Precipitation Datasets Research Articles

Evaluation of ERA5 Reanalysis Precipitation Data in the Yarlung Zangbo River Basin of the Tibetan Plateau

Abstract Atmospheric simulation-based gridded precipitation datasets have been widely used in hydrological and land surface modeling, but may contain larger uncertainties in mountainous regions. This study compared the performance of the fifth European Centre for Medium-Range Weather Forecasts reanalysis (ERA5) precipitation data with two fused precipitation datasets [China Meteorological Administration Land Data Assimilation System version 2.0 (CLDAS2.0) and China Meteorological Forcing Dataset (CMFD)] in the Yarlung Zangbo River basin (YZRB), which has a complex terrain and climate. Compared to in situ observations, ERA5 could capture the spatial–temporal pattern of precipitation but showed high precipitation, especially in the downstream region (lower Nuxia discharge station). In terms of the correlation coefficient, the overall performance of the ERA5 data was slightly worse than that for CMFD data at both the monthly and yearly scales, and was comparable to that of the CLDAS2.0 data. Given that the spatial mismatch between the gridded and in situ data may influence the evaluation, we also employed the water balance method to constrain basinwide precipitation amounts. We found that CLDAS2.0 and CMFD precipitation data tended to cause long-term water imbalance, and ERA5, with a much larger multiyear average annual precipitation, could better close the water budget. Further analysis showed that the differences in multiyear average annual precipitation between ERA5 and in situ observations were closely related to the slope and standard deviation of the subgrid-scale orography, indicating the substantial influence of subgrid topography on precipitation simulation. These findings highlight that ERA5 could be a potential reference dataset for hydrological modeling of the YZRB.

Asymmetric response of gross primary productivity to precipitation changes over China’s karst areas

China’s karst areas play an important role in regulating the global carbon dynamic and mitigating atmospheric CO2 concentration. Yet, extreme drought events have occurred frequently in recent years over China’s karst areas, which have a negative effect on the gross primary productivity (GPP) in the region. It is unclear whether gross primary productivity gains in the wet years can compensate for its losses in the dry years over the karst areas. In this study, we selected the asymmetry indices to assess the potential asymmetric response of gross primary productivity to the change in precipitation over China’s karst areas during the 2003 to 2018 period, using the gross primary productivity and precipitation dataset. Our results show that the gross primary productivity exhibits a positive asymmetry in response to precipitation changes, namely, the gains caused by increased precipitation in the wet years overcompensate the losses caused by decreased precipitation in the dry years. In addition, the gross primary productivity asymmetry shows an increasing trend over the study period despite extreme drought events occurring frequently, which is due to the response of gross primary productivity to drought has significantly decreased over the study period. For each biome, grasslands show the highest positive gross primary productivity asymmetry, indicating that the grassland biomes have a stronger capacity to utilize the increased precipitation during the wet years to increase gross primary productivity compared to other biomes. Furthermore, the gross primary productivity asymmetry over China’s karst areas can be effected by the precipitation asymmetry as well as mean annual precipitation. Our results will contribute to our knowledge of the response of gross primary productivity to precipitation changes in the karst areas of China.

Performance variations of global precipitation products in detecting drought episodes in three wet seasons in Ethiopia: Part II—statistical analysis

AbstractThe need to evaluate global climate data has increased in recent times. In this study, we evaluate the ability of global precipitation products to monitor drought during three wet seasons (Belg/March–May, Kiremt/June–September and Autumn/September–November) and associated rainfall regions in Ethiopia. We employed statistical methods to quantify and evaluate precipitation products based on probability of drought detection (POD), the extent of false alarms (FAR) and the critical success index (CSI) to see the overall performance of the studied precipitation products. The majority of the studied precipitation datasets were relatively better in capturing the Autumn drought in southern Ethiopia, and 18 out of 21 precipitation products captured accurately more than 50% of observed droughts. The CSI scores for this season are also above 0.5 for 14 precipitation products. On the other hand, 15 and 14 precipitation products accurately captured more than 50% of the seasonal drought in Kiremt and Belg rainfall seasons in north‐eastern Ethiopia. In contrast, most precipitation products do not clearly represent the drought conditions of the Kiremt season in north‐western Ethiopia. Only 8 of the 21 precipitation products accurately captured more than 50% of the observed drought in this region, and only 6 precipitation products had a CSI score greater than 0.5. The results can facilitate the selection of precipitation products for drought monitoring purposes, for use in specific wet seasons and regions of Ethiopia.

Evaluation of Seasonal Differences among Three NOAA Climate Data Records of Precipitation

Abstract Three satellite precipitation datasets—CMORPH, PERSIANN-CDR, and GPCP—from the NOAA/Climate Data Record program were evaluated in their ability to capture seasonal differences in precipitation for the period 2007–18 over the conterminous United States. Data from the in situ U.S. Climate Reference Network (USCRN) provided reference precipitation measurements and collocated atmospheric conditions (temperature) at the daily scale. Satellite precipitation products’ (SPP) performance with respect to cold season precipitation was compared to warm season and full-year analysis for benchmarking purposes. Considering an ensemble of typical performance metrics including accuracy, false alarm ratio, probability of detection, probability of false detection, and the Kling–Gupta efficiency (KGE) that combines correlation, bias, and variability, we found that the three SPPs displayed better performances during the warm season than during the cold season. Among the three datasets, CMORPH displayed better performance—smaller bias, higher correlation, and a better KGE score—than the two other SPPs on an annual basis and during the warm season. During the cold season, CMORPH showed the worst performance at higher latitudes over areas experiencing recurring snow or frozen and mixed precipitation. CMORPH’s performances were further degraded compared to PERSIANN-CDR and GPCP when considering freezing temperatures (T < 0°C) due to the inability to microwave sensors to retrieve precipitation over snow-covered surface. However, for cold rainfall events detected simultaneously by the satellite and the USCRN stations (i.e., conditional case), CMORPH performance noticeably improved but remained inferior to the two other datasets. The quantification of seasonal precipitation errors and biases for three satellite precipitation datasets presented in this work provides an objective basis for the improvement of rainfall retrieval algorithms of the next generation of satellite precipitation products.

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.

Quantifying the snowfall variations in the Third Pole region from 1980 to 2020

Snowfall is critical in the mountainous cryosphere cycle and is vulnerable to climate change. Due to data-scarcity and strong surface heterogeneity, its information is not sufficiently revealed across the Third Pole (TP) under warming. Accordingly, this study investigated the spatiotemporal variability of snowfall and its entropy information across the TP from 1980 to 2020 based on a new high-accuracy snowfall dataset (TPHiS), which was estimated from a long-term high-resolution precipitation dataset for the TP (TPHiPr) and High Asia Refined analysis version 2 reanalysis data (HARv2) using a threshold temperature method. The results indicated that the correlation coefficient, mean bias, and root-mean-square error of the daily TPHiS were 0.77, 0.03 mm/daily, and 0.18 mm/daily, respectively, compared against in-situ observation, illustrating that this dataset performs better than other snowfall datasets. The annual mean snowfall measured 259.38 mm across the TP and experienced a slight negative trend (−0.77 mm/decade, p > 0.05). A significantly positive trend in annual snowfall was observed in the intersection of the West Kunlun, Karakoram, and Western Inner Tibet, with elevations over 5500 m. In contrast, a significantly negative trend prevailed in the Pamir, Hengduan Shan, and West and Central Himalaya intersections, especially at elevations varied from 3000 to 4500 m. Compared with other seasons, a significantly negative trend in snowfall was observed in summer (−1.94 mm/decade, p < 0.05). Change in snowfall frequency entropy was more significant than that in snowfall amount entropy. A significantly positive trend in light snowfall frequency dominated the augmentation in snowfall frequency, causing a negative trend in snowfall frequency entropy, while a substantial reduction in snowfall frequency associated with the decline in light snowfall frequency induced a positive trend in snowfall frequency entropy. The outcomes provide a finer snowfall dataset for regional cryospheric research and might inform sustainable water resource management in alpines.

Evaluation of the RF-MEP Method for Merging Multiple Gridded Precipitation Products in the Chongqing City, China

Precipitation is a major component of the water cycle. Accurate and reliable estimation of precipitation is essential for various applications. Generally, there are three main types of precipitation products: satellite based, reanalysis, and ground measurements from rain gauge stations. Each type has its advantages and disadvantages. Recent efforts have been made to develop various merging methods to improve precipitation estimates by combining multiple precipitation products. This study evaluated for the first time the performance of the random forest-based merging procedure (RF-MEP) method in enhancing the accuracy of daily precipitation estimates in Chongqing city, China with a complex terrain and sparse observational data. The RF-MEP method was used to merge three widely used gridded precipitation products (CHIRPS, ERA5-Land, and GPM IMERG) with ground measurements from a limited number of rain gauge stations to produce the merged precipitation dataset. Eight stations (approximately 70% of the available stations) were used to train the RF-MEP approach, while four stations (30%) were used for independent testing. Various statistical metrics were employed to assess the performance of the merged precipitation dataset and the three existing precipitation products against the ground measurements. Our results demonstrated that the RF-MEP approach significantly enhances the accuracy of daily precipitation estimates, surpassing the performance of the individual precipitation products and two other merging methods (the simple linear regression model and the simple averaging). Among the three existing products, ERA5-Land exhibited the best performance in capturing daily precipitation, followed by GPM IMERG, while CHIRPS performed the worst. Regarding precipitation intensity, all three existing products and the RF-MEP merged dataset performed well in capturing light precipitation events with an intensity of less than 1 mm/day, which accounts for the majority (more than 70%) of occurrences. However, all datasets showed rather poor capability in capturing precipitation events beyond 1 mm/day, with the worst performance observed for extreme heavy precipitation events exceeding 50 mm/day. The RF-MEP approach significantly improves the detection ability for all precipitation intensities, except for the most extreme intensity (&gt;50 mm/day), where only marginal improvement is observed. Analysis of the spatial pattern of precipitation estimates and the temporal bias of daily precipitation estimates further confirms the superior performance of the RF-MEP merged precipitation dataset over the three existing products.

Review on spatial downscaling of satellite derived precipitation estimates

The present work aims at reviewing and identifying gaps in knowledge and future perspectives of satellite-derived precipitation downscaling algorithms. Here, various aspects related to statistical and dynamical downscaling approaches of the precipitation data sets from the Tropical Rainfall Measuring Mission (TRMM) and its successor Intergraded Multi-Satellite Retrievals for Global Precipitation Measurement (IMERG–GPM) mission are reviewed and the existing downscaling methods are categorized and analysed, to highlight the usefulness and applicability of the produced downscaled precipitation data sets. In addition, a critical comparison of the various statistical and dynamical methods for spatial or spatiotemporal downscaling of GPM and TRMM precipitation estimates was conducted, in terms of their advantages and disadvantages, simplicity of application and their suitability at different regional and temporal scales and hydroclimatic conditions. Finally, the adequacy of downscaling remotely sensed precipitation estimates as an effective way to obtain precipitation with sufficient spatial and temporal resolution is discussed and future challenges are highlighted.

Assessing mountain block water storage changes in river basins using water balance and GRACE: A case study on Lake Urmia Basin of Iran

Study regionLake Urmia Basin (LUB), northwestern Iran, encloses the largest salt lake in Iran, where severe drawdown of the lake water level is causing salt storms that threaten the health of the basin's inhabitants. Study focusDifferent in-situ and RS-based datasets are used to estimate the water storage change (WSC) in LUB and partition it into the basin floor WSC and mountain block WSC components. Water balance components for LUB were estimated for the 2001–2011 period, using RS-based datasets for precipitation, actual evapotranspiration, soil moisture and snow storage. Ground-based data were used for validation of RS data as well as water balance closure purpose. GRACE terrestrial water storage data were also used to validate the water balance method results. New hydrological insights for the regionGRACE results confirmed an average WSC of − 921 million cubic meters per year (MCM/yr) (−17.6 mm over the basin area). WSC time series in the basin floor (average: −913 MCM/yr, or −53 mm over the basin floor area) was estimated by summing the related water storage components, and then subtracted from the basin WSC to estimate the mountain block WSC time series (average: −8 MCM/yr, or −0.2 mm over the mountain block area). To evaluate its authenticity, the estimated mountain block WSC time series was then compared to mountain block rivers’ base flow and springs' discharge, showing a considerable correlation between the time series. Considering the faster water resources depletion in the basin floor (where anthropological factors are significant) than in the mountain block (where anthropological factors are rare), one can conclude that anthropological factors may have the primary role in deteriorating water resources conditions in the basin.

Assessing applicability of two gridded precipitation datasets for hydrological simulation in a poorly gauged river basin towards supporting sustainable water resources management

Reliable and accurate precipitation estimates are important for hydrological studies and sustainable water resource management. However, networks of rain gauges are often sparsely and unevenly distributed in many large river basins in the world including the Red River basin (RRB). Thus this study aimed to comprehensively evaluate the applicability of two widely used gridded precipitation products, gauge-based APHRODITE and gauge satellite-based GSMaP-Gauge, over the RRB using both statistical and hydrological assessment approaches. The accuracy assessment of the gridded precipitation datasets was performed by comparing with the reference precipitation dataset derived from the local weather stations. The hydrological performance of both gridded products was evaluated through the Variable Infiltration Capacity (VIC) hydrological modelling scheme for simulation of daily streamflow at the hydrological stations in the RRB. The results demonstrated that both gridded products could generally capture the spatiotemporal variation of the reference precipitation over the RRB during the period of 2005–2014, although both underestimated the reference precipitation. Results of statistical analysis showed that the APHRODITE data outperformed the GSMaP-Gauge data in precipitation estimation. The performance of the VIC model driven by the gridded precipitation products in streamflow simulation was satisfactory, although simulations forced with APHRODITE data displayed the better performance. Generally, the APHRODITE product showed its encouraging potential for hydrological studies over the RRB.

Lightning‐Ignited Wildfires in the Western United States: Ignition Precipitation and Associated Environmental Conditions

AbstractCloud‐to‐ground lightning with minimal rainfall (“dry” lightning) is a major wildfire ignition source in the western United States (WUS). Although dry lightning is commonly defined as occurring with &lt;2.5 mm of daily‐accumulated precipitation, a rigorous quantification of precipitation amounts concurrent with lightning‐ignited wildfires (LIWs) is lacking. We combine wildfire, lightning and precipitation data sets to quantify these ignition precipitation amounts across ecoprovinces of the WUS. The median precipitation for all LIWs is 2.8 mm but varies with vegetation and fire characteristics. “Holdover” fires not detected until 2–5 days following ignition occur with significantly higher precipitation (5.1 mm) compared to fires detected promptly after ignition (2.5 mm), and with cooler and wetter environmental conditions. Further, there is substantial variation in precipitation associated with promptly‐detected (1.7–4.6 mm) and holdover (3.0–7.7 mm) fires across ecoprovinces. Consequently, the widely‐used 2.5 mm threshold does not fully capture lightning ignition risk and incorporating ecoprovince‐specific precipitation amounts would better inform WUS wildfire prediction and management.

Asymmetric Right-Skewed Size-Biased Bilal Distribution with Mathematical Properties, Reliability Analysis, Inference and Applications

Asymmetric distributions, as opposed to symmetric distributions, may be more resilient to extreme values or outliers. Furthermore, when data show substantial skewness, asymmetric distributions can shed light on the underlying processes or phenomena being investigated. In this direction, the size-biased Bilal distribution (SBBD) is suggested in this study as a generalization to the Bilal distribution. The length-biased and area-biased Bilal distributions are discussed in detail as two special cases. The main statistical properties of the distribution including the rth moment, coefficients of variation, skewness, kurtosis, moment generating function, incomplete moments, moments of residual life, harmonic mean, Fisher’s information, and the Rényi entropy as a measure of uncertainty are presented. Graphical representations of the cumulative distribution, probability density, odds, survival, hazard, reversed hazard rate, and cumulative hazard functions are presented for further explanation of the distribution behavior. In addition, the methods of moments and maximum likelihood estimates are taken into account for estimating the model parameters. A simulation study is carried out to see the efficiency of the maximum likelihood in terms of standard errors and bias. Real data sets of precipitation and myeloid leukemia patients are considered to show the practical significance of the suggested distributions as an alternative to some well-known distributions such as the Rama, Rani, Bilal, and exponential distributions. It is found that the size-biased Bilal distribution is right-skewed and has a superior fitting performance compared to the other distributions in this study.

Machine Learning Approaches for Streamflow Modeling in the Godavari Basin with CMIP6 Dataset

Accurate streamflow modeling is crucial for effective water resource management. This study used five machine learning models (support vector regressor (SVR), random forest (RF), M5-pruned model (M5P), multilayer perceptron (MLP), and linear regression (LR)) to simulate one-day-ahead streamflow in the Pranhita subbasin (Godavari basin), India, from 1993 to 2014. Input parameters were selected using correlation and pairwise correlation attribution evaluation methods, incorporating a two-day lag of streamflow, maximum and minimum temperatures, and various precipitation datasets (including Indian Meteorological Department (IMD), EC-Earth3, EC-Earth3-Veg, MIROC6, MRI-ESM2-0, and GFDL-ESM4). Bias-corrected Coupled Model Intercomparison Project Phase 6 (CMIP6) datasets were utilized in the modeling process. Model performance was evaluated using Pearson correlation (R), Nash–Sutcliffe efficiency (NSE), root mean square error (RMSE), and coefficient of determination (R2). IMD outperformed all CMIP6 datasets in streamflow modeling, while RF demonstrated the best performance among the developed models for both CMIP6 and IMD datasets. During the training phase, RF exhibited NSE, R, R2, and RMSE values of 0.95, 0.979, 0.937, and 30.805 m3/s, respectively, using IMD gridded precipitation as input. In the testing phase, the corresponding values were 0.681, 0.91, 0.828, and 41.237 m3/s. The results highlight the significance of advanced machine learning models in streamflow modeling applications, providing valuable insights for water resource management and decision making.

Gridded Precipitation Datasets and Gauge Precipitation Products for Driving Hydrological Models in the Dead Sea Region, Jordan

The consistency of hydrological process modeling depends on reliable parameters and available long-term gauge data, which are frequently restricted within the Dead Sea/Jordan regions. This paper proposes a novel method of utilizing six satellite-based and reanalysis precipitation datasets, which are assessed, evaluated, and corrected, particularly for the cases of ungauged basins and poorly monitored regions, for the first time. Due to natural processes, catchments fluctuate dramatically annually and seasonally, making this a challenge. This variability, which is significantly impacted by topo-geomorphological and climatic variables within the basins themselves, leads to increased uncertainty in models and significant restrictions in terms of runoff forecasting. However, quality evaluations and bias corrections should be conducted before the application of satellite data. Moreover, the hydrological HEC-HMS model was utilized to predict the runoff under different loss methods. Furthermore, this loss method was used with an integrated model that might be efficiently employed when designing hydraulic structures requiring high reliability in predicting peak flows. The models’ performance was evaluated using R-squared (R2), the root mean square error (RMSE), the mean absolute error (MAE), and Nash–Sutcliffe efficiency (NSE). In addition, these statistical metrics were implemented to quantitatively evaluate the data quality based on the observed data collected between 2015 and 2020. The results show that AgERA5 exhibited better agreement with the gauge precipitation data than other reanalysis precipitation and satellite-based datasets. The results demonstrate that the data quality of these products could be affected by observational bias, the spatial scale, and the retrieval method. Moreover, the SC loss method demonstrated satisfactory values for the R2, RMSE, NSE, and bias compared to the IC and GA loss, indicating its effectiveness in predicting peak flows and designing hydraulic structures that require high reliability. Overall, the study suggests that AgERA5 can provide better precipitation estimates for hydrological modeling in the Dead Sea region in Jordan. Moreover, integrating the SC, IC, and GA loss methods in hydraulic structure design can enhance prediction accuracy and reliability.

Occurrence of heavy precipitation influenced by solar wind high-speed streams through vertical atmospheric coupling

A tendency of heavy rainfall-induced floods in Canada to follow arrivals of solar wind high-speed streams (HSSs) from coronal holes is observed. Precipitation events during the winter, including extreme freezing rain events in the province of New Brunswick, also tend to occur following HSSs. More direct evidence is provided using the satellite-based gridded precipitation dataset Integrated Multi-satellitE Retrievals for GPM (IMERG) in the superposed epoch analysis of high-rate precipitation. The results show an increase in the high-rate daily precipitation occurrence over Canada following arrivals of major HSSs. This is consistent with previously published results for other mid-latitude geographic regions. The ERA5 meteorological reanalysis is used to evaluate the slantwise convective available potential energy (SCAPE) that is of importance in the development of storms. The role of the solar wind-magnetosphere-ionosphere-atmosphere coupling, mediated by globally propagating aurorally excited atmospheric gravity waves releasing the conditional symmetric instability in the troposphere leading to convection and precipitation, is proposed.

A new daily gridded precipitation dataset for the Chinese mainland based on gauge observations

Abstract. High-quality, freely accessible, long-term precipitation estimates with fine spatiotemporal resolution play essential roles in hydrologic, climatic, and numerical modeling applications. However, the existing daily gridded precipitation datasets over China are either constructed with insufficient gauge observations or neglect topographic effects and boundary effects on interpolation. Using daily observations from 2839 gauges located across China and nearby regions from 1961 to the present, this study compared eight different interpolation schemes that adjusted the climatology based on a monthly precipitation constraint and topographic characteristic correction, using an algorithm that combined the daily climatology field with a precipitation ratio field. Results from these eight interpolation schemes were validated using 45 992 high-density daily gauge observations from 2015 to 2019 across China. Of these eight schemes, the one with the best performance merges the Parameter-elevation Regression on Independent Slopes Model (PRISM) in the daily climatology field and interpolates station observations into the ratio field using an inverse-distance weighting method. This scheme had median values of 0.78 for the correlation coefficient, 8.8 mm d−1 for the root-mean-square deviation, and 0.69 for the Kling–Gupta efficiency for comparisons between the 45 992 high-density gauge observations and the best interpolation scheme for the 0.1∘ latitude × longitude grid cells from 2015 to 2019. This scheme had the best overall performance, as it fully considers topographic effects in the daily climatology field and it balances local data fidelity and global fitting smoothness in the interpolation of the precipitation ratio field. Therefore, this scheme was used to construct a new long-term, gauge-based gridded precipitation dataset for the Chinese mainland (called CHM_PRE, as a member of the China Hydro-Meteorology dataset) with spatial resolutions of 0.5, 0.25, and 0.1∘ from 1961 to the present. This precipitation dataset is expected to facilitate the advancement of drought monitoring, flood forecasting, and hydrological modeling. Free access to the dataset can be found at https://doi.org/10.6084/m9.figshare.21432123.v4 (Han and Miao, 2022).

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.

Statistical learning of water budget outcomes accounting for target and feature uncertainty

Statistical learning seeks to learn statistics-based rules for data analysis tasks from known examples of inputs, or features, and corresponding outcomes and includes machine learning (ML) and deep learning (DL) algorithms. Data sets that are noisy, include significant uncertainty, and have extreme values hinder the learning process. In this study, aquifer recharge predictors are developed using four, random forest or gradient boosting ML methods and Long Short-Term Memory (LSTM) networks, a DL method to: (1) examine predictive skill when trained using noisy and uncertain data and (2) identify advantages of statistical learning implementations for prediction of water budget outcomes relative to process-based water budget calculations. Recharge was selected as the learning outcome because it is not observed and inherently uncertain. Precipitation, potential evapotranspiration (PET), and river discharge are the features, or inputs, and are calculated, or modelled, values and are not directly observed; consequently, they are expected to be noisy and uncertain because of contamination with measurement and model error. A common-sense baseline is developed and implemented to account for uncertainty and noise in outcomes for training and validation; the baseline provides delineation of a lower goodness-of-fit threshold that identifies when trained ML and DL models generate prediction skill and an upper goodness-of-fit threshold above which the models are learning to reproduce noise and bias. For statistical learning regression implementations, features and outcomes need to be transformed to be Gaussian-like. Inherent variability and extreme events in precipitation, discharge, and recharge data sets require power transformation, or at least scaling of logarithms, to enhance predictive skill. Identified advantages to statistical learning of water budget outcomes are the ability to use dimensionless trends for features and to represent a complex study site with the same level of effort as a simple site.

Performance evaluation of multiple satellite rainfall data sets in central highlands of Abbay Basin, Ethiopia

ABSTRACT This study evaluates four satellite-based precipitation datasets with gauged rainfall observations at a daily and wet season time scales. Satellite Precipitation Estimators from Climate Hazards Group Infrared Precipitation Stations (CHIRPSv2), the Climate prediction center (CPC) morphing technique (CMORPH), the Integrated Multi-satellite Retrieval for GPM (IMERG-06) and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks Cloud Classification System (PERSSIANN-CCS) were used. Categorical and continuous statistical techniques were applied to evaluate the detection and estimation ability of SPEs. Among the SPEs, CMORPH was superior, 0.90 and followed by IMERG-06, 0.82 in the Probability of detection (POD), Critical success index (CSI), 0.74 and 0.70 and Accuracy, 0.75 and 0.72 at wet season, respectively. The result of bias ratio showed underestimation for all SPEs at wet season and daily time scale. The comparison of SPEs in terms of RMSE indicated that IMERG-06 was better, 75.9 followed by CMORPH, 86.0. In terms of Relative bias (RBias), IMERG-06 was the second next to CHIRPSv2. On the other hand, IMERG-06 had better RMSE at JJAS and daily time scale. Overall, the findings of this evaluation study showed the capacity of IMERG-06 product could be reliable rainfall data sources for climate and hydrological analysis of central highlands of Abbay Basin.

Validation of precipitation reanalysis products for rainfall-runoff modelling in Slovenia

Abstract. Observational data scarcity often limits the potential of rainfall-runoff modelling around the globe. In ungauged catchments, earth observations or reanalysis products could be used to replace missing ground-based station data. However, performance of different datasets needs to be thoroughly tested, especially at finer temporal resolutions such as hourly time steps. This study evaluates the performance of ERA5-Land and COSMO-REA6 precipitation reanalysis products (PRPs) using 16 meso-scale catchments (41–460 km2) located in Slovenia, Europe. These two PRPs are firstly compared with a gridded precipitation dataset that was constructed based on ground observational data. Secondly, a comparison of the temperature data of these reanalysis products with station-based air temperature data is conducted. Thirdly, several data combinations are defined and used as input for the rainfall-runoff modelling using the GR4H model. A special focus is on the application of an additional snow module. Both tested PRPs underestimate, for at least 20 %, extreme rainfall events that are the driving force of natural hazards such as floods. In terms of air temperature, both tested reanalysis products show similar deviations from the observational dataset. Additionally, air temperature deviations are smaller in winter compared to summer. In terms of rainfall-runoff modelling, the ERA5-Land yields slightly better performance than COSMO-REA6. If a recalibration with PRP has been carried out, the performance is similar compared to the simulations where station-based data were used as input. Model recalibration proves to be essential in providing relatively sufficient rainfall-runoff modelling results. Hence, tested PRPs could be used as an alternative to the station-based data in case precipitation or air temperature data are lacking, but model calibration using discharge data would be needed to improve the performance.