Search

Deep learning (DL) and machine learning (ML) are widely used in hydrological modelling, which plays a critical role in improving the accuracy of hydrological predictions. However, the trade-off between model performance and computational cost has always been a challenge for hydrologists when selecting a suitable model, particularly for probabilistic post-processing with large ensemble members. This study aims to systematically compare the quantile regression forest (QRF) model and countable mixtures of asymmetric Laplacians long short-term memory (CMAL-LSTM) model as hydrological probabilistic post-processors. Specifically, we evaluate their ability in dealing with biased streamflow simulations driven by three satellite precipitation products across 522 nested sub-basins of the Yalong River basin in China. Model performance is comprehensively assessed using a series of scoring metrics from both probabilistic and deterministic perspectives. Our results show that the QRF model and the CMAL-LSTM model are comparable in terms of probabilistic prediction, and their performances are closely related to the flow accumulation area (FAA) of the sub-basin. The QRF model outperforms the CMAL-LSTM model in most sub-basins with smaller FAA, while the CMAL-LSTM model has an undebatable advantage in sub-basins with FAA larger than 60 000 km 2 in the Yalong River basin. In terms of deterministic predictions, the CMAL-LSTM model is preferred, especially when the raw streamflow is poorly simulated and used as input. However, setting aside the differences in model performance, the QRF model with 100-member quantiles demonstrates a noteworthy advantage by exhibiting a 50 % reduction in computation time compared to the CMAL-LSTM model with the same ensemble members in all experiments. As a result, this study provides insights into model selection in hydrological post-processing and the trade-offs between model performance and computational efficiency. The findings highlight the importance of considering the specific application scenario, such as the catchment size and the required accuracy level, when selecting a suitable model for hydrological post-processing. [ABSTRACT FROM AUTHOR]

Aim: Gross primary productivity (GPP) of vegetation is an important component of the carbon balance of terrestrial ecosystems. With the development of science and technology, there has been a proliferation of multiple GPP products, yet large uncertainties exist in the global GPP estimated by different GPP products. The objectives of this study were to assess the uncertainty of 45 GPP products in 4 different categories (machine learning, Fluxnet, light use efficiency, land surface model) and to perform an attribution analysis. Location: Global. Time period: 2003–2010. Major taxa studied: Terrestrial ecosystems. Methods: The three‐cornered hat (TCH) method was used to assess GPP uncertainty because it does not require a priori knowledge of the true GPP values. We also used generalized additive models to explore the contribution of different factors to GPP uncertainty. Results: The results show that the magnitude of absolute uncertainty (AU) is generally higher in the GPP high‐value region and also indicate large spatial variability of relative uncertainty (RU) among products. Deciduous broadleaf forest (20.2 g C/m2/year) and deciduous needleleaf forest (27.2 g C/m2/ year) have lower uncertainty compared to that of woody savanna (128.4 g C/m2/ year) and savanna (128.7 g C/m2/year). The RU in machine learning (ML) products overall is low (average value = 14.3%), with ORCHIDEE‐CNP (weight = 4.19%), being more recognized by GPP products are more consistent in the magnitude of spatial variability (similar in the standard deviations) and more uncertain in spatial distribution patterns. Excluding the combined effects of other factors, soil moisture (15.2%) and precipitation (8.3%) are the two factors that contribute most to GPP uncertainty, indicating that focusing on water constraints is a guarantee of GPP product quality. Main conclusions: The TCH approach quantifies the uncertainty of GPP products well and the spatial distribution pattern variability contributes more uncertainty compared to the magnitude of spatial variability. Future GPP estimates could consider combining the advantages of different GPP products to limit uncertainty. [ABSTRACT FROM AUTHOR]

Meteorological and soil drought is disastrous to natural and social systems. It is expected that the occurrence of meteorological and soil drought compound events will become more frequent and extreme in the future. The propagation process of meteorological drought to soil drought plays an essential role in the occurrence of drought. However, the drought propagation process is not clear, especially the quantification of drought characteristic propagation process is rarely realized. This study constructs drought intensity propagation index (DIP) and drought duration propagation index (DDP), and then puts forward drought propagation partition in China. Furthermore, this study preliminarily discussed the driving factors of drought propagation and the relationship between land‐atmosphere interaction and main drought propagation partition (arid type drought propagation area, peer‐to‐peer type drought propagation area, and humid type drought propagation area). The results show that DDP and DIP are significantly negatively correlated in China from 1981 to 2020. From southeast to northwest, the propagation of drought intensity from meteorological drought to soil drought gradually increased and the propagation of drought duration gradually decreased. We further concluded that 68% of regions with similar intensity and duration of meteorological drought and soil drought (peer‐to‐peer type drought propagation area) are concentrated in semiarid and dry subhumid, which is exactly the area with strong land‐atmosphere interaction. In all land use types, grasslands are most prone to peer‐to‐peer propagation of drought. This study further reveals the law and mechanism of drought propagation and provides a new idea and attempt to clarify the drought propagation process. Key Points: Drought intensity propagation index and drought duration propagation index for quantifying drought propagation characteristics are proposedDividing drought propagation partitionDiscovery of drought propagation close peer‐to‐peer in areas with strongest land‐atmosphere coupling [ABSTRACT FROM AUTHOR]

Accurate and reliable near‐real‐time satellite precipitation estimation is of great importance for operational large‐scale flood forecasting and drought monitoring. The state‐of‐the‐art precipitation post‐processing model is based on a deterministic approach to construct relationships between satellites estimates and ground observations. We propose a probabilistic postprocessor, the Probabilistic Post‐Processing of Near‐Real‐Time Satellite Precipitation Estimates using Quantile Regression Forests (QRF4P‐NRT), based on quantile modeling, yielding both deterministic and probabilistic predictions. The experimental design incorporates different solutions of near‐real‐time predictors to further improve the model performance. Using the Integrated Multi‐satellitE Retrievals Early Run for Global Precipitation Measurement Mission (IMERG‐E) product as an example, we illustrate that the proposed method significantly improves the overall quality of the raw IMERG‐E and is also superior to the bias‐corrected product (IMERG Final Run, IMERG‐F) at daily scale in a complex mountain basin. Evaluations of the corrected IMERG‐E, raw IMERG‐E, and IMERG‐F using ground observation show that the corrected IMERG‐E improves correlation coefficients (0.7), mean error (−0.14 mm/day) and root mean square error (3.3 mm/day) relative to the raw IMERG‐E (0.31, −0.72 and 5.5 mm/day) and IMERG‐F (0.34, −0.09 and 6.0 mm/day). The error decomposition further confirms that the QRF4P‐NRT improves on the various deficiencies of the raw IMERG‐E product. The ensemble assessment also demonstrates that the quantile outputs provide reliable prediction spread and sharp prediction intervals. The promising results indicate the great potential of the proposed method for probabilistic post‐processing for near‐real‐time satellite precipitation estimates, and for further applications such as hydrological ensemble forecasting. Plain Language Summary: Errors in near‐real‐time satellite precipitation estimates limit their applications. The use of error correction models is better able to reduce the errors. However, current deterministic error correction models reduce errors while losing uncertain information. In this study, we propose a probabilistic error correction method that has been used in the field of ensemble numerical weather forecasts. While reducing the error, it is also possible to quantify the probabilistic information. Our method obtains the best score compared to both the raw product and bias‐corrected product. This is of great interest for the application of near‐real‐time satellite precipitation estimates and can be further applied to operational flood forecasting and drought monitoring. Key Points: An ensemble postprocessor based on quantile regression forests for bias correction of satellite precipitation estimates is proposedThe Probabilistic Post‐Processing of Near‐Real‐Time Satellite Precipitation Estimates using Quantile Regression Forests remarkably improved raw satellite precipitation estimates and provide reliable probabilistic outputs in a near‐real‐time wayA static proxy of dynamic near‐real‐time predictor is an acceptable solution for operational application [ABSTRACT FROM AUTHOR]

Long‐term, accurate soil moisture (SM) estimates are crucial for climate, hydrology, and agriculture. Multisource SM products are available at high spatiotemporal coverage, but their uncertainties have yet to be fully evaluated in China. Ten SM products (i.e., reanalysis, assimilation, and satellite) were intercompared and evaluated over 18 hydroclimatic regions of China during 2000–2018 at different scales. The results show that all SM products share similar spatial patterns but exhibit significant differences in regions and products. The interannual fluctuations and differences between these products are relatively small, except for the anomalies of ESA CCI ACTIVE and PASSIVE in southern China. At a seasonal scale, the estimates of NCEP R2, ACTIVE, and PASSIVE for specific regions are significantly different from other products. Moreover, all SM products correlate well with precipitation in summer and have their own dominant areas. Among them, ERA5 shows a comprehensive potential for application in China. The temporal correlation coefficient (R), root mean square difference (RMSD), and standard deviation (σ) demonstrate that ERA5 performs better in southern China, while CLSM performs better in the northeast and throughout China, with R exceeding 0.7 for both products. For the Equitable Threat Score (ETS), ERA5 has a better overall performance with scores greater than 0.03 distributed in more regions of China, followed by ACTIVE and CFSR/CFSV2. This study not only provides a reasonable reference for the targeted selection of SM products in different hydroclimatic regions of China but also establishes a preliminary data basis for the integration of multiple products. Key Points: Ten soil moisture products from multiple sources were intercompared and evaluated in 18 hydroclimatic regions of China during 2000–2018All products share similar spatial patterns to precipitation in China, but with subtle interannual and seasonal variationThe optimal SM products are found in different regions of China, among which ERA5 shows the superior overall performance(R>0.7&ETS¯>0.03) [ABSTRACT FROM AUTHOR]

It is very important to quantify errors of precipitation estimation products. However, the existing methods do not describe all error components and are therefore not comprehensive enough. In this study, we propose a four‐component error decomposition method (4CED) that decomposes the total errors of precipitation products into four independent parts: hit positive bias, hit negative bias, false bias, and missed bias. And we use it to evaluate the performance of the three latest satellite precipitation products in the eastern monsoon region of China. Our study reveals 4CED has apparent improvements compared with the previous method. Results also provide new insights for tracking error sources and quantifying the error magnitudes of precipitation products. Moreover, the proposed 4CED can be extended to different spatial and temporal scales. Our new method will not only contribute to product upgrades, but also provide guidance for potential applications. Plain Language Summary: Precipitation products still have biases compared to the precipitation that actually occurs. The possible biases are divided into precipitation overestimation and precipitation underestimation. And they can be further divided into hit precipitation overestimation, false alarm overestimation, hit precipitation underestimation and missed underestimation. Recognizing and understanding these biases can better leverage the full effectiveness of precipitation products. Here, we have quantified these biases using statistical metrics and found that we may have underestimated the magnitude of the biases when using the classical method, especially in some regions with complex topography where precipitation estimation remains a challenge. Key Points: Four‐component error decomposition (4CED) is proposed to quantify positive and negative error components for precipitation productsThe 4CED incorporates and outdoes the original error decomposition methodThe 4CED can reveal distinct spatial patterns and seasonal characteristics of individual error components [ABSTRACT FROM AUTHOR]

Human activities and climate change have changed the vegetation in China. The analysis of the changes in vegetation that have occurred over the past 30 years in China remains a great challenge due to intense human activity and lack of field observations. The use of various Normalized Difference Vegetation Index (NDVI) datasets to study vegetation coverage changes has received much attention. In this paper, we selected the early versions of Advanced Very High Resolution Radiometer (AVHRR) Global Inventory Monitoring and Modelling Studies (GIMMS), GIMMS3g (third generation GIMMS NDVI from AVHRR sensors) and Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI data including the fusion data (GIMMS+MODIS). We analysed spatial and temporal changes in vegetation cover in different ecosystems and basins in the mainland of China. Different contributions of ecosystems and variations in NDVI trends exist in different ecosystems in 17 basins. The results show that different NDVI from different data sources yield different results: (1) Vegetation increased in 74.62–77.7% of the area of the Chinese mainland during 1982–2015, mainly in the Yellow River and the middle reaches of the Yangtze River basin; (2) 2000–2017 MODIS NDVI in mainland China has increased more area (79.67%). (3) Farmland and Forest ecosystems were significantly enhanced in the eastern monsoon region; (4) High-resolution NDVI can provide more information than domain average NDVI. GIMMS and MODIS NDVI data have complementary spatial and temporal distributions. Our study improves the understanding of vegetation dynamics over long time periods and large areas and, moreover, has potential for supporting ecological managers in mainland China. [ABSTRACT FROM AUTHOR]

Endorheic and arid regions around the world are suffering from serious drought problems. In this study, a drought forecasting system based on eight state-of-the-art climate models from the North American Multi-Model Ensemble (NMME) and a Distributed Time-Variant Gain Hydrological Model (DTVGM) was established and assessed over the upstream and midstream of Heihe River basin (UHRB and MHRB), a typical arid endorheic basin. The 3-month Standardized Precipitation Index (SPI3) and 1-month Standardized Streamflow Index (SSI1) were used to capture meteorological and hydrological drought, and values below -1 indicate drought events. The skill of the forecasting systems was evaluated in terms of anomaly correlation (AC) and Brier score (BS) or Brier skill score (BSS). The predictability for meteorological drought was quantified using AC and BS with a “perfect model" assumption, referring to the upper limit of forecast skill. The hydrological predictability was to distinguish the role of initial hydrological conditions (ICs) and meteorological forcings, which was quantified by root-mean-square error (RMSE) within the ESP (Ensemble Streamflow Prediction) and reverse ESP framework. The UHRB and MHRB showed season-dependent meteorological drought predictability and forecast skill, with higher values during winter and autumn than that during spring. For hydrological forecasts, the forecast skill in the UHRB was higher than that in MHRB. Predicting meteorological droughts more than 2 months in advance became difficult because of complex climate mechanisms. However, the hydrological drought forecasts could show some skills up to 3-6 lead months due to memory of ICs during cold and dry seasons. During wet seasons, there are no skillful hydrological predictions from lead month 2 onwards because of the dominant role of meteorological forcings. During spring, the improvement of hydrological drought predictions was the most significant as more streamflow was generated by seasonal snowmelt. Besides meteorological forcings and ICs, human activities have reduced the hydrological variability and increased hydrological drought predictability during the wet seasons in the MHRB. [ABSTRACT FROM AUTHOR]

High‐resolution modeling became popular in recent years due to the availability of multisource observations, advances in understanding fine‐scale processes, and improvements in computing facilities. However, modeling of hydrological changes over mountainous regions is still a great challenge due to the sensitivity of highland water cycle to global warming, tightly coupled hydrothermal processes, and limited observations. Here we show a successful high‐resolution (3 km) land surface modeling over the Sanjiangyuan region located in the eastern Tibetan plateau, which is the headwater of three major Asian rivers. By developing a new version of a Conjunctive Surface‐Subsurface Process model named as CSSPv2, we increased Nash‐Sutcliffe efficiency by 62–130% for streamflow simulations due to the introduction of a storage‐based runoff generation scheme, reduced errors by up to 31% for soil moisture modeling after considering the effect of soil organic matter on porosity and hydraulic conductivity. Compared with ERA‐Interim and Global Land Data Assimilation System version 1.0 reanalysis products, CSSPv2 reduced errors by up to 30%, 69%, 92%, and 40% for soil moisture, soil temperature, evapotranspiration, and terrestrial water storage change, respectively, as evaluated against in situ and satellite observations. Moreover, CSSPv2 well captured the elevation‐dependent ground temperature warming trends and the decreased frozen dates during 1979–2014, and significant increasing trends (p < 0.05) in evapotranspiration and terrestrial water storage during 1982–2011 and 2003–2014 respectively, while ERA‐Interim and Global Land Data Assimilation System version 1.0 showed no trends or even negative trends. This study implies the necessity of developing high‐resolution land surface models in realistically representing hydrological changes over highland areas that are sentinels to climate change. Key Points: CSSPv2 land model was developed over headwaters with conjunctive surface‐subsurface flows, storage‐based runoff, and soil organic matterCSSPv2 outperformed global reanalysis for simulations of hydrological cycle as validated with multisource observationsCSSPv2 captured elevation‐dependent warming, increasing ET and TWS in the satellite era, while global reanalysis failed over headwaters [ABSTRACT FROM AUTHOR]

High-resolution data on the spatial pattern of water use are a prerequisite for appropriate and sustainable water management. Based on one well-validated hydrological model, the Distributed Time Variant Gains Model (DTVGM), this paper obtains reliable high-resolution spatial patterns of irrigation, industrial and domestic water use in continental China. During the validation periods, ranges of correlation coefficient (R) and Nash-Sutcliffe efficiency (NSE) coefficient are 0.67–0.96 and 0.51–0.84, respectively, between the observed and simulated streamflow of six hydrological stations, indicating model applicability to simulate the distribution of water use. The simulated water use quantities have relative errors (RE) less than 5% compared with the observed. In addition, the changes in streamflow discharge were also correctly simulated by our model, such as the Zhangjiafen station in the Hai River basin with a dramatic decrease in streamflow, and the Makou station in the Pearl River basin with no significant changes. These changes are combined results of basin available water resources and water use. The obtained high-resolution spatial pattern of water use could decrease uncertainty of hydrological simulation and guide water management efficiently.Editor M.C. Acreman; Associate editor X. Fang [ABSTRACT FROM PUBLISHER]

The specification of model parameters in numerical weather prediction ( NWP) models has great influence on model performance. However, how to specify model parameters properly is not a trivial task because a typical NWP model like the Weather Research and Forecasting ( WRF) model contains many model parameters and many model outputs. This article presents the results of an investigation into the sensitivities of different WRF model outputs to the specification of its model parameters. Using a global sensitivity analysis method, the sensitivities are evaluated for surface meteorological variables such as precipitation, surface air temperature, humidity and wind speed, as well as for atmospheric variables such as total precipitable water, cloud cover, boundary-layer height and outgoing long-wave radiation at the top of the atmosphere, all simulated by the WRF model using different model parameters. The goal of this study is to identify the parameters that exert most influence on the skill of short-range meteorological forecasts. The study was performed over the Greater Beijing Region of China. A total of 23 adjustable parameters from seven different physical parametrization schemes were considered. The results indicate that parameter sensitivities vary with different model outputs. However, some of the 23 model parameters considered are shown to be sensitive to all model outputs evaluated, while other parameters may be sensitive to a particular output. The sensitivity results from this research are a basis for further optimizations of the WRF model parameters. [ABSTRACT FROM AUTHOR]

Parameter specification is an important source of uncertainty in large, complex geophysical models. These models generally have multiple model outputs that require multiobjective optimization algorithms. Although such algorithms have long been available, they usually require a large number of model runs and are therefore computationally expensive for large, complex dynamic models. In this paper, a multiobjective adaptive surrogate modeling-based optimization (MO-ASMO) algorithm is introduced that aims to reduce computational cost while maintaining optimization effectiveness. Geophysical dynamic models usually have a prior parameterization scheme derived from the physical processes involved, and our goal is to improve all of the objectives by parameter calibration. In this study, we developed a method for directing the search processes toward the region that can improve all of the objectives simultaneously. We tested the MO-ASMO algorithm against NSGA-II and SUMO with 13 test functions and a land surface model - the Common Land Model (CoLM). The results demonstrated the effectiveness and efficiency of MO-ASMO. [ABSTRACT FROM AUTHOR]

There is an increasing focus on the usefulness of climate model--based seasonal precipitation forecasts as inputs for hydrological applications. This study reveals that most models from the North American Multi-Model Ensemble (NMME) have potential to forecast seasonal precipitation over 17 hydroclimatic regions in continental China. In this paper, we evaluated the NMME precipitation forecast against observations. The evaluation indices included the correlation coefficient (R), relative root-mean-square error (RRMSE), rank histogram (RH), and continuous ranked probability skill score (CRPSS).We presented the RRMSE-R diagram to distinguish differences between the performances of individual models. We find that the predictive skill is seasonally and regionally dependent, exhibiting higher values in autumn and spring and lower values in summer. Higher predictive skill is observed over most regions except the southeastern monsoon regions, which may be attributable to local climatology and variability. Among the 11 NMME models, CFS, especially CFSv2, exhibits the best predictive skill. The GFDL and NASA models, which are followed by CMC, perform worse than CFS. The performances of IRI and CCSM3 are relatively worse than that of the other models. The forecast skills are significantly improved in multi-model mean forecasts based on simple model averaging (SMA). The improvement is more obvious for Bayesian model averaging (BMA), which is employed to further improve the forecast skill and address model uncertainty using multiple model outputs, than individual model and SMA. [ABSTRACT FROM AUTHOR]