Calibrated SWAT Model Research Articles

SWAT Model Calibration and Validation with Remotely Sensed Evapotranspiration data in Maklang-Tuyungbi-Taret lok Ungauged Basin in Manipur, India

SWAT Model Calibration and Validation with Remotely Sensed Evapotranspiration data in Maklang-Tuyungbi-Taret lok Ungauged Basin in Manipur, India

Simulating the land use change effects on non-point source pollution in the Duliujian River Basin.

The uncertainty in the generation and formation of non-point source pollution makes it challenging to monitor and control this type of pollution. The SWAT model is frequently used to simulate non-point source pollution in watersheds and is mainly applied to natural watersheds that are less affected by human activities. This study focuses on the Duliujian River Basin (Xiqing section), which is characterized by a dense population and rapid urbanization. Based on the calibrated SWAT model, this study analyzed the effects of land use change on non-point source pollution both temporally and spatially. It was found that nitrogen and phosphorus non-point source pollution load losses were closely related to land use type, with agricultural land and high-density urban land (including rural settlements) being the main contributors to riverine nitrogen and phosphorus pollution. This indicates the necessity of analyzing the impact of land use changes on non-point source pollution loads by identifying critical source areas and altering the land use types that contribute heavily to pollution in these areas. The simulation results of land use type changes in these critical source areas showed that the reduction effect on non-point source pollution load is in the order of forest land > grassland > low-density residential area. To effectively curb surface source pollution in the study area, strategies such as modifying urban land use types, increasing vegetation cover and ground infiltration rate, and strictly controlling the discharge of domestic waste and sewage from urban areas can be implemented.

A new interpretable streamflow prediction approach based on SWAT-BiLSTM and SHAP.

Streamflow is a crucial variable for assessing the available water resources for both human and environmental use. Accurate streamflow prediction plays a significant role in water resource management and assessing the impacts of climate change. This study explores the potential of coupling conceptual hydrological models based on physical processes with machine learning algorithms to enhance the performance of streamflow simulations. Four coupled models, namely SWAT-Transformer, SWAT-LSTM, SWAT-GRU, and SWAT-BiLSTM, were constructed in this research. SWAT served as a transfer function to convert four meteorological features, including precipitation, temperature, relative humidity, and wind speed, into six hydrological features: soil water content, lateral flow, percolation, groundwater discharge, surface runoff, and evapotranspiration. Machine learning algorithms were employed to capture the underlying relationships between these ten feature variables and the target variable (streamflow) to predict daily streamflow in the Sandu-River Basin (SRB). Among the four coupled models and the calibrated SWAT model, SWAT-BiLSTM exhibited the best streamflow simulation performance. During the calibration period (training period), it achieved R2 and NSE values of 0.92 and 0.91, respectively, and maintained them at 0.90 during the validation period (testing period). Additionally, the performance of all four coupled models surpassed that of the calibrated SWAT model. Compared to the tendency of the SWAT model to underestimate streamflow, the absolute values of PBIAS for all coupled models are below 10%, which indicates that there is no significant systematic bias evident. SHapley Additive exPlanations (SHAP) were used to analyze the impact of different feature variables on streamflow prediction. The results indicated that precipitation contributed the most to streamflow prediction, with a global importance of 29.7%. Hydrological feature variable output by the SWAT model played a dominant role in the Bi-LSTM's prediction process. Coupling conceptual hydrological models with machine learning algorithms can significantly enhance the predictive performance of streamflow. The application of SHAP improves the interpretability of the coupled models and enhances researchers' confidence in the prediction results.

Quantifying Uncertainty in Hydrological Drought Index Using Calibrated SWAT Model

The hydrological model requires precise parameter calibration to reflect the basin characteristicsproperly. The optimum parameter values depend highly on the runoff characteristics; thus, the simulated runoff data affects the uncertainty in hydrological drought assessment using hydrological models. This study quantified the hydrological drought uncertainty based on all available calibrated parameters for 12 different observation periods. The Soil and water assessment Tool was used for hydrological modelling, and the uncertainty was analyzed using Bayesian model averaging (BMA). The hydrological drought index using the simulated runoff was analyzed to vary significantly with the optimized hydrological parameter cases. The drought index calculation using the estimated runoff showed that the number of droughts occurred frequently as the duration increased, But, it was observed that severe droughts of 21-month duration occur most frequently. The uncertainty analysis conducted with BMA revealed that for durations shorter than 18 months, the uncertainties in drought index calculations were higher than those in runoff simulations. As the duration increased, uncertainties decreased, with values of 57%, 54%, 50%, 23%, 23%, 19%, 16% and 16% for 3, 6, 9, 12, 15, 18, 21 and 24-month durations, respectively. The average uncertainty was larger for runoff estimation than for drought calculation. Still, the uncertainty variation by each PC was larger for drought calculation, ranging from 0 to 94% as the duration increased to 24 months. Therefore,the results of this study confirm that hydrological drought analysis using hydrological models contains uncertainties depending on the hydrological model parameter cases.

Impacts of irrigation development on water quality in the San Salvador watershed (Part 2): Implementation of scenarios in SWAT

Intensive agricultural activities pose a significant threat to water quality as critical non-point sources of pollution. Effective mitigation strategies demand understanding the causes and processes of water pollution. This study aimed to quantify the impacts of irrigation development on water quality and assess best management practices for sustainable agriculture intensification. Employing the calibrated SWAT model for the San Salvador watershed (baseline scenario), two scenarios were implemented and evaluated: the first one depicted irrigation development from a future reservoir, and the second integrated riparian buffer zones to minimize nutrient and sediment losses. Notably the baseline scenario did not achieve nutrient water quality objectives. Results revealed that irrigation development increases nutrient yields, driving the future reservoir toward eutrophication. Implementing riparian buffer zones reduced nutrient loss, but additional measures are necessary for sustainable environmental goals at the basin scale. This research contributes with valuable insights for formulating effective management strategies to minimize nutrient pollution in water and safeguard water quality and biodiversity in the basin.

Assessment of Irrigation Potential in Jewuha Watershed, Middle Awash River Basin, Ethiopia

Ethiopia is endowed with immense water and land resources that could be tapped and used for irrigation development. However, little has been done so far due mainly to lack of comprehensive knowledge on the potential of these resources. This study was, therefore, taken up to assess the irrigation potential in data scarce Jewuha watershed of the Awash River Basin. To achieve the objective of the study, the suitability of the land for both surface and pressurized irrigations was estimated using a parametric evaluation technique combined with the FAO land evaluation framework in the GIS environment with multi-criteria analysis of irrigation suitability factors. The irrigation water demands of major crops adopted in the area were determined using CROPWAT software. The surface water potential at the sub watershed was estimated using a combination of SWAT model and the spatial proximity regionalization technique. The performance of the SWAT model was checked by statistical parameters. The irrigation suitability analysis reveals that 51372ha of the study area are suitable and 16274ha are unsuitable for surface irrigation system. Furthermore, 52768ha of the area are suitable and 15088ha are unsuitable for a sprinkler irrigation system. For drip irrigation, 52751ha are suitable and 14855ha are unsuitable. The calibration and validation of SWAT model showed that the model has performed well to simulate the hydrology of the watershed with a coefficient of determination (R2), Nash-Sutcliffe efficiency (NSE) and Percent of bias (PBIAS) of 0.74, 0.73 and 0.80 for calibration and 0.71, 0.70 and 7.90 for validation, respectively. From a combined analysis of the available land and water resources, the gross irrigation potential of the area is estimated to be 12,997ha, of which 3098ha of land is exclusively suitable for surface irrigation and 9,899ha is suitable for pressurized irrigation systems.

Calibration of SWAT model in the Pernambuco state watersheds to support the SUPer system

O SUPer (Sistema de Unidades de Resposta Hidrológica do Estado de Pernambuco, Brasil) foi desenvolvido através de uma grande parceria entre universidades e órgãos públicos para apoiar decisões econômicas e políticas, baseadas em simulações em tempo real, e de curto, médio e longo prazos, com a Ferramenta de Avaliação de Solo e Água (SWAT). Para que o sistema SUPer funcione efetivamente, ele precisa de uma grande quantidade de bacias calibradas, de forma a abranger o comportamento hidrológico de todo o estado de Pernambuco, incluindo características físicas, hidrológicas, de relevo e uso da água. Nesta pesquisa, construímos e calibramos um modelo hidrológico integrado do Estado de Pernambuco utilizando o programa Soil and Water Assessment Tool (SWAT), a fim de subsidiar o sistema SUPer e torná-lo uma ferramenta ativa para a Agência Pernambucana de Águas e Clima (APAC), contribuindo para o processo de planejamento e gestão dos recursos hídricos do estado. Para isso, selecionamos oito bacias do Estado de Pernambuco: Goiana, Sirinhaém, Capibaribe, Una, Mundaú, Pajeú e Brígida. Diferentes componentes dos recursos hídricos foram simulados e as características dos reservatórios foram consideradas a nível da Unidade de Resposta Hidrológica (HRU). Os recursos hídricos foram quantificados a nível de sub-bacia com intervalos de tempo mensais. Além disso, o SWAT-CUP foi usado com uma abordagem semiautomática (SUFI2) para análise de sensibilidade e para os procedimentos de calibração e validação. Em uma análise geral, foram observados ajustes e desempenho estatístico satisfatórios entre os valores medidos e simulados de vazão durante os processos de calibração e validação na maioria das bacias analisadas, onde a bacia do Rio Brígida apresentou o pior desempenho e a bacia do Rio Goiana apresentou o melhor desempenho entre as oito bacias analisadas.

Parallelization of AMALGAM algorithm for a multi-objective optimization of a hydrological model

A calibration procedure is essential step to achieve a realistic model simulation particularly in hydrological model which simulates water cycle in the basin. This process is always faced with challenges due to selection of objective function and highly time-consuming. This study aimed to take advantage of parallel processing to accelerate the computations involved with simulation process of hydrologic model linked with the multi-objective optimization algorithm of AMALGAM for multi-site calibration of SWAT hydrologic model parameters. In order to illustrate how meaningful SWAT model calibration trade-off between the four objective functions involved in AMALGAM optimization program, the Pareto solution sets were provided. Furthermore, it is implemented a group of model runs with a number of cores involved (from one to eight) to demonstrate and evaluate the running of parallelized AMALGAM with taking advantages of “spmd” method to decrease the running time of the SWAT model. The results revealed the robustness of the method in reducing computational time of the parameter calibration significantly. This strategy with 4-objective functions focuses on high streamflow (Nash–Sutcliffe coefficient), low streamflow (Box–Cox transformed root–mean–square error), water balance (runoff coefficient error), and flashiness (slope of the flow duration curve error) provided an efficient tool to decide about the best simulation based on the investigated objective functions. This study also provides a strong basis for multi-objective optimization of hydrological and water quality models and its general analytical framework could be applied to other parts of the world.

Modeling the impact of climate change on streamflow responses in the Kessem watershed, Middle Awash sub-basin, Ethiopia

Abstract In this study, we examined how future climate change will affect streamflow responses in the Kessem watershed. Climate variables from SSP2-4.5 and SSP5-8.5 emission scenarios were extracted from GCMs for the 2040s (2031–2060) and 2070s (2061–2090). The bias-corrected precipitation and temperature were converted into streamflow using a calibrated SWAT model. The simulated output of the future streamflow for the periods 2040s and 2070s was compared with the base period (1992–2020) and presented as percentage changes. During calibration and validation, the SWAT model showed Nash–Sutcliffe efficiency (NSE) values of 0.79 and 0.77, as well as coefficient of determination (R2) values of 0.8 and 0.79, demonstrating its capability of simulating streamflow. The annual mean maximum and minimum temperatures are predicted to increase, with a pronounced increase in the minimum temperature for the mid-term and long-term futures under both emission scenarios. As we approach the end of the century, we see an increase in annual mean rainfall and streamflow under the SSP5-8.5 emission scenario. The increment in annual mean rainfall (streamflow) is expected to be 3% (12.5%) and 23% (48.8%) for the 2040s and 2070s, respectively, under the SSP5-8.5 emission scenario.

SWAT model calibration for hydrological modeling using concurrent methods, a case of the Nile Nyabarongo River basin in Rwanda

The Nile Nyabarongo, which is Rwanda's largest river, is facing stress from both human activities and climate change. These factors have a substantial contribution to the water processes, making it difficult to effectively manage water resources. To address this issue, it is important to find out the most accurate techniques for simulating hydrological processes. This study aimed to calibrate the SWAT model employing various algorithms such as GLUE, ParaSol, and SUFI-2 for the simulation of hydrology in the basin of the Nile Nyabarongo River. Different data sources, such as DEM, Landsat images, soil data, and daily meteorological data, were utilized to input information into the SWAT modeling process. To divide the basin area effectively, 25 sub-basins were created, with due consideration of soil characteristics and the diverse land cover. The outcomes point out that SUFI-2 outperformed the other algorithms for SWAT calibration, requiring fewer computing model runs and producing the best results. ParaSol established residing the least effective algorithm. After calibration with SUFI-2, the most sensitive parameters for modeling were revealed to be (1) the Effective Channel Hydraulic Conductivity (CH K2) measuring how well water can flow through a channel, with higher values indicating better conductivity, (2) Manning's n value (CH N2) representing the roughness or resistance to flow within the channel, with smaller values suggesting a smoother channel, (3) Surface Runoff Lag Time (SURLAG) quantifying the delay between rainfall and the occurrence of surface runoff, with shorter values indicating faster runoff response, (4) the Universal Soil-Loss Equation (USLE P) estimating the amount of soil loss. The average evapotranspiration within the basin was calculated to be 559.5 mma-1. These calibration results are important for decision-making and updating policies related to water balance management in the basin.

Asymmetry of blue and green water changes in the Yangtze river basin, China, examined by multi-water-variable calibrated SWAT model

Renewable freshwater is a fundamental natural resource for sustaining human social development and the health of terrestrial ecosystems. While blue water (BW), which can be directly utilized by human beings, has received significant attention, the importance of green water (GW), crucial for ecosystem safety, has often been overlooked in water resource assessments. To date, there has been a lack of integrated research incorporating the detection and attribution of changes in both BW and GW in the Yangtze River basin (YRB), China’s largest river basin, resulting in a deficiency of comparative and interactive analyses between them. In this study, we examined the historical trends and long-term memory of BW and GW flows across 272 sub-basins of the YRB using a multi-water-variable calibrated SWAT model, which was validated by streamflow, evapotranspiration, and terrestrial water storage change. We determined the individual contributions of climate variability and land use changes to BW and GW changes. Additionally, we identified the responses characteristics of BW and GW changes in four different ecosystems primarily characterized by alpine grasslands, cropland, forests, and river–lake systems, respectively. The results indicate a non-significant decreasing trend in precipitation and BW in most regions, while an increasing trend is observed in GW, which is expected to persist in the future. Climate variability played a predominant role in BW and GW changes, while land use change also appeared to have significant impact on monthly GW changes. GW showed a positive correlation with precipitation/humidity index in relatively dry and moderately humid areas, but a negative correlation in highly humid areas. Moreover, a stronger correlation was found between the gross primary production (GPP) of terrestrial ecosystems and GW compared to GPP and BW. The findings contribute to guiding the harmonious utilization of water resources between humans and ecosystems in the YRB, as well as optimizing water allocation.

Investigation of climate change impact on the optimal operation of koka reservoir, upper awash watershed, Ethiopia

The objectives of this study were to predict the inflow and optimal operation of the Koka reservoir under the impact of climate change for the 2020s (2011–2040), 2050s (2041–2070), and 2080s (2071–2100) with respect to the reference period (1981–2010). The optimal elevation, storage, and hydropower capacity were modeled using the HEC-ResPRM, whereas the inflow to Koka reservoir was simulated using the calibrated SWAT model. Based on the result, the average annual inflow of the reference period was 139.675 Million Cubic Meter (MCM). However, from 2011 to 2100 an increase of +4.179% to +11.694 is expected. The inflow analysis at different flow regimes shows that the high flow may decline by (−28.528%) to (−22.856%) due to climate change. On the other hand, the low flow is projected to increase by (+78.407%) to (+90.401%) as compared to the low flow of the reference period. Therefore, the impact of climate change on the inflow to the Koka reservoir is positive. The study also indicates that the optimum values of elevation and storage capacity of the Koka reservoir during the reference period were 1590.771 m above mean sea level (a.m.s.l) and 1860.818 MCM, respectively. However, the optimum level and storage capacity are expected to change by (−0.016%) to (−0.039%) and (−2.677%) to (+6.164%), respectively from 2020s to 2080s as compared with their corresponding values during the reference period. On the other hand, the optimum power capacity during the reference period was 16.489 MCM, while it will likely fluctuates between (−0.948%) - (+0.386%) in the face of climate change. The study shows that the optimum elevation, storage, and power capacity were all higher than the corresponding observed values. However, the occurrence month of their peak value will likely shift due to climate change. The study can be used as a first-hand information for the development of reservoir operation guidelines that can account for the uncertainty caused by the impacts of climate change.

SWAT model calibration approaches in an integrated paddy-dominated catchment-command

SWAT model calibration approaches in an integrated paddy-dominated catchment-command

Sensitivity analysis, Calibration and Validation of SWAT model for the Lower Mahanadi River Basin

Sensitivity analysis, Calibration and Validation of SWAT model for the Lower Mahanadi River Basin

Can Gauss‐Newton Algorithms Outperform Stochastic Optimization Algorithms When Calibrating a Highly Parameterized Hydrological Model? A Case Study Using SWAT

AbstractThe calibration of highly parameterized hydrological models is a major computational challenge, especially for models with long run times. This challenge motivates the reconsideration of gradient‐based algorithms often overlooked for their perceived lack of robustness. Our study evaluates two Gauss‐Newton algorithms, robust Gauss‐Newton (RGN), and Levenberg‐Marquardt (PEST), and two stochastic algorithms, Shuffled Complex Evolution (SCE), and Dynamically Dimensioned Search (DDS), on a 38‐parameter SWAT model calibration problem. Algorithm performance is comprehensively assessed using trajectory plots from 100 invocations and by analyzing the distribution of estimated optima at fixed budgets of 200, 500, 1,000, 2,000, 3,000, and 5,000 objective function evaluations (model runs). Empirical results indicate that: (a) Gauss‐Newton algorithms are more likely than stochastic algorithms to locate good solutions for the budgets considered in this work, and more likely to locate satisfactory solutions when budget is tight (200–500 model runs) and (b) RGN shows the fastest initial convergence amongst the algorithms under consideration and has the highest chance of finding satisfactory solutions when the budget is tight. The results indicate that Gauss‐Newton algorithms offer an attractive choice for the calibration of highly parameterized hydrological models.

Validation and calibration of SWAT model for Kollur River Basin, Kundapura Taluk, Udupi District, Karnataka, India

Validation and calibration of SWAT model for Kollur River Basin, Kundapura Taluk, Udupi District, Karnataka, India

Determination of rainfed wheat agriculture potential through assimilation of remote sensing data with SWAT model case study: ZarrinehRoud Basin, Iran

Abstract Considering the importance of rainfed agriculture in adaptation to nature and long-term sustainability in the human food supply and livelihood of farmers, the main purpose of this study is to investigate the potential of rainfed agriculture in the Zarrinehroud basin as this basin is one of the most important sub-basins of Lake Urmia. For this study, the remote sensing data of surface soil moisture and evapotranspiration were combined with the SWAT model using the Data Assimilation method, Ensemble Kalman Filter (EnKF). Calibration of runoff flow rate in the SWAT model showed the correlation coefficient ranging between 0.69 and 0.84 in the calibration period (2000–2009) and between 0.64 and 0.86 for the validation period (2010–2014). The assimilation of the remote sensing data with the calibrated SWAT model showed that the model simulations for both the variables of surface soil moisture and actual evapotranspiration improved by at least 25% in both 2010 and 2014. It has been determined that 10.5 and 25.4% of the region's lands have a Very Appropriate and Appropriate potential for rainfed wheat agriculture, respectively. Areas with Moderate and Inappropriate potential occupy 64.1% of the lands in the region.

Multi-site multi-objective calibration of SWAT model using a large dataset for improved performance in Ethiopia

Multi-site multi-objective calibration of SWAT model using a large dataset for improved performance in Ethiopia

Multi-site calibration of hydrological model and the response of water balance components to land use land cover change in a rift valley Lake Basin in Ethiopia

This study aims to perform multi-site calibration and validation of SWAT model using streamflow data, and investigating the responses of water balance components to land use land cover (LULC) change in the Gidabo river sub-basin of rift valley lake basin in southern part of Ethiopia. SUFI-2 algorithm embedded in the SWAT-CUP was employed for sensitivity analysis, calibration, and validation on a monthly basis. The 17 years (1990–2006) streamflow data for three (Aposto, Bedessa, and Measso) gauging stations were used for calibrating the model while 8 years (2007–2014) streamflow data was used for validating the model. The calibrated and validated SWAT model was then used to investigate the response of water balance components to LULC change for three periods (1990, 2005, and 2019) which were performed using ERDAS Imagine 2014 with a maximum likelihood classifier. The most common statistical model performance evaluation indices namely Coefficient of Determination (R2), Nash-Sutcliffe Efficiency (NSE), and Percent Bias (PBIAS) were used to evaluate performance of the SWAT model in simulating sub-basin hydrology; in addition to physical inspection of observed and simulated streamflow hydrograph. The findings of the multi-site model performance evaluation indicated that the values of R2 ranged from 0.80 to 0.64 and 0.74 to 0.72 during calibration and validation periods respectively. The values of NSE ranged from 0.74 to 0.61 and 0.71 to 0.65 during calibration and validation periods respectively whereas PBIAS ranged from 19.70 to -3.20 and 18.10 to 0.80 during calibration and validation periods respectively. The calibration and validation results indicated that the SWAT model would simulate fairly well the historical streamflow at three gauging stations. The mean annual streamflow response to LULC change for the periods 1990–2005 and 2005–2019 was observed to increase by 2.13% (1.16m3/s) and 3.62% (2.04m3/s), respectively and the mean seasonal streamflow was obtained to increase during wet season (April-September) while decreasing trend was observed during dry season (October-March) in all three gauged stations. Results also revealed that there were significant spatiotemporal variations of surface runoff, groundwater flow, lateral flow, and evapotranspiration in the sub-basin. The multi-site calibration and validation together with uncertainty analysis detects spatial variability and simulates the water balance components under changing LULC, which is paramount importance for planning and formulating appropriate integrated land and water resources management and development strategies in the rift valley lake basin of Gidabo river sub-basin in Ethiopia.

Optimizing land use systems of an agricultural watershed in China to meet ecological and economic requirements for future sustainability

The contradiction between economic development and water pollution is ever so obvious in most agricultural watersheds in China. The continuous popularization of commercial crops in hills, especially tea, has intensified water eutrophication, which has attracted close attention of the Chinese government. Under this circumstance, local authorities have proposed the returning of tea plantations to forests, which would distinctly inhibit the development of rural economics. Hence, we employed a calibrated SWAT model to determine the total nitrogen/total phosphorus (TN/TP) loads in the reservoir, nutrient contribution of each land use type, and landscape pattern in a representative agricultural watershed. Aimed at the synchronous development of the economy and environment, this research study investigated the correlation between landscape pattern and nutrient output to accordingly propose alternative management strategies based on the correlation and actual watershed situation. Correlation analysis suggested that the landscape pattern prominently affects nutrient output in farmlands, tea plantations, orchards, and grasslands, and that the impact varies with the land use types and landscape parameters. Furthermore, a stepwise multiple regression model revealed that the landscape pattern could effectively predict nutrient output in farmlands, tea plantations, orchards, and grasslands. The aggregation index ( AI ) ( R 2 = 0.996, p < 0.05) and contiguity index ( CONTIG ) ( R 2 = 0.963, p < 0.01) are the primary landscape drivers for the export of TN and TP in farmlands, respectively, and the percentage of landscape ( PLAND ) is the key driver for TN loss in tea plantations ( R 2 = 0.841, p < 0.05). The fractal dimension index ( FRAC ) is the pivotal driver both for TN ( R 2 = 0.969, p < 0.01) and TP ( R 2 = 0.845, p < 0.05) export from orchards. According to the actual watershed situation and the relationship between nutrient output and landscape pattern, six alternative management strategies were performed: planting tea along contour lines (S1), constructing filter strips downstream tea plantations and orchards (S2), changing the shape of orchards (S3), controlling the nutrient emissions of residential areas (S4), returning farmlands to forests (S5), and a combination of S1-S5 (S6). All these measures are superior to the current government strategy (returning tea plantations around reservoirs to forests) in terms of both cutting down reservoir nutrient loads and reducing economic loss. Among them, S6 is the optimal strategy; using this, TN and TP loads are expected to be reduced by 51.64% and 58.96%, respectively, compared with those in 2019, and the economic loss will be far lower than that upon implementation of the current government strategy. This study has provided detailed suggestions for watersheds plagued by the contradiction between agricultural economic development and water pollution. • Correlation between TN/TP output and landscape pattern indices varies with land use types. • Optimal land management strategies were proposed based on landscape pattern and watershed sustainable development. • Optimal strategy can reduce reservoir TN/TP loads by 50% with much lower economic loss than current government strategy.