Bias-Correction of Satellite-Based Precipitation Estimates in Frequency Domain
Abstract
Over the last two decades, numerous satellite-based precipitation estimates (SPE) with a high spatio-temporal resolution, global coverage, and real-time availability have been developed. SPEs have shown immense utility in hydro-meteorological and agricultural applications. However, SPEs are indirect estimates of rainfall (derived from cloud properties) and may show divergence with the ground-based rain-gauge measurements. Significant research efforts are being made to identify and remove the sources of error in SPEs. Currently, to deal with this problem, a variety of bias-correction techniques have been developed. These methods calibrate a transfer function between SPEs and observed rain-gauge observations over the historical period and then post-process the SPEs by applying this transfer function to real-time SPEs. In this study, we propose a quantile mapping (QM) and signal decomposition-based (SD) technique that performs bias-correction of SPEs in the frequency domain. The proposed QM-SD bias correction technique is applied to correct Integrated Multi-Satellite Retrievals for GPM (IMERG) Early precipitation estimates over India. The gauge-based daily gridded dataset of the Indian Meteorological Department (IMD) is considered as the reference. The algorithm is trained for 13 years (2001-2013) and tested for 6 years (2014-2019). The results of this study show a significant reduction in the mean bias and RMSE of the IMERG Early dataset after the application of the QM-SD bias-correction technique in the training as well as testing periods. The all-India annual mean bias in IMERG Early reduced by 77% during the training period and 49% during the testing period. Similarly, the all-India annual mean RMSE reduced by 12% and 19% during the training and testing period respectively. The bias-correction of IMERG Early also leads to improved representation of different extreme precipitation characteristics. Additionally, better performance of the proposed QM-SD technique is observed when compared to three widely adopted bias-correction techniques. The QM-SD bias-correction approach shows promising results and significantly aid in correcting SPEs for real-time hydrological and climatological predictions.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.H15T1272C