Statistical downscaling of GCM simulations to streamflow using relevance vector machine

General circulation models (GCMs), the climate models often used in assessing the impact of climate change, operate on a coarse scale and thus the simulation results obtained from GCMs are not particularly useful in a comparatively smaller river basin scale hydrology. The article presents a methodology of statistical downscaling based on sparse Bayesian learning and Relevance Vector Machine (RVM) to model streamflow at river basin scale for monsoon period (June, July, August, September) using GCM simulated climatic variables. NCEP/NCAR reanalysis data have been used for training the model to establish a statistical relationship between streamflow and climatic variables. The relationship thus obtained is used to project the future streamflow from GCM simulations. The statistical methodology involves principal component analysis, fuzzy clustering and RVM. Different kernel functions are used for comparison purpose. The model is applied to Mahanadi river basin in India. The results obtained using RVM are compared with those of state-of-the-art Support Vector Machine (SVM) to present the advantages of RVMs over SVMs. A decreasing trend is observed for monsoon streamflow of Mahanadi due to high surface warming in future, with the CCSR/NIES GCM and B2 scenario.     

Regional impacts of climate change on irrigation water demands

This paper presents an approach to model the expected impacts of climate change on irrigation water demand in a reservoir command area. A statistical downscaling model and an evapotranspiration model are used with a general circulation model (GCM) output to predict the anticipated change in the monthly irrigation water requirement of a crop. Specifically, we quantify the likely changes in irrigation water demands at a location in the command area, as a response to the projected changes in precipitation and evapotranspiration at that location. Statistical downscaling with a canonical correlation analysis is carried out to develop the future scenarios of meteorological variables (rainfall, relative humidity (RH), wind speed (U₂), radiation, maximum (Tmax) and minimum (Tmin) temperatures) starting with simulations provided by a GCM for a specified emission scenario. The medium resolution Model for Interdisciplinary Research on Climate GCM is used with the A1B scenario, to assess the likely changes in irrigation demands for paddy, sugarcane, permanent garden and semidry crops over the command area of Bhadra reservoir, India. Results from the downscaling model suggest that the monthly rainfall is likely to increase in the reservoir command area. RH, Tmax and Tmin are also projected to increase with small changes in U₂. Consequently, the reference evapotranspiration, modeled by the Penman–Monteith equation, is predicted to increase. The irrigation requirements are assessed on monthly scale at nine selected locations encompassing the Bhadra reservoir command area. The irrigation requirements are projected to increase, in most cases, suggesting that the effect of projected increase in rainfall on the irrigation demands is offset by the effect due to projected increase/change in other meteorological variables (viz., Tmax and Tmin, solar radiation, RH and U₂). The irrigation demand assessment study carried out at a river basin will be useful for future irrigation management systems. Copyright © 2012 John Wiley & Sons, Ltd.     

Downscaling of precipitation for climate change scenarios: A support vector machine approach

The Climate impact studies in hydrology often rely on climate change information at fine spatial resolution. However, general circulation models (GCMs), which are among the most advanced tools for estimating future climate change scenarios, operate on a coarse scale. Therefore the output from a GCM has to be downscaled to obtain the information relevant to hydrologic studies. In this paper, a support vector machine (SVM) approach is proposed for statistical downscaling of precipitation at monthly time scale. The effectiveness of this approach is illustrated through its application to meteorological sub-divisions (MSDs) in India. First, climate variables affecting spatio-temporal variation of precipitation at each MSD in India are identified. Following this, the data pertaining to the identified climate variables (predictors) at each MSD are classified using cluster analysis to form two groups, representing wet and dry seasons. For each MSD, SVM- based downscaling model (DM) is developed for season(s) with significant rainfall using principal components extracted from the predictors as input and the contemporaneous precipitation observed at the MSD as an output. The proposed DM is shown to be superior to conventional downscaling using multi-layer back-propagation artificial neural networks. Subsequently, the SVM-based DM is applied to future climate predictions from the second generation Coupled Global Climate Model (CGCM2) to obtain future projections of precipitation for the MSDs. The results are then analyzed to assess the impact of climate change on precipitation over India. It is shown that SVMs provide a promising alternative to conventional artificial neural networks for statistical downscaling, and are suitable for conducting climate impact studies.     

Univariate streamflow forecasting using commonly used data-driven models: literature review and case study

Eight data-driven models and five data pre-processing methods were summarized; the multiple linear regression (MLR), artificial neural network (ANN) and wavelet decomposition (WD) models were then used in short-term streamflow forecasting at four stations in the East River basin, China. The wavelet–artificial neural network (W-ANN) method was used to predict 1-month-ahead monthly streamflow at Longchuan station (LS). The results indicate better performance of MLR and wavelet–multiple linear regression (W-MLR) in analysing the stationary trained dataset. Four models showed similar performance in 1-day-ahead streamflow forecasting, while W-MLR and W-ANN performed better in 5-day-ahead forecasting. Three reservoirs were shown to have more influence on downstream than upstream streamflow and models had the worst performance at Boluo station. Furthermore, the W-ANN model performed well for 1-month-ahead streamflow forecasting at LS with consideration of a deterministic component.     

Prediction of daily rainfall state in a river basin using statistical downscaling from GCM output

Conventional statistical downscaling techniques for prediction of multi-site rainfall in a river basin fail to capture the correlation between multiple sites and thus are inadequate to model the variability of rainfall. The present study addresses this problem through representation of the pattern of multi-site rainfall using rainfall state in a river basin. A model based on K-means clustering technique coupled with a supervised data classification technique, namely Classification And Regression Tree (CART), is used for generation of rainfall states from large-scale atmospheric variables in a river basin. The K-means clustering is used to derive the daily rainfall state from the historical daily multi-site rainfall data. The optimum number of clusters in the observed rainfall data is obtained after application of various cluster validity measures to the clustered data. The CART model is then trained to establish relationship between the daily rainfall state of the river basin and the standardized, dimensionally-reduced National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) reanalysis climatic data set. The relationship thus developed is applied to the General Circulation Model (GCM)-simulated, standardized, bias free large-scale climate variables for prediction of rainfall states in future. Comparisons of the number of days falling under different rainfall states for the observed period and the future give the change expected in the river basin due to global warming. The methodology is tested for the Mahanadi river basin in India.     

Assessment and evaluation of potential climate change impact on monsoon flows using machine learning technique over Wainganga River basin,India

In this study, classification- and regression-based statistical downscaling is used to project the monthly monsoon streamflow over the Wainganga basin, India, using 40 global climate model (GCM) outputs and four representative concentration pathways (RCP) scenarios. Support vector machine (SVM) and relevance vector machine (RVM) are considered to perform downscaling. The RVM outperforms SVM and is used to simulate future projections of monsoon flows for different periods. In addition, variability in water availability with uncertainty and change point (CP) detection are accomplished by flow–duration curve and Bayesian analysis, respectively. It is observed from the results that the upper extremes of monsoon flows are highly sensitive to increases in temperature and show a continuous decreasing trend. Medium and low flows are increasing in future projections for all the scenarios, and high uncertainty is noticed in the case of low flows. An early CP is detected in the case of high emissions scenarios.     

Assessing regression‐based statistical approaches for downscaling precipitation over North America

This paper assesses linear regression‐based methods in downscaling daily precipitation from the general circulation model (GCM) scale to a regional climate model (RCM) scale (45‐ and 15‐km grids) and down to a station scale across North America. Traditional downscaling experiments (linking reanalysis/dynamical model predictors to station precipitation) as well as nontraditional experiments such as predicting dynamic model precipitation from larger‐scale dynamic model predictors or downscaling dynamic model precipitation from predictors at the same scale are conducted. The latter experiments were performed to address predictability limit and scale issues. The results showed that the downscaling of daily precipitation occurrence was rarely successful at all scales, although results did constantly improve with the increased resolution of climate models. The explained variances for downscaled precipitation amounts at the station scales were low, and they became progressively better when using predictors from a higher‐resolution climate model, thus showing a clear advantage in using predictors from RCMs driven by reanalysis at its boundaries, instead of directly using reanalysis data. The low percentage of explained variances resulted in considerable underestimation of daily precipitation mean and standard deviation. Although downscaling GCM precipitation from GCM predictors (or RCM precipitation from RCM predictors) cannot really be considered downscaling, as there is no change in scale, the exercise yields interesting information as to the limit in predictive ability at the station scale. This was especially clear at the GCM scale, where the inability of downscaling GCM precipitation from GCM predictors demonstrates that GCM precipitation‐generating processes are largely at the subgrid scale (especially so for convective events), thus indicating that downscaling precipitation at the station scale from GCM scale is unlikely to be successful. Although results became better at the RCM scale, the results indicate that, overall, regression‐based approaches did not perform well in downscaling precipitation over North America. Copyright © 2013 John Wiley & Sons, Ltd.     

Comparison of transfer functions in statistical downscaling models for daily temperature and precipitation over Canada

This study compares three linear models and one non-linear model, specifically multiple linear regression (MLR) with ordinary least squares (OLS) estimates, robust regression, ridge regression, and artificial neural networks (ANNs), to identify an appropriate transfer function in statistical downscaling (SD) models for the daily maximum and minimum temperatures (T_max and T_min) and daily precipitation occurrence and amounts (P_occ and P_amount). This comparison was made over twenty-five observation sites located in five different Canadian provinces (British Columbia, Saskatchewan, Manitoba, Ontario, and Québec). Reanalysis data were employed as atmospheric predictor variables of SD models. Predictors of linear transfer functions and ANN were selected by linear correlations coefficient and mutual information, respectively. For each downscaled case, annual and monthly models were developed and analysed. The monthly MLR, annual ANN, annual ANN, and annual MLR yielded the best performance for T_max, T_min, P_occ and P_amont according to the modified Akaike information criterion (AICu). A monthly MLR is recommended for the transfer functions of the four predictands because it can provide a better performance for the T_max and as good performance as the annual MLR for the T_min, P_occ, and P_amount. Furthermore, a monthly MLR can provide a slightly better performance than an annual MLR for extreme events. An annual MLR approach is also equivalently recommended for the transfer functions of the four predictands because it showed as good a performance as monthly MLR in spite of its mathematical simplicity. Robust and ridge regressions are not recommended because the data used in this study are not greatly affected by outlier data and multicollinearity problems. An annual ANN is recommended only for the T_min, based on the best performance among the models in terms of both the RMSE and AICu.     

A comparison of three methods for downscaling daily precipitation in the Punjab region

Many downscaling techniques have been developed in the past few years for projection of station‐scale hydrological variables from large‐scale atmospheric variables simulated by general circulation models (GCMs) to assess the hydrological impacts of climate change. This article compares the performances of three downscaling methods, viz. conditional random field (CRF), K‐nearest neighbour (KNN) and support vector machine (SVM) methods in downscaling precipitation in the Punjab region of India, belonging to the monsoon regime. The CRF model is a recently developed method for downscaling hydrological variables in a probabilistic framework, while the SVM model is a popular machine learning tool useful in terms of its ability to generalize and capture nonlinear relationships between predictors and predictand. The KNN model is an analogue‐type method that queries days similar to a given feature vector from the training data and classifies future days by random sampling from a weighted set of K closest training examples. The models are applied for downscaling monsoon (June to September) daily precipitation at six locations in Punjab. Model performances with respect to reproduction of various statistics such as dry and wet spell length distributions, daily rainfall distribution, and intersite correlations are examined. It is found that the CRF and KNN models perform slightly better than the SVM model in reproducing most daily rainfall statistics. These models are then used to project future precipitation at the six locations. Output from the Canadian global climate model (CGCM3) GCM for three scenarios, viz. A1B, A2, and B1 is used for projection of future precipitation. The projections show a change in probability density functions of daily rainfall amount and changes in the wet and dry spell distributions of daily precipitation. Copyright © 2011 John Wiley & Sons, Ltd.     

Estimation of future precipitation change in the Yangtze River basin by using statistical downscaling method

In this study, the applicability of the statistical downscaling model (SDSM) in downscaling precipitation in the Yangtze River basin, China was investigated. The investigation includes the calibration of the SDSM model by using large-scale atmospheric variables encompassing NCEP/NCAR reanalysis data, the validation of the model using independent period of the NCEP/NCAR reanalysis data and the general circulation model (GCM) outputs of scenarios A2 and B2 of the HadCM3 model, and the prediction of the future regional precipitation scenarios. Selected as climate variables for downscaling were measured daily precipitation data (1961–2000) from 136 weather stations in the Yangtze River basin. The results showed that: (1) there existed good relationship between the observed and simulated precipitation during the calibration period of 1961–1990 as well as the validation period of 1991–2000. And the results of simulated monthly and seasonal precipitation were better than that of daily. The average R ² values between the simulated and observed monthly and seasonal precipitation for the validation period were 0.78 and 0.91 respectively for the whole basin, which showed that the SDSM had a good applicability on simulating precipitation in the Yangtze River basin. (2) Under both scenarios A2 and B2, during the prediction period of 2010–2099, the change of annual mean precipitation in the Yangtze River basin would present a trend of deficit precipitation in 2020s; insignificant changes in the 2050s; and a surplus of precipitation in the 2080s as compared to the mean values of the base period. The annual mean precipitation would increase by about 15.29% under scenario A2 and increase by about 5.33% under scenario B2 in the 2080s. The winter and autumn might be the more distinct seasons with more predicted changes of precipitation than in other seasons. And (3) there would be distinctive spatial distribution differences for the change of annual mean precipitation in the river basin, but the most of Yangtze River basin would be dominated by the increasing trend.     

Estimation of potential evapotranspiration over the Yellow River basin: reference crop evaporation or Shuttleworth–Wallace?

Potential evapotranspiration (PET) is a key input to hydrological models. Its estimation has often been via the Penman–Monteith (P–M) equation, most recently in the form of an estimate of reference evapotranspiration (RET) as recommended by FAO‐56. In this paper the Shuttleworth–Wallace (S–W) model is implemented to estimate PET directly in a form that recognizes vegetation diversity and temporal change without reference to experimental measurements and without calibration. The threshold values of vegetation parameters are drawn from the literature based on the International Geosphere–Biosphere Programme land cover classification. The spatial and temporal variation of the LAI of vegetation is derived from the composite NOAA‐AVHRR normalized difference vegetation index (NDVI) using a method based on the SiB2 model, and the Climate Research Unit database is used to provide the required meteorological data. All these data inputs are publicly and globally available. Consequently, the implementation of the S–W model developed in this study is applicable at the global scale, an essential requirement if it is to be applied in data‐poor or ungauged large basins. A comparison is made between the FAO‐56 method and the S–W model when applied to the Yellow River basin for the whole of the last century. The resulting estimates of RET and PET and their association with vegetation types and leaf area index (LAI) are examined over the whole basin both annual and monthly and at six specific points. The effect of NDVI on the PET estimate is further evaluated by replacing the monthly NDVI product with the 10‐day product. Multiple regression relationships between monthly PET, RET, LAI, and climatic variables are explored for categories of vegetation types. The estimated RET is a good climatic index that adequately reflects the temporal change and spatial distribution of climate over the basin, but the PET estimated using the S–W model not only reflects the changes in climate, but also the vegetation distribution and the development of vegetation in response to climate. Although good statistical relationships can be established between PET, RET and/or climatic variables, applying these relationships likely will result in large errors because of the strong non‐linearity and scatter between the PET and the LAI of vegetation. It is concluded that use of the implementation of the S–W model described in this study results in a physically sound estimate of PET that accounts for changing land surface conditions. Copyright © 2006 John Wiley & Sons, Ltd.     

Extraction of information content from stochastic disaggregation and bias corrected downscaled precipitation variables for crop simulation

We applied a simple statistical downscaling procedure for transforming daily global climate model (GCM) rainfall to the scale of an agricultural experimental station in Katumani, Kenya. The transformation made was two-fold. First, we corrected the rainfall frequency bias of the climate model by truncating its daily rainfall cumulative distribution into the station’s distribution based on a prescribed observed wet-day threshold. Then, we corrected the climate model rainfall intensity bias by mapping its truncated rainfall distribution into the station’s truncated distribution. Further improvements were made to the bias corrected GCM rainfall by linking it with a stochastic disaggregation scheme to correct the time structure problem inherent with daily GCM rainfall. Results of the simple and hybridized GCM downscaled precipitation variables (total, probability of occurrence, intensity and dry spell length) were linked with a crop model for a more objective evaluation of their performance using a non-linear measure based on mutual information based on entropy. This study is useful for the identification of both suitable downscaling technique as well as the effective precipitation variables for forecasting crop yields using GCM’s outputs which can be useful for addressing food security problems beforehand in critical basins around the world.     

Impact of the spatial variability of daily precipitation on hydrological projections: A comparison of GCM‐ and RCM‐driven cases in the Han River basin,Korea

In this study, we investigate the impact of the spatial variability of daily precipitation on hydrological projections based on a comparative assessment of streamflow simulations driven by a global climate model (GCM) and two regional climate models (RCMs). A total of 12 different climate input datasets, that is, the raw and bias‐corrected GCM and raw and bias‐corrected two RCMs for the reference and future periods, are fed to a semidistributed hydrological model to assess whether the bias correction using quantile mapping and dynamical downscaling using RCMs can improve streamflow simulation in the Han River basin, Korea. A statistical analysis of the daily precipitation demonstrates that the precipitation simulated by the GCM fails to capture the large variability of the observed daily precipitation, in which the spatial autocorrelation decreases sharply within a relatively short distance. However, the spatial variability of precipitation simulated by the two RCMs shows better agreement with the observations. After applying bias correction to the raw GCM and raw RCMs outputs, only a slight change is observed in the spatial variability, whereas an improvement is observed in the precipitation intensity. Intensified precipitation but with the same spatial variability of the raw output from the bias‐corrected GCM does not improve the heterogeneous runoff distributions, which in turn regulate unrealistically high peak downstream streamflow. GCM‐simulated precipitation with a large bias correction that is necessary to compensate for the poor performance in present climate simulation appears to distort streamflow patterns in the future projection, which leads to misleading projections of climate change impacts on hydrological extremes.     

Comparison of SDSM and LARS-WG for simulation and downscaling of extreme precipitation events in a watershed

Future climate projections of Global Climate Models (GCMs) under different emission scenarios are usually used for developing climate change mitigation and adaptation strategies. However, the existing GCMs have only limited ability to simulate the complex and local climate features, such as precipitation. Furthermore, the outputs provided by GCMs are too coarse to be useful in hydrologic impact assessment models, as these models require information at much finer scales. Therefore, downscaling of GCM outputs is usually employed to provide fine-resolution information required for impact models. Among the downscaling techniques based on statistical principles, multiple regression and weather generator are considered to be more popular, as they are computationally less demanding than the other downscaling techniques. In the present study, the performances of a multiple regression model (called SDSM) and a weather generator (called LARS-WG) are evaluated in terms of their ability to simulate the frequency of extreme precipitation events of current climate and downscaling of future extreme events. Areal average daily precipitation data of the Clutha watershed located in South Island, New Zealand, are used as baseline data in the analysis. Precipitation frequency analysis is performed by fitting the Generalized Extreme Value (GEV) distribution to the observed, the SDSM simulated/downscaled, and the LARS-WG simulated/downscaled annual maximum (AM) series. The computations are performed for five return periods: 10-, 20-, 40-, 50- and 100-year. The present results illustrate that both models have similar and good ability to simulate the extreme precipitation events and, thus, can be adopted with confidence for climate change impact studies of this nature.     

Evaluation of linear regression methods as downscaling tools in temperature projections over the Pichola Lake Basin in India

In this paper, downscaling models are developed using various linear regression approaches, namely direct, forward, backward and stepwise regression, for obtaining projections of mean monthly maximum and minimum temperatures (Tmax and Tmin) to lake‐basin scale in an arid region in India. The effectiveness of these regression approaches is evaluated through application to downscale the predictands for the Pichola lake region in the state of Rajasthan in India, which is considered to be a climatically sensitive region. The predictor variables are extracted from (i) the National Centers for Environmental Prediction (NCEP) reanalysis dataset for the period 1948–2000 and (ii) the simulations from the third‐generation Canadian Coupled Global Climate Model (CGCM3) for emission scenarios A1B, A2, B1 and COMMIT for the period 2001–2100. The selection of important predictor variables becomes a crucial issue for developing downscaling models as reanalysis data are based on a wide range of meteorological measurements and observations. A simple multiplicative shift was used for correcting predictand values. Direct regression was found to yield better performance among all other regression techniques for the training data set, while the forward regression technique performed better in the validation data set, explored in the present study. For trend analysis, the Mann–Kendall non‐parametric test was performed. The results of downscaling models show that an increasing trend is observed for Tmax and Tmin for A1B, A2 and B1 scenarios, whereas no trend is discerned with the COMMIT scenario by using predictors. Copyright © 2010 John Wiley & Sons, Ltd.     

A spatial temporal downscaling approach to development of IDF relations for Perth airport region in the context of climate change

ABSTRACT

Downscaling of climate projections is the most adapted method to assess the impacts of climate change at regional and local scales. This study utilized both spatial and temporal downscaling approaches to develop intensity–duration–frequency (IDF) relations for sub-daily rainfall extremes in the Perth airport area. A multiple regression-based statistical downscaling model tool was used for spatial downscaling of daily rainfall using general circulation models (GCMs) (Hadley Centre’s GCM and Canadian Global Climate Model) climate variables. A simple scaling regime was identified for 30 minutes to 24 hours duration of observed annual maximum (AM) rainfall. Then, statistical properties of sub-daily AM rainfall were estimated by scaling an invariant model based on the generalized extreme value distribution. RMSE, Nash-Sutcliffe efficiency coefficient and percentage bias values were estimated to check the accuracy of downscaled sub-daily rainfall. This proved the capability of the proposed approach in developing a linkage between large-scale GCM daily variables and extreme sub-daily rainfall events at a given location. Finally IDF curves were developed for future periods, which show similar extreme rainfall decreasing trends for the 2020s, 2050s and 2080s for both GCMs.

Editor M.C. Acreman; Associate editor S. Kanae     

Dynamical downscaling of regional climate: A review of methods and limitations

The traditional dynamical downscaling (TDD) method employs continuous integration of regional climate models (RCM) with the general circulation model (GCM) providing the initial and lateral boundary conditions. Dynamical downscaling simulations are constrained by physical principles and can generate a full set of climate information, providing one of the important approaches to projecting fine spatial-scale future climate information. However, the systematic biases of climate models often degrade the TDD simulations and hinder the application of dynamical downscaling in the climate-change related studies. New methods developed over past decades improve the performance of dynamical downscaling simulations. These methods can be divided into four groups: the TDD method, the pseudo global warming method, dynamical downscaling with GCM bias corrections, and dynamical downscaling with both GCM and RCM bias corrections. These dynamical downscaling methods are reviewed and compared in this paper. The merits and limitations of each dynamical downscaling method are also discussed. In addition, the challenges and potential directions in progressing dynamical downscaling methods are stated.     

Wet and dry spell analysis of Global Climate Model-generated precipitation using power laws and wavelet transforms

Climate model simulations for the twenty-first century point toward changing characteristics of precipitation. This paper investigates the impact of climate change on precipitation in the Kansabati River basin in India. A downscaling method, based on Bayesian Neural Network (BNN), is applied to project precipitation generated from six Global Climate Models (GCMs) using two scenarios (A2 and B2). Wet and dry spell properties of monthly precipitation series at five meteorologic stations in the Kansabati basin are examined by plotting successive wet and dry durations (in months) against their number of occurrences on a double-logarithmic paper. Straight-line relationships on such graphs show that power laws govern the pattern of successive persistent wet and dry monthly spells. Comparison of power-law behaviors provides useful interpretation about the temporal precipitation pattern. The impact of low-frequency precipitation variability on the characteristics of wet and dry spells is also evaluated using continuous wavelet transforms. It is found that inter-annual cycles play an important role in the formation of wet and dry spells.     

Monthly total sediment forecasting using adaptive neuro fuzzy inference system

Accurate forecasting of sediment is an important issue for reservoir design and water pollution control in rivers and reservoirs. In this study, an adaptive neuro-fuzzy inference system (ANFIS) approach is used to construct monthly sediment forecasting system. To illustrate the applicability of ANFIS method the Great Menderes basin is chosen as the study area. The models with various input structures are constructed for the purpose of identification of the best structure. The performance of the ANFIS models in training and testing sets are compared with the observed data. To get more accurate evaluation of the results ANFIS models, the best fit model structures are also tested by artificial neural networks (ANN) and multiple linear regression (MLR) methods. The results of three methods are compared, and it is observed that the ANFIS is preferable and can be applied successfully because it provides high accuracy and reliability for forecasting of monthly total sediment.     

基于改进U-Net的CMIP5全球气候模式降尺度方法及其在鄱阳湖流域的应用

利用降尺度方法对CMIP5全球气候模式进行空间降尺度并以此研究鄱阳湖流域未来气候时空变化趋势,能够为流域生态环境保护提供数据、技术和理论上的支持.通过简化原始网络结构,在网络首部添加插值层,采用反卷积算法作为上采样算法对传统U-Net网络进行改进,建立基于深度学习的气候模式空间降尺度模型(DLDM).以1965-200...