Figure - available from: Stochastic Environmental Research and Risk Assessment
This content is subject to copyright. Terms and conditions apply.
The map showing Harris reservoir at the Alabama–Coosa–Tallapoosa River Basin (ACT) and Navajo reservoir at the Upper Colorado River Basin (UCRB) in the United States
Source publication
In this study, we developed and evaluated a hybrid framework for reservoir inflow forecast. This framework is unprecedented, which integrates new quasi-globally available observation-, satellite-, or model-based datasets using machine learing models to forecast inflow at the local scale. Under this framework, we compared random forests, gradient bo...
Similar publications
The development of a computationally efficient machine learning-based framework to understand the underlying causes for biases in climate model simulated fields is presented in this study. The framework consists of a two-step approach, with the first step involving the development of a Random Forest (RF) model, trained on observed data of the clima...
Citations
... Skillful precipitation forecasts are vital for environmental modeling, energy production, and disaster risk management [4][5][6][7][8][9][10][11][12][13]. As an important climate forcing, skillful precipitation forecasts usually lead to skillful hydrological forecasts [14][15][16][17]. Operational seasonal climate forecasting has been performed at major climate centers around the world, including the Geophysical Fluid Dynamics Laboratory (GFDL), the National Centers for Environmental Prediction (NCEP) and the European Centre for Medium-Range Weather Forecasts [1,18,19]. ...
Global climate models (GCMs) provide valuable forecasts of precipitation around the world. This paper has presented an in-depth investigation of the overlapping versus differing information for 2 sets of GCM forecasts based on the classic set operations. Specifically, by using the coefficient of determination to measure the amount of information of precipitation observations contained in GCM forecast, the common part of the 2 sets of forecasts is quantified by the intersection operation and the unique part of 1 set of forecasts is quantified by the difference operation. A case study is devised for the global precipitation forecasts in December-January-February generated by the Seamless System for Prediction and EArth System Research (SPEAR) and the Climate Forecast System version 2 (CFSv2). Their overlapping and differing information are diagnosed. It is found that significant information common to the 2 sets of forecasts exists over 54.61% of global land grid cells, significant information unique to SPEAR forecasts over 23.59% of global land grid cells, and significant information unique to CFSv2 forecasts over 18.15% of global land grid cells. While the information unique to the SPEAR forecasts suggests that the SPEAR forecasts provide new information compared to the CFSv2 forecasts and the information unique to the CFSv2 forecasts suggests that the CFSv2 forecasts also provide new information compared to the SPEAR forecasts, the common information of the 2 sets of forecasts indicates that they present substantial amount of similar information. Overall, the diagnosis of the overlapping and differing information for different sets of GCM forecasts yields insights into GCM predictive performances.
... Hybrid method, integrating data-driven methods (statistical or machine learning) and predictions from dynamical, physics-based models (such as climate, hydrology, and crop growth models), is a promising way of enhancing the prediction skill in different forecasting (Slater et al., 2023). Many studies proved that climate prediction models combined with ML improved the accuracy of seasonal climate forecast and extreme events (Ju-Young et al., 2020;Tian et al., 2022). For crop yield forecasting, some researches merged climate prediction systems and crop growth models (Doi et al., 2020;Ogutu et al., 2018), more like land surface model (Boas et al., 2023). ...
Effective yield forecasting is a key strategy for adaptation when facing food loss to climate variability. Currently, solar-induced chlorophyll fluorescence (SIF) is an emerging remote-sensing index owing to its high relevance to plant photosynthesis, and sensitivity to drought. Despite many studies have focused on drought monitoring and production assessment by SIF, little puts it into practice for in-season yield prediction. In this study, we combined multi-source satellite and meteorological data, especially coupling with subseasonal-to-seasonal (S2S) dynamic atmospheric prediction climate model (IAP-CAS FGOALS-f2), with an addition of SIF, to predict maize yields in the U.S. Corn Belt, based on the developed machine learning dynamical hybrid model (MHCF). By comparison, we found that SIF performed well in the correlation analysis with yield, with average correlations up to 0.719 in August. Then we utilized different algorithms, different models (S2S data for MHCF, climate data for the Benchmark), and different input combinations to train and predict maize yields. All four algorithms using SIF significantly improved prediction performance. S2S + VIs + SIF combination (FGOALS-f2、NDVI、EVI、SIF) can achieve the best performance, while the XGBoost algorithm reached 0.897 of R². With the best combination, it can achieve 4 months before maize harvest (with R² value of 0.85, and RMSE < 13 bu/acre). In 2012, the year had a severe drought, although predictive capability decreased in all the predictions, the models with SIF still maintained robust and improved the prediction (improved R² by 5.92%, and RMSE decreased by 18.08% of XGBoost). According to the study, it can be expected, the combination of MHCF and SIF will play a greater role in subseasonal yield prediction. We also provide an operational proposition of hybrid yield forecasting method to fully integrating climate prediction and machine learning for early notice of crop production losses.
... Seasonal Climate Forecasts (SCFs) provide a middle-to long-range outlook of changes in the Earth system over periods of a few weeks to several months, through predictable changes in some of the slow-varying components of the system, such as ocean temperatures (Johnson et al. 2019). Skilful seasonal precipitation forecasts provide invaluable support to a wide range of sectors, including agriculture, construction, mining, hydrology, and water resources management (Falamarzi et al. 2023;Jin et al. 2022;Merryfield et al. 2020; the Centre for International Economics 2014; Tian et al. 2021). For example, site-specific information on monthly rainfall for the growing season, typically up to three months ahead, can help farmers make informed decisions about which crop types or varieties to plant (Weisheimer and Palmer 2014). ...
Seasonal precipitation forecasting is vital for weather-sensitive sectors. Global Circulation Models (GCM) routinely produce ensemble Seasonal Climate Forecasts (SCFs) but suffer from issues like low forecast resolution and skills. To address these issues in this study, we introduce a post-processing method, Quantile Ensemble Bayesian Model Averaging (QEBMA). It utilises quantiles from a GCM ensemble forecast to create a pseudo-ensemble forecast. Through their reasonable linear relationships with observations, each pseudo-member connects a hurdle distribution with a point mass at zero for dry months and a gamma distribution for wet months. These distributions are mixed to construct a forecast probability distribution with their weights, proportional to the quantiles’ historical forecast performance. QEBMA is applied to three GCMs, including GloSea5 from the United Kingdom, ECMWF from Europe and ACCESS-S1 from Australia, for monthly precipitation forecasts in 32 locations across four climate zones in Australia. Leave-one-month-out cross-validation results illustrate that QEBMA enhances forecast skills compared to raw GCMs and other post-processing techniques, including quantile mapping and Extended Copula Post-Processing (ECPP), for forecast lead time of 0 to 2 months, based on five metrics. The skill improvements achieved by QEBMA are often statistically significant, particularly when compared to raw GCM forecasts across the 32 study locations. Among these post-processing models, only QEBMA consistently outperforms the SCF benchmark climatology, offering a promising alternative for improving seasonal precipitation forecasts.
... Dams are built to deliver water for human consumption, irrigation and water supply, or to be used in hydropower generation, flood control, recreation, navigation, and sediment control (Jackson & Brown 2023). Dams in Ethiopia are used mainly for hydropower generation (Grand Ethiopian Renaissance Dam, Gibe-I and -III, Tekeze, Koysha, Aba Samuel, Genale Dawa III, Genale Dawa VI, and Melka Wakena), water supply (Geferesa, Legedadi, etc.), and irrigation (Kessem, Tendaho, Arjo In addition, in recent years, models based on machine learning techniques have been developed/analyzed for forecasting reservoir inflow, estimating instantaneous peak flow, and analyzing hydrological risk (Jimeno-Sáez et al. 2017;Gabriel-Martin et al. 2019;Hong et al. 2020;Huang et al. 2022;Tian et al. 2022). Senent-Aparicio et al. (2019) estimated the instantaneous peak flow by combining the Soil and Water Assessment Tool (SWAT) and machine learning models. ...
Flooding due to overtopping during peak flow in embankment dams primarily causes dam failure. The Kessem River watershed of the Awash basin in the Rift Valley of the Afar region in Ethiopia has been studied intricately to predict the causes of the Kessem Dam safety using machine learning predictive models and Risk Management Centre-Reservoir Frequency Analysis (RMC-RFA). Recently developed recurrent neural network (RNN) predictive models with hybrid with Soil Conservation Service Curve Number (SCS-CN) were used for simulation of the river flow. Peak daily inflow to the reservoir is predicted to be 467.72, 435.88, and 513.55 m3/s in 2035, 2061, and 2090, respectively. The hydrologic hazard analysis results show 2,823.57 m3/s and 935.21 m; 2,126.3 m3/s and 934.18 m; and 11,491.1 m3/s and 942.11 m peak discharge and maximum reservoir water level during the periods of 2022–2050, 2051–2075, and 2076–2100, respectively, for 0.0001 annual exceedance probability (AEP). The Kessem Dam may potentially be overtopped by a flood with a return period of about 10,000 years during the period of 2076–2100. Quantitative hydrologic risk assessment of the dam is used for dam safety evaluation to decide whether the existing structure provides an adequate level of safety, and if not, what modifications are necessary to improve the dam's safety.
... Considering the complexity of residual patterns in hydrologic and hydraulic models (Beven, 2016), it is necessary to conduct numerical experiments based on pre-defined model errors as this is the first study to apply the MCMC method for estimating BMA parameters. Given the Gaussian assumption of model errors in the EM algorithm and the relatively small size (less than 10) of the model ensemble in practice (Cao et al., 2021;Huo et al., 2019;Moknatian & Mukundan, 2023;Raftery et al., 2005;Tian et al., 2021;Tsai, 2010;Vrugt et al., 2008), a total of 10 sets of hydrologic data (100 days of daily observed water stage data, denoted as D, of the White river in Indiana, see Table A1 for basic statistics of daily water stage data) that contain random errors (denoted as ε) are generated. These synthetic data sets are created assuming a normal distribution with zero mean and standard deviations ranging from 0.06 m (0.2 ft) to 0.30 m (1 ft) with an increment of 0.06 m (0.2 ft) (the other numerical experiment including model errors with larger magnitudes around 1 m has also been conducted and the conclusions are consistent with those obtained in the presented one). ...
As all kinds of numerical models are emerging in hydrologic and hydraulic engineering, Bayesian model averaging (BMA) is one of the popular multi‐model methods used to account for various uncertainties in the flood modeling process and generate robust ensemble predictions. The reliability of BMA parameters (weights and variances) determines the accuracy of BMA predictions. However, the uncertainty in BMA parameters with fixed values, which are usually obtained from Expectation‐Maximization (EM) algorithm, has not been adequately investigated in BMA‐related applications over the past few decades. Given the limitations of the default EM algorithm, Metropolis‐Hastings (M‐H) algorithm, which is one of the most widely used algorithms in Markov Chain Monte Carlo (MCMC), is proposed to estimate BMA parameters. Both numerical experiments and one‐dimensional Hydrologic Engineering Center‐River Analysis System models are employed to examine the applicability of M‐H algorithm with multiple independent Markov chains. The performances of EM and M‐H algorithms are compared based on the daily water stage predictions from 10 model members. Results show that BMA weights estimated from both algorithms are comparable, while BMA variances obtained from M‐H algorithm are closer to the given variances in the numerical experiment. Moreover, the normal proposal distribution used in M‐H algorithm can yield narrower distributions for BMA weights than those from the uniform proposal distribution. Overall, MCMC approach with multiple chains can provide more information associated with the uncertainty of BMA parameters and its performance is better than the default EM algorithm in terms of both deterministic and probabilistic evaluation metrics as well as algorithm flexibility.
... Many studies indicated that the CHIRPS satellite dataset is suitable for describing the spatial distribution of precipitation and able to capture the occurrence and characteristics of drought events, suggesting that the CHIRPS dataset could be used as an alternative precipitation source for monitoring drought [22], [42], [50], [57]. Furthermore, there were a significant number of studies used CHIRPS as data set for machine learning models [5], [10], [24], [36], [37], [53], [59], [62]. In this study, the CHIRPS precipitation dataset for 1981-2021 was obtained from CHIRPS: Rainfall Estimates from Rain Gauge and Satellite Observations | Climate Hazards Center -UC Santa Barbara (ucsb.edu). ...
Drought is a recurring natural disaster that can cause significant damage to agricultural production, human livelihoods, and the environment. Drought forecasting is an important tool for managing and mitigating the impacts of drought. This study aimed to improve drought forecasting through the use of machine learning models. Specifically, the study evaluated the performance of three machine learning models, namely Extreme Learning Machine (ELM), Random Forest (RF), and Support Vector Regression (SVR), for forecasting Standardized Precipitation Index (SPI) drought. These models were trained using precipitation data of Hiran region, Somalia from 1980 to 2021, to evaluate their ability to accurately predict drought conditions. The results showed that the SVR model performed the best, with an R 2 value of 0.753, MAE of 0.344, and RMSE of 0.488. The ELM and RF models also performed well. The study highlights the potential of machine learning models to improve drought forecasting, and the importance of evaluating multiple models to select the one that performs best for a specific dataset.
... For example, Bradley et al. (2015) enhance seasonal streamflow forecasts in the Blue Nile basin by postprocessing available ensemble forecasts based on the relationship between streamflow and an ENSO index. Tian et al. (2022) illustrate the benefit of including antecedent climate indices as predictors in reservoir inflow forecasts. ...
Streamflow forecasts play an important role in water resources operation and management and skillful seasonal forecasts can significantly facilitate the decision-making process. Streamflow variability is often associated with various large-scale, slowly evolving climate phenomena (e.g. ENSO), promoting the value of climate indices for streamflow forecast development, and has been investigated extensively. Separately, global climate models (GCMs), which provide climate forecasts out to 12-months globally, have been demonstrated to enhance seasonal streamflow predictability. However, neither the combination nor the interaction of these two sources of predictability has been given much attention. In this work, we propose a framework which can simultaneously account for antecedent climate indices and GCM forecasts in statistical streamflow forecasts and address how their interactions affect predictability. More specifically, we build a streamflow forecast framework combining statistical forecast models conditioned on climate indices with dynamical North American Multi-Model Ensemble (NMME) precipitation forecasts to generate streamflow forecast ensembles, and merge ensemble members with a BMA-bagging (Bayesian Model Averaging with bootstrap aggregating) approach. The framework is applied to streamflow on the Blue Nile River upstream of the Grand Ethiopian Renaissance Dam (GERD). Most NMME models tend to improve one-month-ahead GERD inflow forecasts however longer lead times prove challenging, and require specific NMME model selection. In contrast, although the value of climate indices varies with forecast lead time, their potential contribution grows at multi-month leads. The GERD inflow forecasts including NMME models and climate indices simultaneously can take advantage of both sources of predictability and prove superior across lead times as compared to including either source individually.
... In an attempt to further increase these models' efficiency, deep learning neural network models have been applied to reservoir inflow forecasting with promising results [6][7][8][9][10][11][12] Recurrent Neural Networks (RNN), Gated Recurrent Unit (GRU), and Long-Short Tem Memory (LSTM) models have been used for reservoir inflow forecasting, with LSTM proving to be the most effective [13]. A hybrid framework using machine learning for reservoir inflow forecast has been proposed by Tian et al. [14] with interesting results as an outcome as compared to classical methods. Luo et al. [15] proposed an ensemble model combining the Deep believe network (DBN) and LSTM with the ensemble result better than that of the individual models. ...
Accurate reservoir inflow forecasting is crucial for efficient water management. In this study, different deep learning models, including Dense, Long short-term memory (LSTM), and one-dimensional convolutional neural networks (Conv1D), were used to build ensembles. Seasonal-trend decomposition using loess (STL) was applied to decompose reservoir inflows and precipitations into random, seasonal, and trend components. Seven ensemble models, namely STL-Dense, STL-Conv1D, STL-LSTM, STL-Dense-LSTM-Conv1D, STL-Dense multivariate, STL-LSTM multivariate, and STL-Conv1D multivariate, were proposed and evaluated using daily inflows and precipitation decomposed data from the Lom Pangar reservoir from 2015 to 2020. Evaluation metrics, such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Nash Sutcliff Efficiency (NSE), were applied to assess model performance. Results showed that the STL-Dense multivariate model was the best ensemble among the thirteen models with MAE of 14.636 m3/s, RMSE of 20.841 m3/s, MAPE of 6.622%, and NSE of 0.988. These findings stress the importance of considering multiple inputs and models for accurate reservoir inflow forecasting and optimal water management. Not all ensemble models were good for Lom pangar inflow forecast as the Dense, Conv1D, and LSTM models performed better than their proposed STL monovariate ensemble models.
... Ravuri et al., 2021;Neri et al., 2019), and geographical domains (from point to street level; from a single river catchment through to global approaches). Hybrid models have been applied to predict a variety of hydrometeorological variables, including extreme heat and precipitation (Najafi et al., 2021;Miao et al., 2019;Ma et al., 2022), seasonal climate variables (Golian et al., 2022;Baker et al., 2020), tropical cyclones/hurricanes (Vecchi et al., 2011;Murakami et al., 2016;Kang and Elsner, 2020;Klotzbach et al., 2020), streamflow (Wood and Schaake, 2008;Mendoza et al., 2017;Rasouli et al., 2012;Duan et al., 2020), flooding (Slater and Villarini, 2018), drought (Madadgar et al., 2016;Wu et al., 2022), sea level (Khouakhi et al., 2019), and reservoir levels (Tian et al., 2022), over a range of timescales (Table 2). Certain other examples discussed in this review are not fully hybrid (e.g. ...
... DelSole and Shukla, 2009). Hybrid hydroclimatic forecasts and predictions have numerous operational and strategic applications, including water resources planning, reservoir inflow management (Tian et al., 2022;Essenfelder et al., 2020), surface water flooding (Rözer et al., 2021), flood risk mitigation, navigation (Meißner et al., 2017), and agricultural crop forecasting (Cao et al., 2022;Slater et al., 2022). ...
... As multiple predictor variables can be included within a statistical or ML model, it is feasible to combine predictors that have very different time-varying impacts, such as reservoir management decisions or initial hydrological conditions impacting the short term, versus annual-to-multi-decadal climate oscillations for longer-term predictability. For instance, Tian et al. (2022) present a reservoir inflow forecasting framework combining a suite of different ML models (including gradient-boosting machine, random forests, and elastic net) with climate model outputs from the FLOR model for reservoirs in the Upper Colorado River basin. They also included soil moisture and evaporation to represent antecedent conditions, which significantly improved the forecasts of reservoir inflow. ...
Hybrid hydroclimatic forecasting systems employ data-driven (statistical or machine learning) methods to harness and integrate a broad variety of predictions from dynamical, physics-based models - such as numerical weather prediction, climate, land, hydrology, and Earth system models - into a final prediction product. They are recognized as a promising way of enhancing the prediction skill of meteorological and hydroclimatic variables and events, including rainfall, temperature, streamflow, floods, droughts, tropical cyclones, or atmospheric rivers. Hybrid forecasting methods are now receiving growing attention due to advances in weather and climate prediction systems at subseasonal to decadal scales, a better appreciation of the strengths of AI, and expanding access to computational resources and methods. Such systems are attractive because they may avoid the need to run a computationally expensive offline land model, can minimize the effect of biases that exist within dynamical outputs, benefit from the strengths of machine learning, and can learn from large datasets, while combining different sources of predictability with varying time horizons. Here we review recent developments in hybrid hydroclimatic forecasting and outline key challenges and opportunities for further research. These include obtaining physically explainable results, assimilating human influences from novel data sources, integrating new ensemble techniques to improve predictive skill, creating seamless prediction schemes that merge short to long lead times, incorporating initial land surface and ocean/ice conditions, acknowledging spatial variability in landscape and atmospheric forcing, and increasing the operational uptake of hybrid prediction schemes.
... Under the NMME protocol, forecasts from different GCMs are operationally provided at unified horizontal resolution (1 • ×1 • ), delivery date (eighth of each month) and lead time (at least nine months) (Becker et al., 2022). Furthermore, the NMME forecasts have been widely incorporated into decision support systems for environmental planning, flow maintenance and hazards management (Koppa et al., 2019;Arsenault et al., 2020;Muñoz et al., 2020;Tian et al., 2022). ...
