Modified Kling-Gupta Efficiency Research Articles

A combined extended triple collocation and cumulative distribution function merging framework for improved daily precipitation estimates over mainland China

Accurate monitoring of daily precipitation is essential for many applications, such as simulation and prediction of watershed water-related disasters. However, precipitation data from satellite and gauge-based precipitation products face challenges in capturing true observations because of their large uncertainties. To improve daily precipitation estimates, this study proposed a new merging framework, ETCCCDF, that combines an extended triple collocation (ETCC) with maximized correlation and cumulative distribution function (CDF) matching with reduced bias. The ETCCCDF framework was applied to systematically integrate three independent precipitation products from SM2RAIN-ASCAT satellite soil moisture estimator, IMERG Early Run satellite retrieval, and CPC unified gauge-based analysis. A new merged precipitation product, with a fine spatiotemporal resolution of 0.25°/1 d, was developed and then thoroughly evaluated using 2304 stations over mainland China from 2007 to 2014. The results showed that the ETCCCDF framework significantly improved the precipitation representation in terms of the correlation, bias, and absolute difference, and it increased modified Kling-Gupta efficiency value relative to any of the parent precipitation products, demonstrating the effectiveness of the proposed merging framework. In addition, the ETCCCDF framework is more robust than the ETCC, triple collocation (TC) merging, and combined strategy of TC with CDF, in which CDF matching successfully corrects the precipitation biases in collocation fusions. The accuracy of the merged products in heavy rainfall estimation during Typhoon Morakot in 2009 and Typhoon Fitow in 2013 was further evaluated. The results confirmed that the ETCCCDF-merged product performed satisfactorily for heavy rainfall events and presented superior statistical scores. This study proves that the ETCCCDF framework is a promising solution for obtaining high-quality precipitation data, particularly for limited-gauge areas.

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.

Efficient Super‐Resolution of Near‐Surface Climate Modeling Using the Fourier Neural Operator

AbstractDownscaling methods are critical in efficiently generating high‐resolution atmospheric data. However, state‐of‐the‐art statistical or dynamical downscaling techniques either suffer from the high computational cost of running a physical model or require high‐resolution data to develop a downscaling tool. Here, we demonstrate a recently proposed zero‐shot super‐resolution method, the Fourier neural operator (FNO), to efficiently perform downscaling without the need for high‐resolution data. Because the FNO learns dynamics in Fourier space, FNO is a resolution‐invariant emulator; it can be trained at a coarse resolution and produces emulation at any high resolution. We applied FNO to downscale a 4‐km resolution Weather Research and Forecasting (WRF) Model simulation of near‐surface heat‐related variables over the Great Lakes region. The FNO is driven by the atmospheric forcings and topographic features used in the WRF model at the same resolution. We incorporated a physics‐constrained loss in FNO by using the Clausius–Clapeyron relation to better constrain the relations among the emulated states. Trained on merely 600 WRF snapshots at 4‐km resolution, the FNO shows comparable performance with a widely‐used convolutional network, U‐Net, achieving averaged modified Kling–Gupta Efficiency of 0.88 and 0.94 on the test data set for temperature and pressure, respectively. We then employed the FNO to produce 1‐km emulations to reproduce the fine climate features. Further, by taking the WRF simulation as ground truth, we show consistent performances at the two resolutions, suggesting the reliability of FNO in producing high‐resolution dynamics. Our study demonstrates the potential of using FNO for zero‐shot super‐resolution in generating first‐order estimation on atmospheric modeling.

Monthly streamflow forecasting by machine learning methods using dynamic weather prediction model outputs over Iran

Seasonal hydrological forecasts play a critical role in water resources management. The Copernicus Climate Change Service (C3S) data store provides open access to monthly hydrological forecasts for up to six-months. This study aims to evaluate, for the first time, 1- to 3-month runoff forecasts using the European Centre for Medium-Range Weather Forecasts (ECMWF) ensembles of precipitation, runoff, and temperature in 1981–2015 period over a total of 30 s-level basins in Iran. We adopted the 5th, 50th and 95th ECMWF ensemble quantiles for each variable that represent low, medium and high probability of occurrence, respectively. Pearson correlation analysis (Pca), Recursive Feature Elimination (RFE) via random forest (RF) model, and Bayesian Networks (BN) feature selection algorithms were used in order to reduce input variable dimension and select potential predictors to be fed to the machine learning models. Multiple Linear Regression (MLR), Artificial Neural Networks (ANN), Support Vector Regression (SVR), Random Forest (RF), and eXtreme Gradient Boosting (XGBoost) machine learning models were used with Repeated K-Fold cross validation (rK-Fold CV) while model efficiency was evaluated using modified Kling-Gupta efficiency coefficient (KGE’), Nash-Sutcliffe Efficiency coefficient (NSE), and Normalized Root Mean Square Error (NRMSE). Results of this study revealed that C3S runoff ensembles have the highest impact on forecast accuracy of streamflow, followed by precipitation and temperature. Overall, model performance yield a best-to-worst ranking of ANN, XGBoost, RF, MLR, and SVR with KGE’ values of 0.70, 0.68, 0.66, 0.57, and 0.41, respectively. The predictive performance of all models decreased with lead times beyond 1-month, where ANN and XGBoost outperformed other models with KGE’ of 0.65 for 2-month lead time and 0.60 for 3-month lead time. The three superior models of XGBoost, ANN, and RF, were employed with RFE and BN FSAs most frequently across Iran’s 30 s level basins in all lead times. Almost all models in the arid central region of Iran showed the lowest performance while highest skills were achieved in the western regions of Iran. Finally, for all models and over all regions, the model performance reduced by increase in lead-time.

Resolving new and old phosphorus source contributions to subsurface tile drainage with weighted regressions on discharge and season.

Agricultural losses of dissolved reactive phosphorus (DRP) emanate from both historic P applications (i.e., "old P") and recently applied fertilizer (i.e., "new P"). Understanding the relative contributions of these sources is important for mitigating DRP losses from agriculture. This study provides a proof-of-concept for resolving new P vs. old P source contributions to DRP losses in subsurface tile drainage using edge-of-field water quality data and management records from eight fields in Ohio. Weighted regressions on discharge and season (WRDS) were fitted using data from periods without P fertilizer applications and then used to predict DRP losses in tile drainage during new P loss risk periods (default length, 90 d) after fertilizer applications. Differences between observed and predicted DRP concentrations during the new P loss risk period were attributed to the new P source. Remaining losses were attributed to the old soil P source. The WRDS model performance was modest (modified Kling-Gupta efficiency ranged from -0.074 to 0.484). New P sources contributed between 0 and 17% of overall DRP losses (average, 7%), with old soil P contributing 83-100%. Individual P fertilizer applications were associated with new DRP losses up to 192g P ha-1 . Increasing the length of the risk period for new P losses up to 180 d after fertilizer application marginally increased the estimated contribution of the new P source. The WRDS-based analysis provides a novel approach for resolving the contributions of new and old sources to edge-of-field DRP losses.

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.

Comprehensive evaluation of parameter importance and optimization based on the integrated sensitivity analysis system: A case study of the BTOP model in the upper Min River Basin, China

Parameter optimization is mandatory in many numerical modeling studies of hydrology sciences. However, optimizing all parameters (APs) is highly inefficient, including the equifinality phenomenon. This paper proposes an integrated sensitivity analysis system consisting of, multiple Global sensitivity analysis methods, Design of experiment algorithms, Indicators, and Sample size design Combinations (GDISCs), for screening the sensitive parameters (SPs) robustly and reducing parameter optimization burden efficaciously. In this study, four design of experiment (DoE) algorithms are applied to collect samples with varying sizes, and the corresponding three indicator values. We evaluate the effectiveness and efficiency of one qualitative and two quantitative global sensitivity analysis (GSA) methods by identifying the appropriate DoE algorithm and the sufficient sample size. To apply the GDISCs, the Block-wise use of the TOPMODEL (BTOP) model is set up for the hydrological simulation in the upper Min River Basin, China. The results show that the GDISCs system robustly screens out three SPs of the BTOP model, and the DoE algorithms significantly impact the effectiveness and efficiency of the GSA methods. The SA performance with applying the modified Kling–Gupta efficiency (KGE') as the indicator is more robust than the root mean squared error (RMSE) and the Nash-Sutcliffe efficiency (NSE). Compared to the traditional APs optimization, the optimal values are identified quickly through almost 1/2 or even fewer model runs and duration for the SPs optimization, with a very close accuracy (the maximum NSE, KGE' and relative bias (RBias) differences are 0.137, 0.037 and 2.70%, and the minimum are 0.002, 0.001 and 0.20%). Optimizing SPs also has a noticeable improvement for the default parameter values, and the maximum differences of NSE, KGE' and RBias reach 0.764, 0.321, and −20.80%. In summary, the GDISCs system is especially applicable for obtaining the robust SA results to considerably improve model calibration efficiency and reduce model computational burden while ensuring satisfactory model simulation performance.

Revisiting 2013 Uttarakhand flash floods through hydrological evaluation of precipitation data sources and morphometric prioritization

With advancements in computational technology, data assimilation techniques, high-resolution remote sensing, and complex climate models, numerous precipitation products are available with different spatiotemporal resolutions; however, their evaluation, especially in the Himalayan region, is unexplored. Therefore, this study attempts to assess four sources (gridded observation dataset, reanalysis, satellite, and numerical weather prediction models) of precipitation through hydrological modelling for the catastrophic 2013 floods of Uttarakhand, India. The Upper Ganga Basin located in Western Himalayas is selected as the study area consisting of Alaknanda and Bhagirathi streams in the eastern and western parts. The Hydrologic Engineering Center’s Hydrologic Modeling System (HEC-HMS) is employed for rainfall-runoff modelling. The rainfall from IMD, ERA-5, GPM-IMERG-Final, and WRF model outputs are forced into the calibrated HEC-HMS model for assessing their performance in hydrological simulations. The correlation coefficient of IMD, ERA-5, GPM-IMERG-Final, and WRF simulations with respect to the observed flow is 0.89, 0.88, 0.55, and 0.89, respectively, whereas their corresponding Modified Kling-Gupta Efficiency (KGE) is 0.66, 0.72, 0.48, and 0.71. Flash flood prioritization of the sub-watersheds based on morphometric characteristics suggests that the Alaknanda basin is relatively more vulnerable to flash floods due to their elongated nature, highest relative relief, and high mean slope.

Evaluation of Global Forecast System (GFS) Medium-Range Precipitation Forecasts in the Nile River Basin

Abstract Reliable weather forecasts are valuable in a number of applications, such as agriculture, hydropower, and weather-related disease outbreaks. Global weather forecasts are widely used, but detailed evaluation over specific regions is paramount for users and operational centers to enhance the usability of forecasts and improve their accuracy. This study presents evaluation of the Global Forecast System (GFS) medium-range (1–15 day) precipitation forecasts in the nine subbasins of the Nile basin using NASA’s Integrated Multisatellite Retrievals (IMERG) Final Run satellite–gauge merged rainfall observations. The GFS products are available at a temporal resolution of 3–6 h and a spatial resolution of 0.25°, and the version-15 products are available since 12 June 2019. GFS forecasts are evaluated at a temporal scale of 1–15 days, a spatial scale from 0.25° to all the way to the subbasin scale, and for a period of one year (15 June 2019–15 June 2020). The results show that performance of the 1-day lead daily basin-averaged GFS forecast performance, as measured through the modified Kling–Gupta efficiency (KGE), is poor (0 < KGE < 0.5) for most of the subbasins. The factors contributing to the low performance are 1) large overestimation bias in watersheds located in wet climate regimes in the northern hemispheres (Millennium watershed, Upper Atbara and Setit watershed, and Khashm El Gibra watershed), and 2) lower ability in capturing the temporal dynamics of watershed-averaged rainfall that have smaller watershed areas (Roseires at 14 110 km2 and Sennar at 13 895 km2). GFS has better bias for watersheds located in the dry parts of the Northern Hemisphere or wet parts of the Southern Hemisphere, and better ability in capturing the temporal dynamics of watershed-average rainfall for large watershed areas. IMERG Early has better bias than GFS forecast for the Millennium watershed but still comparable and worse bias for the Upper Atbara and Setit and Khashm El Gibra watersheds. The variation in the performance of the IMERG Early could be partly explained by the number of rain gauges used in the reference IMERG Final product, as 16 rain gauges were used for the Millennium watershed but only one rain gauge over each Upper Atbara and Setit and Khashm El Gibra watershed. A simple climatological bias correction of IMERG Early reduces in the bias in IMERG Early over most watersheds, but not all watersheds. We recommend exploring methods to increase the performance of GFS forecasts, including postprocessing techniques through the use of both near-real-time and research-version satellite rainfall products.

A Morphology-Based Adaptively Spatio-Temporal Merging Algorithm for Optimally Combining Multisource Gridded Precipitation Products With Various Resolutions

Gridded precipitation products with fine resolutions and qualities are of great importance for understanding the global water–carbon-energy cycles at various spatiotemporal scales. Though continuous developments in Satellite Remote Sensing fields have been providing great strengths for measuring the precipitation from space, merging precipitation products from different sources, especially the gauge observations, is still the optimal way for obtaining high-quality precipitation data. Currently, the mainstream merging methods mainly focus on merging the rain rates without the considerations of rain events. In this study, we propose a new assumption that both rain events and rain rates should be considered in the merging procedures rather than only the rain rates. To meet our assumption, a morphology-based adaptive spatio-temporal merging algorithm (MASTMA) for combining various precipitation products is proposed, in which the morphology theory is first introduced to comprehensively consider the influences from both rain events and rain rates. The multisource and multiscale precipitation products including the gauge-based data (CPC-U, 0.5°, daily), the satellite-based data [Global Satellite Mapping of Precipitation by Moving Vector with Kalman (GSMaP-MVK), 0.1°, hourly; integrated multisatellite retrievals for global precipitation measurement late run (IMERG-LR), 0.1°, half-hourly], and the reanalysis data (ERA5-land, 0.1°, hourly), have been comprehensively considered in MASTMA for generating the final estimates (MASTMA-F, 0.1°, hourly) over the southeastern regions of the Mainland China in the periods from 2016 to 2019. The main conclusions include but are not limited to: 1) considerations on rain events contribute significantly to the final merged results, especially when eliminating false extreme values over the regions where precipitation is greatly overestimated; 2) the MASTMA could optimally integrate the advantages from multisource precipitation products with different resolutions, particularly from the perspective of the spatial distributions; and 3) the final merged estimates using MASTMA outperform the contemporary state-of-the-art precipitation products especially in terms of modified Kling–Gupta Efficiency (mKGE) and critical success index (CSI). Additionally, the results of this study suggest that MASMTA is a new promising merging approach with great robustness and applicability, and has the foreseeable potentials for the operational run to generate the optimal global merged precipitation products.

Optimized algorithm for evapotranspiration retrieval via remote sensing

Many algorithms for surface energy balance (SEB) based on remote sensing (RS) have been advanced to determine evapotranspiration (ET). These algorithms were developed for specific conditions (e.g., sensors, land use, and crop management) in which functions and empirical parameters within its algorithms concur with those conditions. Therefore, this study aims to develop a SEB-RS algorithm for retrieving ET adjusted to in situ observations. The study was conducted in two experimental fields in Brazil with the crops Jatropha curcas, maize, soybean, and sugarcane. We used multispectral images from the orbital sensors, Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) coupled in Landsat 8 satellite and from the terrestrial sensor, Altum, on board of an unmanned aerial vehicle. The proposed algorithm termed as Ground-truthed Surface Energy Balance (GT-SEB) is based on physical formulation of SEB-RS algorithms, where two extra computational processes using in situ ET observations were proposed for originating the new algorithm. The first additional process for optimizing the automatic “anchor” pixels selection and another for algorithm parameters optimization. Thus, both processes aim to reduce the difference between the observed ET and estimated by GT-SEB. Being assessed for both orbital (OLI/TIRS) and suborbital (Altum) sensors, the GT-SEB yielded excellent results (root-mean-square-error, RMSE, ≤ 0.48 mm and modified Kling-Gupta efficiency, KGE, ≥ 0.92). In addition to GT-SEB being an optimized algorithm, it uses a classic parameterization of SEB-RS algorithms, providing efficiency and scalability for other remote sensors, climates, and surfaces.

Multi-Criteria Decision Framework to Evaluate Bias Corrected Climate Change Projections in the Piracicaba River Basin

Abstract Regional climate models (RCM) are the main tools for climate change impacts assessment in hydrological studies. These models, however, often show biases when compared to historical observations. Bias Correction (BC) are useful techniques to improve climate projection outputs. This study presents a multi-criteria decision analysis (MCDA) framework to compare combinations of RCM with selected BC methods. The comparison was based on the modified Kling-Gupta efficiency (KGE’). The criteria evaluated the general capability of models in reproducing the observed data main statistics. Other criteria evaluated were the relevant aspects for hydrological studies, such as seasonality, dry and wet periods. We applied four BC methods in four RCM monthly rainfall outputs from 1961 to 2005 in the Piracicaba river basin. The Linear Scaling (LS) method showed higher improvements in the general performance of the models. The RCM Eta-HadGEM2-ES, corrected with Standardized Reconstruction (SdRc) method, achieved the best results when compared to the observed precipitation. The bias corrected projected monthly precipitation (2006-2098) preserved the main signal of climate change effects when compared to the original outputs regarding annual rainfall. However, SdRc produced significant decrease in monthly average rainfall, higher than 45% for July, August and September for RCP4.5 and RCP8.5 scenarios.

Can a physically-based land surface model accurately represent evapotranspiration partitioning? A case study in a humid boreal forest

One of the roles of Land Surface Models (LSMs) is to partition evapotranspiration (E) into overstory transpiration (ET), understory evapotranspiration (EG), and wet canopy evaporation (EC). Unfortunately, only a handful of studies have evaluated the performance of LSMs with E partitioning. Unlike dry canopies which are dominated by transpiration, wet canopies lead to the evaporation of intercepted water. In this respect, there is no better testing site for LSMs than the humid boreal forest, which is characterized by frequent precipitation and sustained evapotranspiration. This study assesses the performance of the Canadian Land Surface Scheme (CLASS) in simulating evapotranspiration and its components by applying detailed observations of water pathways and residence times in a forest canopy using a variety of methods (eddy covariance; sap flow, throughfall and stemflow measurements; and the stem compression approach). The study site was located in Montmorency Forest in Québec, Canada, a balsam fir boreal forest with ≈1600 mm of annual precipitation. The field campaign was conducted during the summers of 2017 and 2018. To improve the accuracy of the simulations, we adjusted the definition of maximum canopy water storage, S=c×LAI, by using c=0.49 mm instead of the default value of 0.20 mm. This simple modification improved the performance of CLASS when simulating components of evapotranspiration, indicated by the increase in the modified Kling-Gupta efficiency (KGE′) with rises between 0.01 and 0.23. Overall, CLASS performed well in simulating evapotranspiration and overstory transpiration in dry canopy conditions at half-hourly time steps with KGE′ values between 0.67 and 0.81. The contribution of simulated ET,EG, and EC to total E during the two sampling periods were 46%,24%, and 30%, respectively, which were comparable with the field observations (35%,22%, and 25%, respectively). In conclusion, CLASS was able to simulate E partitioning in dry and wet canopy conditions reasonably well at both seasonal and half-hourly time scales.

Evaluation and Comparison of Satellite-Derived Estimates of Rainfall in the Diverse Climate and Terrain of Central and Northeastern Ethiopia

Understanding rainfall processes as the main driver of the hydrological cycle is important for formulating future water management strategies; however, rainfall data availability is challenging for countries such as Ethiopia. This study aims to evaluate and compare the satellite rainfall estimates (SREs) derived from tropical rainfall measuring mission (TRMM 3B43v7), rainfall estimation from remotely sensed information using artificial neural networks—climate data record (PERSIANN-CDR), merged satellite-gauge rainfall estimate (IMERG), and the Global Satellite Mapping of Precipitation (GSMaP) with ground-observed data over the varied terrain of hydrologically diverse central and northeastern parts of Ethiopia—Awash River Basin (ARB). Areal comparisons were made between SREs and observed rainfall using various categorical indices and statistical evaluation criteria, and a non-parametric Mann–Kendall (MK) trend test was analyzed. The monthly weighted observed rainfall exhibited relatively comparable results with SREs, except for the annual peak rainfall shifts noted in all SREs. The PERSIANN-CDR products showed a decreasing trend in rainfall at elevations greater than 2250 m above sea level in a river basin. This demonstrates that elevation and rainfall regimes may affect satellite rainfall data. On the basis of modified Kling–Gupta Efficiency, the SREs from IMERG v06, TRMM 3B43v7, and PERSIANN-CDR performed well in descending order over the ARB. However, GSMaP showed poor performance except in the upland sub-basin. A high frequency of bias, which led to an overestimation of SREs, was exhibited in TRMM 3B43v7 and PERSIANN-CDR products in the eastern and lower basins. Furthermore, the MK test results of SREs showed that none of the sub-basins exhibited a monotonic trend at 5% significance level except the GSMap rainfall in the upland sub-basin. In ARB, except for the GSMaP, all SREs can be used as alternative options for rainfall frequency-, flood-, and drought-monitoring studies. However, some may require bias corrections to improve the data quality.

Spatial and Temporal Evaluation of the Latest High-Resolution Precipitation Products over the Upper Blue Nile River Basin, Ethiopia

Quality and representative precipitation data play an essential role in hydro-meteorological analyses. However, the required reliability and coverage is often unavailable from conventional gauge observations. As a result, globally available precipitation datasets are being used as an alternative or supplementary to gauge observations. In this study, the accuracy of three recently released, high-resolution precipitation datasets with a spatial resolution of 0.1° and a daily temporal resolution is evaluated over the Upper Blue Nile River Basin (UBNRB) for the period of 2007 to 2016. The datasets are Integrated Multi-satellitE Retrievals for GPM version 6 (IMERG6), Multi-Source Weighted-Ensemble Precipitation version 2.2 (MSWEP2.2) and soil moisture to rain using Advanced SCATterometer version 1.1 (SM2RAIN-ASCAT1.1). The comparison was made between rain gauge observations and two other high-resolution precipitation datasets named Enhancing National Climate Services (ENACTS) and Climate Hazards Group Infrared Precipitation with Stations version 2 (CHIRPS2). The modified Kling-Gupta efficiency (KGE’) and four categorical indices named probability of detection (POD), false alarm ratio (FAR), critical success index (CSI), and frequency bias (fBIAS) was used to measure the skills of each dataset. Results revealed that, except SM2RAIN-ASCAT1.1, all other datasets show a better ability on a monthly time scale for areas with an elevation below 1500 m above sea level (m.a.s.l). The overall performance was better in the wetter months of March to August than the drier months of September to February. Besides, all products including SM2RAIN-ASCAT1.1 could detect no rain events (rain < 1 mm) correctly, but their skill deteriorates on identifying higher intensity events. By comparison, ENACTS (calibrated with most quality gauges of Ethiopia) and CHIRPS2 exhibited the best performance due to their high-resolution nature and inclusion of physiographic information in their data generation procedures. IMERG6 and MSWEP2.2 showed the next best performance according to both the continuous and categorical indices used. SM2RAIN-ASCAT1.1 demonstrates the least skill everywhere due to problems that could be associated with misinterpretations of soil moisture signals by the SM2RAIN algorithm. Considering the scarcity of gauged datasets over UBNRB, IMERG6 and MSWEP2.2 could be regarded as valuable datasets for hydro-climatic analysis, mainly where gauge density is low. SM2RAIN-ASCAT1.1, on the other hand, needs significant bias correction to treat its apparent wet biases before any application.

Reliability of SM2RAIN precipitation datasets in comparison to gauge observations and hydrological modelling over arid regions

AbstractNumerous satellite‐based precipitation datasets have been successively made available. Their precipitation estimates rely on clouds properties derived from microwave and thermal sensors in a so‐named ‘top‐down’ approach. Recently, a ‘bottom‐up’ approach to infer precipitation from soil moisture (SM) estimates has resulted in the release of two new precipitation datasets (P‐datasets). One uses satellite‐based SM estimates from the European Spatial Agency (ESA) Climate Change Initiative (CCI) (SM2RAIN‐CCI) while the other uses satellite‐based SM from European Organization for the Exploitation of Meteorological Satellites (EUMETSAT) Advanced SCATterometer (ASCAT) (SM2RAIN‐ASCAT). This study assesses SM2RAIN‐ASCAT and ‐CCI reliability over two arid regions: Bolivian and Peruvian Altiplano and Pakistan (South Asia) using (a) direct comparisons with rain gauges and (b) testing the sensitivity of streamflow modelling to the P‐datasets. Selecting two different regions and different indicators helps to assess whether the P‐dataset reliability varies depending on the assessment method and location. For comparison purposes, the most reliable P‐datasets from the literature are also considered (IMERG‐E v.6, IMERG‐L v.6, IMERG‐F v.6, CHIRPS v.2 and MSWEP v.2.2). Compared to rain gauge observations and based on the modified Kling–Gupta Efficiency (KGE) values, the SM2RAIN‐ASCAT and ‐CCI are more accurate in the Altiplano than in Pakistan. This difference is explained by a more favourable physical context for satellite‐based SM estimates in the Altiplano. Over the Altiplano and despite an overall positive bias, SM2RAIN‐ASCAT describes rain gauges temporal dynamics as well as IMERG‐F v.6, CHIRPS v.2 and MSWEP v.2.2 and provides streamflow simulations very close to those obtained when using IMERG‐F v.6, CHIRPS v.2 and MSWEP v.2.2 as forcing data.

Snowmelt-Driven Streamflow Prediction Using Machine Learning Techniques (LSTM, NARX, GPR, and SVR)

Although machine learning (ML) techniques are increasingly popular in water resource studies, they are not extensively utilized in modeling snowmelt. In this study, we developed a model based on a deep learning long short-term memory (LSTM) for snowmelt-driven discharge modeling in a Himalayan basin. For comparison, we developed the nonlinear autoregressive exogenous model (NARX), Gaussian process regression (GPR), and support vector regression (SVR) models. The snow area derived from moderate resolution imaging spectroradiometer (MODIS) snow images along with remotely sensed meteorological products were utilized as inputs to the models. The Gamma test was conducted to determine the appropriate input combination for the models. The shallow LSTM model with a hidden layer achieved superior results than the deeper LSTM models with multiple hidden layers. Out of seven optimizers tested, Adamax proved to be the aptest optimizer for this study. The evaluation of the ML models was done by the coefficient of determination (R2), mean absolute error (MAE), modified Kling–Gupta efficiency (KGE’), Nash–Sutcliffe efficiency (NSE), and root-mean-squared error (RMSE). The LSTM model (KGE’ = 0.99) enriched with snow cover input achieved the best results followed by NARX (KGE’ = 0.974), GPR (KGE’ = 0.95), and SVR (KGE’ = 0.949), respectively. The outcome of this study proves the applicability of the ML models, especially the LSTM model, in predicting snowmelt driven discharge in the data-scant mountainous watersheds.

Validation of GPM IMERG V05 and V06 Precipitation Products over Iran

AbstractThis study attempts to assess the validity of the Integrated Multisatellite Retrievals for Global Precipitation Measurement (IMERG) products across Iran. Six IMERG precipitation products (IPPs) including early, late, and final runs for versions 05 and 06 were compared with precipitation data from 76 synoptic stations on a daily scale for the period from June 2014 to June 2018. According to the results, V05 performed better than V06, particularly in early and late runs. The IPPs overestimate precipitation ranging from 5% to 32%; however, IPPs tended to underestimate (overestimate) the amount of precipitation for wet (dry) areas and precipitation classes higher than 5 mm day−1 (less than 5 mm day−1). The probability of detection (POD) in IPPs was almost similar (with a median equal to 0.60), whereas other categorical validation metrics like false alarm ratio (FAR) improved in the final run. Our assessments revealed that the dependency of IPPs to the elevation was low, while the error characteristics of IPPs were strongly dependent on the climate and precipitation intensity. For instance, the systematic error varied between less than 12% in dry regions to more than 60% in wet regions. Also, according to modified Kling–Gupta efficiency (KGE) and relative bias (RBias), the performance of IPPs in winter with the highest KGE (ranging from 0.47 to 0.63) and lowest RBias (ranging from 0% to 16%) was better than other seasons. Further improvement is recommended in the satellite sensors and the precipitation retrieval algorithms to achieve a reliable precipitation source.

A 439-year simulated daily discharge dataset (1861–2299) for the upper Yangtze River, China

Abstract. The outputs of four global climate models (GFDL-ESM2M, HadGEM2-ES, IPSL-CM5A-LR and MIROC5), which were statistically downscaled and bias corrected, were used to drive four hydrological models (Hydrologiska Byråns, HBV; Soil and Water Assessment Tool, SWAT; Soil and Water Integrated Model, SWIM; and Variable Infiltration Capacity, VIC) to simulate the daily discharge at the Cuntan hydrological station in the upper Yangtze River from 1861 to 2299. As the performances of hydrological models in various climate conditions could be different, the models were first calibrated in the period from 1979 to 1990. Then, the models were validated in the comparatively wet period, 1967–1978, and in the comparatively dry period, 1991–2002. A multi-objective automatic calibration programme using a univariate search technique was applied to find the optimal parameter set for each of the four hydrological models. The Nash–Sutcliffe efficiency (NSE) of daily discharge and the weighted least-squares function (WLS) of extreme discharge events, represented by high flow (Q10) and low flow (Q90), were included in the objective functions of the parameterization process. In addition, the simulated evapotranspiration results were compared with the GLEAM evapotranspiration data for the upper Yangtze River basin. For evaluating the performances of the hydrological models, the NSE, modified Kling–Gupta efficiency (KGE), ratio of the root-mean-square error to the standard deviation of the measured data (RSR) and Pearson's correlation coefficient (r) were used. The four hydrological models reach satisfactory simulation results in both the calibration and validation periods. In this study, the daily discharge is simulated for the upper Yangtze River under the preindustrial control (piControl) scenario without anthropogenic climate change from 1861 to 2299 and for the historical period 1861–2005 and for 2006 to 2299 under the RCP2.6, RCP4.5, RCP6.0 and RCP8.5 scenarios. The long-term daily discharge dataset can be used in the international context and water management, e.g. in the framework of Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) by providing clues to what extent human-induced climate change could impact streamflow and streamflow trend in the future. The datasets are available at: https://doi.org/10.4121/uuid:8658b22a-8f98-4043-9f8f-d77684d58cbc (Gao et al., 2019).

Performance Assessment of SM2RAIN-CCI and SM2RAIN-ASCAT Precipitation Products over Pakistan

Accurate estimation of precipitation from satellite precipitation products (PPs) over the complex topography and diverse climate of Pakistan with limited rain gauges (RGs) is an arduous task. In the current study, we assessed the performance of two PPs estimated from soil moisture (SM) using the SM2RAIN algorithm, SM2RAIN-CCI and SM2RAIN-ASCAT, on the daily scale across Pakistan during the periods 2000–2015 and 2007–2015, respectively. Several statistical metrics, i.e., Bias, unbiased root mean square error (ubRMSE), Theil’s U, and the modified Kling–Gupta efficiency (KGE) score, and four categorical metrics, i.e., probability of detection (POD), false alarm ratio (FAR), critical success index (CSI), and Bias score, were used to evaluate these two PPs against 102 RGs observations across four distinct climate regions, i.e., glacial, humid, arid and hyper-arid regions. Total mean square error (MSE) is decomposed into systematic (MSEs) and random (MSEr) error components. Moreover, the Tropical Rainfall Measurement Mission Multi-Satellite Precipitation Analysis (TRMM TMPA 3B42v7) was used to assess the performance of SM2RAIN-based products at 0.25° scale during 2007–2015. Results shows that SM2RAIN-based product highly underestimated precipitation in north-east and hydraulically developed areas of the humid region. Maximum underestimation for SM2RAIN-CCI and SM2RIAN-ASCAT were 58.04% and 42.36%, respectively. Precipitation was also underestimated in mountainous areas of glacial and humid regions with maximum underestimations of 43.16% and 34.60% for SM2RAIN-CCI. Precipitation was overestimated along the coast of Arabian Sea in the hyper-arid region with maximum overestimations for SM2RAIN-CCI (SM2RAIN-ASCAT) of 59.59% (52.35%). Higher ubRMSE was observed in the vicinity of hydraulically developed areas. Theil’s U depicted higher accuracy in the arid region with values of 0.23 (SM2RAIN-CCI) and 0.15 (SM2RAIN-ASCAT). Systematic error components have larger contribution than random error components. Overall, SM2RAIN-ASCAT dominates SM2RAIN-CCI across all climate regions, with average percentage improvements in bias (27.01% in humid, 5.94% in arid, and 6.05% in hyper-arid), ubRMSE (19.61% in humid, 20.16% in arid, and 25.56% in hyper-arid), Theil’s U (9.80% in humid, 28.80% in arid, and 26.83% in hyper-arid), MSEs (24.55% in humid, 13.83% in arid, and 8.22% in hyper-arid), MSEr (19.41% in humid, 29.20% in arid, and 24.14% in hyper-arid) and KGE score (5.26% in humid, 28.12% in arid, and 24.72% in hyper-arid). Higher uncertainties were depicted in heavy and intense precipitation seasons, i.e., monsoon and pre-monsoon. Average values of statistical metrics during monsoon season for SM2RAIN-CCI (SM2RAIN-ASCAT) were 20.90% (17.82%), 10.52 mm/day (8.61 mm/day), 0.47 (0.43), and 0.47 (0.55) for bias, ubRMSE, Theil’s U, and KGE score, respectively. TMPA outperformed SM2RAIN-based products across all climate regions. SM2RAIN-based datasets are recommended for agricultural water management, irrigation scheduling, flood simulation and early flood warning system (EFWS), drought monitoring, groundwater modeling, and rainwater harvesting, and vegetation and crop monitoring in plain areas of the arid region.