Assessing the changeability of precipitation patterns using multiple remote sensing data and an efficient uncertainty method over different climate regions of Iran

While the rain gauge measurements are unevenly distributed in many world regions, it is necessary to use remote sensing-based precipitation data at high spatial and temporal resolutions. On the other hand, quantifying the uncertainty of precipitation is a vital issue for hydrometeorological applications globally. For this purpose, applying a new fusion framework using high-resolution remote sensing datasets can provide accurate precipitation evaluation against local observations with low uncertainty. This study examines three weighted fusion-based models containing the Ordered-Weighted-Averaged (OWA) family approach based on the ORLIKE method (OWA-ORLIKE) and ORNESS method (OWA-ORNESS) as well as the Entropy-weight (EW) method to combine different remote sensing precipitation products over different climate zones of Iran. In this case, multiple monthly remotely sensed datasets, including ERA5, ERA5-Land, TerraClimate, GPM, PERSIANN-CDR, TRMM, and CHIRPS, are utilized to assess precipitation patterns versus local measurements, which were gathered by Google Earth Engine (GEE) platform. Furthermore, the K-means algorithm is employed to cluster ground-based precipitation stations based on the climate zones category across Iran. Additionally, the Genetic Optimization Algorithm (GOA) is applied to specify optimal values of weights in weighting-based models. The performance of single and combination datasets are evaluated using statistical error metrics, including Pearson correlation coefficient (PCC), root mean square error (RMSE), Kling Gupta efficiency (KGE), and bias. The Thiessen polygon method has been applied to calculate each cluster's mean precipitation to obtain the optimal weights of stations. Finally, as an efficient uncertainty approach, Cross Wavelet Transform (XWT) method has been used for uncertainty assessment of monthly and seasonal precipitation series. Results indicated that the OWA family as the best fusion model had almost the lowest uncertainty values and best statistical error indices compared to the EW model and all single products in different clusters of Iran. As a motivation part of this study, unlike the single sources of remotely sensed data, multiple datasets can apply together in different climate conditions with various features using combination models, which mentioned uncertainty method has clearly revealed this claim.