Multi-Source Precipitation Research Articles

Multi-source precipitation products assessment on drought monitoring across global major river basins

Increased availability of precipitation observational datasets allows for accurate drought detection, monitoring and forecast. Most studies have shown accuracy of various precipitation datasets on regional drought detection and/or monitoring, how they perform across different global river basins, however, is less understood. Here, we evaluate five popular precipitation products including GPM-IMERG, ERA5, GLADS, CPC and PREC/L for meteorological drought detection in 46 large river basins globally, by employing the Standardized Precipitation Index (SPI) at 3-month time scale (SPI-3). The results show that there is a large performance difference among these products. On the whole, GPM-IMERG and PREC/L are ideal precipitation datasets for drought monitoring in North America, South America and Asia, while GLDAS is better in Europe. In terms of specific mega drought events over major basins, all precipitation products perform well in capturing mega drought events overall, especially for the Yangtze, Danube, and Colorado river basins, with PREC/L having the highest accuracy. Specifically, PREC/L well captures the centroid and extent of the typical mega drought events over the Orange and Murray basins. The outcomes of this study can provide a deeper insight into applicability of various precipitation datasets in drought monitoring for different global river basins, which guides scientists and engineers to select suitable precipitation products and further better manage drought disasters.

A Method for Spatiotemporally Merging Multi-Source Precipitation Based on Deep Learning

Reliable precipitation data are essential for studying water cycle patterns and climate change. However, there are always temporal or spatial errors in precipitation data from various sources. Most precipitation fusion methods are influenced by high-dimensional input features and do not make good use of the spatial correlation between precipitation and environmental variables. Thus, this study proposed a novel multi-source precipitation spatiotemporal fusion method for improving the spatiotemporal accuracy of precipitation. Specifically, the attention mechanism was used to first select critical input information to dimensionalize the inputs, and the Convolutional long-short-term memory network (ConvLSTM) was used to merge precipitation products and environmental variables spatiotemporally. The Yalong River in the southeastern part of the Tibetan Plateau was used as the case study area. The results show that: (1) Compared with the original precipitation products (IMERG, ERA5 and CHIRPS), the proposed method has optimal accuracy and good robustness, and its correlation coefficient (CC) reaches 0.853, its root mean square coefficient (RMSE) decreases to 3.53 mm/d and its mean absolute error (MAE) decreases to 1.33 mm/d. (2) The proposed method can reduce errors under different precipitation intensities and greatly improve the detection capability for strong precipitation. (3) The merged precipitation generated by the proposed method can be used to describe the rainfall–runoff relationship and has good applicability. The proposed method may greatly improve the spatiotemporal accuracy of precipitation in complex terrain areas, which is important for scientific management and the allocation of water resources.

A deep learning-based framework for multi-source precipitation fusion

Accurate quantitative precipitation estimation (QPE) is essential for various applications, including land surface modeling, flood forecasting, drought monitoring and prediction. In situ precipitation datasets, remote sensing-based estimations, and reanalysis products have heterogeneous uncertainty. Numerous models have been developed to merge precipitation estimations from different sources to improve the accuracy of QPE. However, many of these attempts are mainly focused on spatial or temporal correlations between various remote sensing sources and/or gauge data separately, and thus, the developed model cannot fully capture the inherent spatiotemporal dependencies that could potentially improve the precipitation estimations. In this study, we developed a general framework that can simultaneously merge and downscale multiple user-defined precipitation products by using rain gauge observations as target values. A novel deep learning-based convolutional neural network architecture, namely, the precipitation data fusion network (PDFN), that combines multiple layers of 3D-CNN and ConvLSTM was developed to fully exploit the spatial and temporal patterns of precipitation. This architecture benefits from techniques such as batch normalization, data augmentation schemes, and dropout layers to avoid overfitting and address skewed class proportions due to the highly imbalanced nature of the precipitation datasets. The results showed that the fused daily product remarkably improved the mean square error (MSE) and Pearson correlation coefficient (PCC) by 35% and 16%, respectively, compared to the best-performing product.

Machine learning approaches for reconstructing gridded precipitation based on multiple source products

Study regionSouth Korea is situated in the northeastern region of Asia Study focus:Recent technological developments have enabled multi-source precipitation products (MSPs), including satellite-based and model-based, to be useful data sources for quantifying the spatiotemporal variations in precipitation. Unfortunately, the main limitation of MSPs in potential applications is inheritance errors with high uncertainty. To tackle this problem, the capabilities of six machine learning algorithms (Ridge Linear Regression, k-Nearest Neighbors, Support Vector Regression, Gradient Boosting Decision Tree, Light Gradient Boosting Machine, and Random Forest) to produce new precipitation product by merging MSPs with ground-based data have investigated. Ground-based data from 2003 to 2017 were utilized for train and valid process. The robustness of the ML algorithms was highlighted using several evaluation metrics such as continuous indices (modified Kling-Gupta Efficiency and root mean square error) and categorical indices (probability of detection, false alarm rate, and critical success index). New hydrological insights for the regionThe results indicate that (1) the ML approaches can merge MSPs with observed data for accurately estimate rainfall, particularly in basins with sparsely distributed rain gauge stations. (2) The merged precipitation products generated from the six ML approaches showed significant agreement and high accuracy with observation data considering rainfall intensity estimation and improved the capability of detecting rain and non-rain events over South Korea.

A framework for merging precipitation retrievals and gauge-based observations based on a novel concept namely virtual gauges

The merging of multi-source precipitation retrievals (PRs) and gauge-based observations (GO) provides a new opportunity for precipitation field estimation. However, the uncertainties associated with PRs may become relatively more evident when the gauge network better captures the spatial distribution of the rainfall fields. To dynamically balance the utilisation of PRs, we propose a merging framework based on a novel concept; namely, the virtual gauge, where the grid cells utilising PRs are regarded as virtual gauges, and the framework is henceforth referred to as VG. The main steps were as follows: i) determine the locations of virtual gauges from multi-source PRs, ii) estimate rainfall at virtual gauges using a basic merging method, and iii) spatially interpolate using both actual and virtual gauges. Accordingly, the case study employed random forest and inverse distance weight as the basic merging and interpolation methods of VG. Evaluation using real world data over a region where nearly each 0.1° grid cell contains a ground gauge indicates that VG improves around 7–11% over its basic methods and improves around 32–240% over its inputting PRs. VG performed better and was more stable than the basic methods under various gauge densities, rainfall intensities, and rainfall distributions. The results showed that VG could supplement the spatial information of rainfall fields missed by the gauge network and reduce interference caused by the uncertainties of PRs. Overall, the framework integrates the advantages of existing merging and spatial interpolation methods by adjusting the number, locations, and rainfall estimations of virtual gauges.

Quantitative Evaluation of Runoff Simulation and Its Driving Forces Based on Hydrological Model and Multisource Precipitation Fusion

The hydrological cycle across the source regions of the Yellow River (SRYR) affects water supply for 324 million people across the Yellow River basin (YRB), and the scarcity of meteorological stations leads to great challenges for the estimation of hydrologic and energy balance. Therefore, our work employs multisource precipitation products across the YRB to develop a new integrated precipitation product with the optimized Bayesian mean algorithm (OBMA). It investigates the performance and hydrological utility of the optimal Bayesian integrated precipitation product (OBIPP). This study found that the OBIPP improved by 14.08% in overall performance relative to the optimal precipitation product across the SRYR, respectively. Meanwhile, the variable infiltration capacity (VIC) model, driven by daily OBIPP, can drastically improve the accuracy of runoff simulation compared with other precipitation products across the SRYR. According to the VIC model driven by daily OBIPP, the average precipitation and runoff depth across the SRYR were approximately 621 mm and 64 mm from 2001 to 2019, respectively, showing a spatial trend increasing from northwest to southeast. Overall, OBIPP is characterized by smaller uncertainty of simulation and higher simulation performance across the SRYR, which should provide a scientific basis for accurate prediction and assessment of water resources in areas where meteorological data are scarce.

Deep learning-based multi-source precipitation merging for the Tibetan Plateau

Deep learning-based multi-source precipitation merging for the Tibetan Plateau

Real-Time Flood Forecasting via Parameter Regionalization and Blending Nowcasts with NWP Forecasts over the Jiao River, China

Abstract Supertyphoon rainstorms are apposite examples to evaluate the utility of multisource precipitation products in monitoring and forecasting short-duration heavy rainfall and the resulting intense floods. In this study, the record-breaking floods induced by Typhoon Lekima in Jiao River, China, were retrospectively forecasted. The Xinanjiang (XAJ) model was calibrated based on parameter regionalization derived from SOM+k-means clustering. Via XAJ, the performance of the currently prevailing atmosphere reanalysis (CLDASv2 and CMA-CMORP), quantitative precipitation estimation (QPE) (IMERG-ER and PERSIANN-CCS), and quantitative precipitation forecasts (QPFs) (GRAPES_MESO, ECMWF, and GFS) in monitoring and forecasting Lekima rainfall and flood was comprehensively evaluated. A three-component blended ensemble was proposed, by blending QPE nowcasts with the weighted ensemble of QPFs through a transition of the regional GRAPES_MESO, and compared with two conventional two-component blending methods. The results indicated that the parameter regionalization enabled an explicit consideration of the spatial heterogeneity of basin attributes as well as meteorology, resulting in a minimum NSE of 0.81. CLDASv2 and CMA-CMORPH provided superior spatiotemporal accuracy with a structural similarity index up to 0.75 and NSE > 0.9 for the flood simulation. PERSIANN-CCS rainfall and the driven flood were seriously underestimated by 70% and 80%, respectively. The real-time application of QPFs during the Lekima flood provided encouraging results with a lead time of 40 h. The three-component blended ensemble method resulted in more stable and accurate flood forecasts, especially for the flood peak on 9 August, which was improved by 80%. Our results are expected to present support for real-time flood preparation and mitigation with practical significance.

Merging multisatellite precipitation products using stacking method and the censored-shifted gamma ensemble model output statistics in china's Beimiaoji basin

High-quality precipitation data are crucial for hydrological analyses, water resource management, and drought monitoring. Merging precipitation products from different sensors and algorithms provides more reliable spatial information and can obtain high-quality precipitation data. To obtain high-spatiotemporal-resolution precipitation, in this study, we proposed a method to merge multisource precipitation products by downscaling coarse resolution products(geographically weighted regression (GWR) algorithm) and then eliminating biases in individual datasets(stacking algorithm), and finally merging the bias-adjusted precipitation by weighted average(the ensemble model output statistics-censored, shifted gamma (EMOS-CSG) algorithm). The multisource precipitation products included the Tropical Rainfall Measuring Mission (TRMM) Multisatellite Precipitation 3B42 in Real-Time (TMPA-3B42RT), Climate Precipitation Center Morphing Technique (CMORPH), Global Satellite Mapping of Precipitation Near-Real-Time (GSMaP_NRT), and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) products). The merging method was applied to the Beimiaoji basin from April to October in the 2016–2019 period. The precipitation observations at 38 gauges were spatially and randomly divided into two parts, 28 gauges were used for the training, while the other for the performance evaluations. The results showed that (1) the daily merged precipitation product had a better performance than the original satellite products in terms of the six considered statistical indexes, with the lowest root mean square error (RMSE) at 4.33 mm and the highest correlation coefficient (CC) at 0.64; (2) the utilized merging method not only increased the spatial resolution to 1 km but also captured more detailed precipitation distribution information; and (3) considering the influence of the gauge density, the performance of the merged product was still improving after the number of gauge stations exceeds 16, but the improvement was not significant.

Fusion of gauge-based, reanalysis, and satellite precipitation products using Bayesian model averaging approach: Determination of the influence of different input sources

Selection of the number and which of multisource precipitation datasets is crucially important for precipitation fusion. Considering the effects of different inputs, this study proposes a new framework based on the Bayesian model averaging (BMA) algorithm to integrate precipitation information from gauge-based analysis CPC, reanalysis-derived dataset ERA5, and satellite-retrieval products IMERG-E and GSMaP-RT. The BMA weights were optimized for the period 2001–2010 using daily measurements and then applied to the period 2011–2015 for model validation. Seven BMA-merged precipitation products (i.e., MCE, MCI, MCG, MCEI, MCEG, MCIG, and MCEIG) were thoroughly evaluated across mainland China and then compared against the state-of-the-art ensemble-based product, MSWEP. The results indicate that the BMA predictions performed substantially better than the reanalysis and satellite precipitation datasets in both daily statistics and seasonal analyses. MCE, MCI, and MCEG demonstrated better performances relative to CPC in terms of individual metrics. Moreover, MCI, MCG, and MCEI generally outperformed MSWEP over the entire study area, particularly in local regions, such as southwestern China and the eastern Tibetan Plateau. During Typhoon Rammasun in 2014, MCG and MCEG provided greater detail for heavy rainfall events than the four ensemble members and the MSWEP product. Thus, the performance of the BMA predictions exhibited evident differences because of various input sources. CPC was the major internal influencing factor with the highest weight score. Meanwhile, the increased-input CPC dataset into the BMA-based schemes exerted a significant influence on the precipitation estimates, which markedly facilitated the performance improvement of the fusion model, and its improved degree (greater than 14 %) was obtained using a ‘changed-initial’ comparison method. Our results demonstrate that the developed modifiable BMA framework is useful for analyzing the impacts of ensemble members on BMA predictions and suggests that it is considerate in the use of different input sources for generating ensemble-based precipitation products.

Quantifying the Reliability and Uncertainty of Satellite, Reanalysis, and Merged Precipitation Products in Hydrological Simulations over the Topographically Diverse Basin in Southwest China

With the continuous emergence of remote sensing technologies and atmospheric models, multi-source precipitation products (MSPs) are increasingly applied in hydrometeorological research, especially in ungauged or data-scarce regions. This study comprehensively evaluates the reliability of MSPs and quantifies the uncertainty of sources in streamflow simulation. Firstly, the performance of seven state-of-the-art MSPs is assessed using rain gauges and the Block-wise use of the TOPMODEL (BTOP) hydrological model under two calibration schemes over Jialing River Basin, China. Then, a variance decomposition approach (Analysis of variance, ANOVA) is employed to quantify the uncertainty contribution of precipitation products, model parameters, and their interaction in streamflow simulation. The MSPs include five satellite-based (GSMaP, IMERG, PERCDR, CHIRPS, CMORPH), one reanalysis (ERA5L), and one ensembled product (PXGB2). The results of precipitation evaluation show that the MSPs have temporal and spatial variability and PXGB2 has the best performance. The hydrologic utility of MSPs is different under different calibration methods. When using gauge-based calibration parameters, the PXGB2-based simulation performs best, whereas CHIRPS, PERCDR, and ERA5L show relatively poor performance. In comparison, the model recalibrated by individual MSPs significantly improves the simulation accuracy of most MSPs, with GSMaP having the best performance. The ANOVA results reveal that the contribution of precipitation products to the streamflow uncertainty is larger than model parameters and their interaction. The impact of interaction suggests that a better simulation attributes to an optimal combination of precipitation products and model parameters rather than solely relying on the best MSPs. These new findings are valuable for improving the suitability of MSPs in hydrologic applications.

SPEI and multi-threshold run theory based drought analysis using multi-source products in China

With the accelerated climate change, the selection of globe-warming related drought monitoring tools is vitally significant. To select the most suitable multi-source precipitation product for drought monitoring, this study comprehensively evaluated the performance of 7 latest version of multi-source precipitation estimate products (MPEs, including PERSIANN-CDR, CHIRPS, ERA5, MERRA2, APHRODITE, GPCC and MSWEP) and compared their drought monitoring applicability combined with GLEAM potential evapotranspiration (ETP) product based on the standardized precipitation evapotranspiration index (SPEI) during 1983–2015 over China. The SPEI estimated by the China monthly Precipitation Analysis Product (CPAP) was used as the reference to assess their performance. In addition, the drought area ratio, frequency, duration and intensity were analyzed based on the multi-threshold run theory. The results showed that: (1) The precision of MPEs was higher in eastern China but lower in western China, and the overall accuracy of MPEs summarized as 4 levels from top to end respectively was GPCC and APHRODITE, MERRA2 and MSWEP, CHIRPS and PERSIANN-CDR as well as ERA5. GLEAM ETP performed well in eastern China but poor in western China with large underestimations. (2) The applicability of MPEs in drought monitoring was regional-specific (low in parts of Northwest Desert and Qinghai-Tibet Plateau, but high in other areas). The accuracy of estimated SPEI for most MPEs dropped with expanding timescale in most sub-regions. Among of MPEs, GPCC product comprehensively performed best in drought monitoring. (3) Drought events occurring in northern China were usually with larger area ratio, lower frequency and longer duration than other regions. Drought events had an exacerbating trend in southeastern China. Overall, this study provides significant information to choose suitable datasets for globe-warming related drought studies.

Quantifying and reducing the uncertainty in multi-source precipitation products using Bayesian total error analysis: A case study in the Danjiangkou Reservoir region in China

Quantifying and reducing the errors and uncertainties in high-resolution gridded precipitation products are essential for hydrological modelling to generate realistic parameter estimates and accurate model predictions. Formal Bayesian inference provides a powerful tool to quantify rainfall uncertainty and other uncertainty factors in hydrological modelling. An efficient way to reduce rainfall uncertainty in hydrological modelling is to merge precipitation estimates from multiple sources prior to their real-world applications. However, it remains a challenging task to optimally determine the weightings of different precipitation estimates based on the predictive ability of a hydrological model, especially from a probabilistic perspective. This study addresses this issue by evaluating and merging six gridded precipitation products using formal Bayesian inference schemes, including the Markov Chain Monte Carlo (MCMC) approach and Bayesian total error analysis (BATEA). MCMC describes the total predictive uncertainty using residual error models. In contrast, BATEA utilizes the storm depth multiplier approach to separately account for rainfall uncertainty in probabilistic uncertainty quantification. A novel signature of model predictability that combines the reliability and precision of probabilistic model predictions is developed to quantify the predictive ability of a hydrological model from precipitation estimates. Using the proposed predictability signature, we examined the benefit of optimally merging multi-source precipitation estimates in improving the accuracy of hydrologic simulations by applying the method to the Danjiangkou Reservoir region (DRR) between 1998 and 2007. The analysis demonstrates that (1) BATEA, compared to MCMC, can provide more realistic estimates of model parameters and tighter predictive intervals of daily streamflow when using rain gauge observations as well as gridded precipitation products; and (2) the proposed predictability signature is efficient for optimally merging precipitation measurements as it can produce tighter uncertainty bands and better streamflow simulations than individual precipitation estimates. This study provides insights into improved quantification and reduction of uncertainty in multi-source precipitation products for hydrological modelling.

Improved Daily Spatial Precipitation Estimation by Merging Multi-Source Precipitation Data Based on the Geographically Weighted Regression Method: A Case Study of Taihu Lake Basin, China.

Accurately estimating the spatial and temporal distribution of precipitation is crucial for hydrological modeling. However, precipitation products based on a single source have their advantages and disadvantages. How to effectively combine the advantages of different precipitation datasets has become an important topic in developing high-quality precipitation products internationally in recent years. This paper uses the measured precipitation data of Multi-Source Weighted-Ensemble Precipitation (MSWEP) and in situ rainfall observation in the Taihu Lake Basin, as well as the longitude, latitude, elevation, slope, aspect, surface roughness, distance to the coastline, and land use and land cover data, and adopts a two-step method to achieve precipitation fusion: (1) downscaling the MSWEP source precipitation field using the bilinear interpolation method and (2) using the geographically weighted regression (GWR) method and tri-cube function weighting method to achieve fusion. Considering geographical and human activities factors, the spatial and temporal distribution of precipitation errors in MSWEP is detected. The fusion of MSWEP and gauge observation precipitation is realized. The results show that the method in this paper significantly improves the spatial resolution and accuracy of precipitation data in the Taihu Lake Basin.

Statistical uncertainty analysis-based precipitation merging (SUPER): A new framework for improved global precipitation estimation

Multi-source merging is an established tool for improving large-scale precipitation estimates. Existing merging frameworks typically use gauge-based precipitation error statistics and neglect the inter-dependence of various precipitation products. However, gauge-observation uncertainties at daily and sub-daily time scales can bias merging weights and yield sub-optimal precipitation estimates, particularly over data-sparse regions. Likewise, frameworks ignoring inter-product error cross-correlation will overfit precipitation observation noise. Here, a Statistical Uncertainty analysis-based Precipitation mERging framework (SUPER) is proposed for addressing these challenges. Specifically, a quadruple collocation analysis is employed to estimate precipitation error variances and covariances for commonly used precipitation products. These error estimates are subsequently used for merging all products via a least-squares minimization approach. In addition, false-alarm precipitation events are removed via a reference rain/no-rain time series estimated by a newly developed categorical variable merging method. As such, SUPER does not require any rain gauge observations to reduce daily random and rain/no-rain classification errors. Additionally, by considering precipitation product inter-dependency, SUPER avoids overfitting measurement noise present in multi-source precipitation products. Results show that the overall RMSE of SUPER-based precipitation is 3.35 mm/day and the daily correlation with gauge observations is 0.71 [−] – metrics that are generally superior to recent precipitation reanalyses and remote sensing products. In this way, we seek to propose a new framework for robustly generating global precipitation datasets that can improve land surface and hydrological modeling skill in data-sparse regions.

Evaluating the Influence of Multisource Typhoon Precipitation Data on Multiscale Urban Pluvial Flood Modeling

Based on station precipitation observations, radar quantitative precipitation estimates (QPE), and radar fusion data during Typhoon Fitow (2013), the influence of multisource precipitation data on multiscale urban typhoon pluvial flood modeling is studied. Using Shanghai, China, as the study area, a simplified 2D hydrodynamic model is applied to simulations. Combined with actual flood incidents reported by the public and soil moisture data, we perform multiscale verifications and determine the applicability of three precipitation datasets in the modeling. The results are as follows: (1) At the city scale, although QPE have higher spatial resolution, these estimates are lower than station observations. Radar fusion data have both high accuracy and high spatial resolution. For flood depths above 5 cm, the radar fusion precipitation scenario can improve the matching probability by 6%. (2) At the neighborhood scale, the radar fusion precipitation scenario can effectively mitigate the problems of an uneven spatial distribution of stations and a weak QPE to accurately capture pluvial details. (3) One fixed-point assessment shows that different precipitation data have little influence on the temporal characteristics of the modeling result—all three types of data can accurately reflect flood occurrence times. This work can provide a scientific basis for constructing effective urban pluvial flood monitoring systems.

Application of bias corrected FGOALS-g3 model products for detecting changes in extreme precipitation in the Tienshan Mountains, Central Asia

This paper investigates the performance of bias-corrected Flexible Global Ocean-Atmosphere-Land System-g3 (FGOALS-g3) model products and then detects changes in extreme precipitation (EP) in the Tienshan Mountains, Central Asia (TMCA), as reflected by 25 EP indices. The reliability of the FGOALS-g3 model outputs is assessed systematically based on multiple statistical indicators against multi-source precipitation datasets and the bias-corrected FGOALS-g3 model products are applied to project EP variations under different global warming levels. Using the geographical detector method, a novel statistical method for detecting spatial heterogeneity and elucidating the underlying causes, the explanatory power of 20 atmospheric circulation factors related to EP is examined. The findings indicate that while the FGOALS-g3 products can detect the spatial pattern of multi-year average precipitation in the TMCA, there is an obvious overestimation in magnitude, especially in the West and Middle Tienshan Mountains. These biases are significantly reduced, however, after downscaling and bias correction. Compared with the Asian Precipitation Highly-Resolved Observational Data Integration Towards Evaluation (APHRODITE) data, the grid-by-grid error in bias-corrected FGOALS-g3 simulated mean precipitation is between −7.64% and 10.95%. In terms of EP, the corrected FGOALS-g3 products not only reproduce the spatial distribution, but also reasonably simulate their magnitudes, with some overestimation in light EP and underestimation in heavy EP. Overall, across the historical period, EP has increased. The intensity and frequency of EP are projected to generally increase under different scenarios. At 1.5 °C warming levels, annual total precipitation in wet days (PRCP) increases by 5.74% (7.74%) under the SSP245 (SSP585). Additionally, as an EP becomes rarer, its rate of change rises. The main driving factors in EP are detected to be 30 hPa zonal wind (30ZW), relative number of sunspots (SF), south Asian summer monsoon (SAM), sea surface temperature anomaly in the region of 5°S-5°N, 170°-120°W (NINO 3.4), and mean surface temperature (T).

Spatiotemporal Evaluation and Estimation of Precipitation of Multi-Source Precipitation Products in Arid Areas of Northwest China—A Case Study of Tianshan Mountains

In the arid areas of Northwest China, especially in the Tianshan Mountains, the scarcity of meteorological stations has brought some challenges in collecting accurate information to describe the spatial distribution of precipitation. In this study, the applicability of TRMM3B42, GPM IMERG, and MSWEP V2.2 in different regions of Tianshan Mountain is comprehensively evaluated by using ten statistical indicators, three classification indicators, and variation coefficients at different time–space scales, and the mechanism of accuracy difference of precipitation products is discussed. The results show that: (1) On the annual and monthly scales, the correlation between GPM and measured precipitation is the highest, and the ability of three precipitation products to capture precipitation in the wet season is stronger than that in the dry season; (2) On the daily scale, TRMM has the highest ability to estimate the frequency of light rain events, and MSWEP has the highest ability to monitor extreme precipitation events; (3) On the spatial scale, GPM has the highest fitting degree with the spatial distribution of precipitation in Tianshan Mountains, MSWEP is the closest to the precipitation differentiation pattern in Tianshan Mountains; (4) The three satellite products generally perform best in low and middle longitude regions and middle elevation regions. This study provides a reference for the selection of grid precipitation datasets for hydrometeorological simulation in northwest arid areas and also provides a basis for multi-source data assimilation and fusion.

A two-step merging strategy for incorporating multi-source precipitation products and gauge observations using machine learning classification and regression over China

Abstract. Although many multi-source precipitation products (MSPs) with high spatiotemporal resolution have been extensively used in water cycle research, they are still subject to various biases, including false alarm and missed bias. Precipitation merging technology is an effective means to alleviate this uncertainty. However, how to efficiently improve precipitation detection efficiency and precipitation intensity simultaneously is a problem worth exploring. This study presents a two-step merging strategy based on machine learning (ML) algorithms, including gradient boosting decision tree (GBDT), extreme gradient boosting (XGBoost), and random forest (RF). It incorporates six state-of-the-art MSPs (GSMaP, IMERG, PERSIANN-CDR, CMORPH, CHIRPS, and ERA5-Land) and rain gauges to improve the accuracy of precipitation identification and estimation from 2000 to 2017 over China. Multiple environment variables and spatial autocorrelation are combined in the merging process. The strategy first employs classification models to identify wet and dry days and then combines regression models to predict precipitation amounts based on classified wet days. The merged results are compared with traditional methods, including multiple linear regression (MLR), ML regression models, and gauge-based Kriging interpolation. A total of 1680 (70 %) rain gauges are randomly chosen for model training and 692 (30 %) for performance evaluation. The results show that (1) the multi-source merged precipitation products (MSMPs) outperformed all original MSPs in terms of statistical and categorical metrics, which substantially alleviates the temporal and spatial biases. The modified Kling–Gupta efficiency (KGE), critical success index (CSI), and Heidke Skill Score (HSS) of original MSPs are improved by 15 %–85 %, 17 %–155 %, and 21 %–166 %, respectively. (2) The spatial autocorrelation plays a significant role in precipitation merging, which considerably improves the model accuracy. (3) The performance of MSMPs obtained by the proposed method is superior to MLR, Kriging interpolation, and ML regression models. The XGBoost algorithm is recommended more for large-scale data merging owing to its high computational efficiency. (4) The two-step merging strategy performs better when higher-density gauges are used to model training. However, it has strong robustness and can also obtain better performance than original MSPs even when the gauge number is reduced to 10 % (237). This study provides an accurate and reliable method to improve precipitation detection accuracy under complex climatic and topographic conditions. It could be applied to other areas well if rain gauges are available.

Evaluation of IMERG, TMPA, ERA5, and CPC precipitation products over mainland China: Spatiotemporal patterns and extremes

A comprehensive assessment of representative satellite-retrieved (Integrated Multi-satellite Retrievals for Global Precipitation Measurement (IMERG) and Tropical Rainfall Measuring Mission Multi-satellite Precipitation Analysis (TMPA)), reanalysis-based (fifth generation of atmospheric reanalysis by the European Centre for Medium Range Weather Forecasts (ERA5)), and gauge-estimated (Climate Prediction Center (CPC)) precipitation products was conducted using the data from 807 meteorological stations across mainland China from 2001 to 2017. Error statistical metrics, precipitation distribution functions, and extreme precipitation indices were used to evaluate the quality of the four precipitation products in terms of multi-timescale accuracy and extreme precipitation estimation. When the timescale increased from daily to seasonal scales, the accuracy of the four precipitation products first increased and then decreased, and all products performed best on the monthly timescale. Their accuracy ranking in descending order was CPC, IMERG, TMPA, and ERA5 on the daily timescale and IMERG, CPC, TMPA, and ERA5 on the monthly and seasonal timescales. IMERG was generally superior to its predecessor TMPA on the three timescales. ERA5 exhibited large statistical errors. CPC provided stable estimated values. For extreme precipitation estimation, the quality of IMERG was relatively consistent with that of TMPA in terms of precipitation distribution and extreme metrics, and IMERG exhibited a significant advantage in estimating moderate and heavy precipitation. In contrast, ERA5 and CPC exhibited poor performance with large systematic underestimation biases. The findings of this study provide insight into the performance of the latest IMERG product compared with the widely used TMPA, ERA5, and CPC datasets, and points to possible directions for improvement of multi-source precipitation data fusion algorithms in order to better serve hydrological applications.