a method for delineating the precipitation area using the GEO systems Elektro-L2 and Insat-3D was developed. For this purpose, Random Forest (RF) models were trained using the multispectral GEO data as predictors and precipitation area information from the MW-only sub-dataset of the IMERG product as target variable. Adjusting the ratio between the number of raining and non-raining pixels during the training process resulted in an improvement in terms of detected precipitation area. The validation results using independent data demonstrate an improved precipitation area delineation compared to the IR-only sub-dataset of the IMERG product based on the 10.8 µm IR channel. Building on the precipitation area detection, the next step was to develop a method for deriving the precipitation rate. For this purpose, RF regression models were trained on a scene basis incorporating the multispectral GEO data and the MW-only precipitation dataset of the IMERG product. An undersampling approach was used to adjust the relationship between precipitation intensities of different frequencies, resulting in an improvement in the derived precipitation rate. The validation results show good performance of the final precipitation product at both 11 km and 4 km resolution and a significant improvement over the IR-only precipitation sub-dataset of the IMERG product. The tendency to overestimate precipitation can be explained by the overestimation of the precipitation area. In contrast, averaging the precipitation rate by the RF regression approach leads to an underestimation of extreme precipitation rates. The studies on precipitation phase discrimination highlight the great importance of a reliable reference data set for RF model training. Given the lack of studies regarding a comprehensive and reliable assessment of the GPM DPR snow product, further research is needed in this area. This requires reliable ground station measurements with high temporal resolution and good spatial coverage, which were not available for the TP. A comparison of the GPM DPR snow product with the High Asia Refined analysis version 2 snowfall product for different temperature classes showed good agreement between both data sets for ground temperatures below 1 °C. Due to its high spatio-temporal resolution, the developed multispectral precipitation derivation method offers the potential of spatio-temporal analyses of precipitation patterns and investigation of the influence of the dominant atmospheric flow conditions and pressure centers on precipitation dynamics and distribution over the TP.
Projektbezogene Publikationen (Auswahl)
- (2019): Precipitation Retrieval over the Tibetan Plateau from the Geostationary Orbit — Part 1: Precipitation Area Delineation with Elektro-L2 and Insat-3D. Remote Sensing 11(19), 2302
Kolbe, C., Thies, B., Egli, S., Lehnert, L., Schulz, M. & Bendix, J.
(Siehe online unter https://doi.org/10.3390/rs11192302) - (2020): Intercomparison of Gridded Precipitation Datasets over a Sub-Region of the Central Himalaya and the Southwestern Tibetan Plateau. Water 12(3271), 1-23
Hamm, A., Arndt, A., Kolbe, C., Wang, X., Thies, B., Boyko, O., Reggiani, P., Scherer, D., Bendix, J. & Schneider, C.
(Siehe online unter https://doi.org/10.3390/w12113271) - (2020): Precipitation Retrieval over the Tibetan Plateau from the Geostationary Orbit—Part 2: Precipitation Rates with Elektro-L2 and Insat-3D. Remote Sensing 12(13), 2114
Kolbe, C., Thies, B., Turini, N., Liu, Z. & Bendix, J.
(Siehe online unter https://doi.org/10.3390/rs12132114)
