Unmonitored Watersheds Research Articles

Assessing effect of best management practices in unmonitored watersheds using the coupled SWAT-BiLSTM approach

In order to enhance the simulation of BMPs (Best Management Practices) reduction effects in unmonitored watersheds, in this study, we combined the physically-based hydrological model Soil & Water Assessment Tool (SWAT) and the data-driven model Bi-directional Long Short-Term Memory (Bi-LSTM), using the very-high-resolution (VHR) Land Use and Land Cover (LULC) dataset SinoLC-1 as data input, to evaluate the feasibility of constructing a water environment model for the Ba-River Basin (BRB) in central China and improving streamflow prediction performance. In the SWAT-BiLSTM model, we calibrated the top five SWAT parameters sorted by P-Value, allowing SWAT to act as a transfer function to convert meteorological data into base flow and storm flow, serving as the data input for the Bi-LSTM model. This optimization improved the Bi-LSTM's learning process for the relationship between the target and explanatory variables. The daily streamflow prediction results showed that the hybrid model had 9 regions rated as "Very good," 2 as "Good," 2 as "Satisfactory," and 1 as "Unsatisfactory" among the 14 regions. The model achieved an NSE of 0.86, R2 of 0.85, and PBIAS of −2.71% for the overall daily streamflow prediction performance during the verification period of the BRB. This indicates that the hybrid model has high predictive accuracy and no significant systematic bias, providing a sound hydrodynamic environment for water quality simulation. The simulation results of different BMPs scenarios showed that in the scenarios with only one BMP measure, stubble mulch had the best reduction effect, with average reductions of 17.83% for TN and 36.17% for TP. In the scenarios with a combination of multiple BMP measures, the combination of stubble mulch, soil testing and formula fertilization, and vegetative filter strip performed the best, achieving average reductions of 42.71% for TN and 50.40% for TP. The hybrid model provides a novel approach to simulate BMPs' reduction effects in regions without measured hydrological data and has the potential for wide application in BMP-related decision-making.

Estimation of long-term series of total nutrient loads flowing into a large perialpine lake (Lake Como, Northern Italy) from incomplete discrete data by governmental monitoring

Continuous series of nutrient loads released into a water body are essential for nutrient budgeting, water quality modelling and watershed management activities. A quintessential example of pursued data are the series of total nutrient loads flowing from multiple tributaries into a lake. However, except for extraordinary cases in which high-frequency monitoring (HFM) stations are installed for both discharge and concentrations, measured nutrient loads are available on a discrete basis. Such observations are typically obtained by governmental agencies for environmental monitoring purposes, with an at best monthly resolution, yet commonly with gaps spanning years. Usually, monitoring activities are limited to major inflows, neglecting minor ones. However, the latters can play a more relevant role in nutrient load budgets than in hydrological ones, in response to different natural features and pollution among tributary watersheds. In this work, we present a methodology we developed to estimate long-term series of total nutrient loads flowing into a water body, employing as case study Lake Como, a large deep lake in the Italian Alps with manifold monitored and unmonitored tributary watersheds. The method uses observed long-term relationships between discharges and concentrations (Q – C) and available discharge measurements and hydrological estimations to estimate continuous load series for the monitored basins. For the unmonitored watersheds, Q – C relationships are estimated from those of the monitored basins, given the observed dependence of the power-law coefficients on basin hydromorphological parameters. This equals to extending the regionalisation approach applied in hydrology for rainfall and discharges to the ecohydrological field for nutrient load estimation. The application of the method to the case study led to overall annual load estimations congruent to traditional techniques and revealed interesting Q – C watershed dynamics at interannual time scales which could not be disclosed through previous approaches. This work represents an exploratory development and application of ecohydrological regionalisation techniques, whose future development is fostered.

Predicting Dynamic Riverine Nitrogen Export in Unmonitored Watersheds: Leveraging Insights of AI from Data-Rich Regions.

Terrestrial export of nitrogen is a critical Earth system process, but its global dynamics remain difficult to predict at a high spatiotemporal resolution. Here, we use deep learning (DL) to model daily riverine nitrogen export in response to hydrometeorological and anthropogenic drivers. Long short-term memory (LSTM) models for the daily concentration and flux of dissolved inorganic nitrogen (DIN) were built in a coastal watershed in southeastern China with a typical subtropical monsoon climate. The DL models exhibited excellent accuracy for both DIN concentration and flux, with Nash-Sutcliffe efficiency coefficients (NSEs) up to 0.67 and 0.92, respectively, a performance unlikely to be achieved by generic process-based models with comparable data quality. The flux model ensemble, without retraining, performed well (mean NSE = 0.32-0.84) in seven distinct watersheds in Asia, Europe, and North America, and retraining with multi-watershed data further improved the lowest NSE from 0.32 to 0.68. DL interpretation confirmed that interbasin consistency of riverine nitrogen export exists across different continents, which stems from the similarities in rainfall-runoff relationships. The multi-watershed flux model projects 0.60-12.4% increases in the nitrogen export to oceans from the studied watersheds under a 20% increase in fertilizer consumption, which rises to 6.7-20.1% with a 10% increase in runoff, indicating the synergistic effect of human activities and climate change. The DL-based method represents a successful case of explainable artificial intelligence in environmental science, providing a potential shortcut to a consistent understanding of the global daily-resolution dynamics of riverine nitrogen export under the currently limited data conditions.

Water quality prediction using SWAT-ANN coupled approach

Efficient and accurate prediction of river water quality is challenging due to the complex hydrological and environmental processes affecting their nature. The challenge is even bigger in unmonitored watersheds. Both process- and data-based approaches are utilized for this purpose, with each having its own strengths and weaknesses. The development of a hybrid model can potentially give robust solutions in this regard. To improve the water quality predictions in unmonitored watersheds, we developed a hybrid model by combining a process-based watershed model and artificial neural network (ANN). Combining these two models helped to optimize the calibration and validation process while accounting for the complex hydrological and water quality processes. The developed model was applied to watersheds in the Atlanta metropolitan area, USA, to predict monthly nitrate, ammonium, and phosphate loads. We treated the watersheds as unmonitored and tested the skill of the hybrid model accordingly. The hybrid model had good skills in predicting all three constituents. The model worked especially well for nitrate. As a matter of fact, it even outperformed SWAT models calibrated at each site. This work emphasizes the potential benefits of the proposed hybrid modeling framework for the prediction of water quality parameters in unmonitored watersheds.

Influence of land use and land cover patterns on seasonal water quality at multi-spatial scales

The influence of land use patterns on stream water quality is scale-dependent in space and time. Understanding the relationship between landscape characteristics and water quality is of great importance to improve water contamination prediction in unmonitored watersheds and for providing guidelines for watershed land use planning. In this study, five water sampling sites in the upper Dan River basin were established to monitor seasonal water chemical contamination over the period of 2000 to 2008. The relationships between land use patterns and water quality were analyzed across multiple-scales using redundancy analysis. The results showed that stream water quality variables displayed highly temporal variations, with electrical conductivity (EC), ammonium nitrogen (NH4+-N), nitrate nitrogen (NO3−N), and total suspended solids (TSS) all generally displaying higher levels in the wet season, while there were higher concentrations of biochemical oxygen demand (BOD5), chemical oxygen demand (CODcr), and dissolved oxygen (DO) in the dry season. The total contribution of land use patterns on overall water quality was stronger at the riparian scale than at the catchment and reach scales during the wet season. However, different land use metrics had different scale effects. Urban land had a higher positive relationship with degraded water quality at small scales than at large scales, whereas agricultural land displayed the opposite scale effects. Forest and grassland explained more water quality variations at the riparian scale than at other scales. Analyses of spatial development patterns suggested that size, density, aggregation, and diversity of landscape patterns were important factors impacting on stream water quality. The results provide important information regarding sustainable land use and landscape planning at multiple-scales that can be used to improve water quality.

Coupling SWAT and ANN models for enhanced daily streamflow prediction

To improve daily flow prediction in unmonitored watersheds a hybrid model was developed by combining a quasi‐distributed watershed model and artificial neural network (ANN). Daily streamflow data from 29 nearby watersheds in and around the city of Atlanta, Southeastern United States, with leave-one-site-out jackknifing technique were used to build the flow predictive models during warm and cool seasons. Daily streamflow was first simulated with the Soil and Water Assessment Tool (SWAT) and then the SWAT simulated baseflow and stormflow were used as inputs to ANN. Out of the total 29 test watersheds, 62% and 83% of them had Nash–Sutcliffe values above 0.50 during the cool and warm seasons, respectively (considered good or better). As the percent forest cover or the size of test watershed increased, the performances of the models gradually decreased during both warm and cool seasons. This indicates that the developed models work better in urbanized watersheds. In addition, SWAT and SWAT Calibration Uncertainty Procedure (SWAT-CUP) program were run separately for each station to compare the flow prediction accuracy of the hybrid approach to SWAT. Only 31% of the sites during the calibration and 34% of validation runs had ENASH values ⩾0.50. This study showed that coupling ANN with semi-distributed models can lead to improved daily streamflow predictions in ungauged watersheds.

Estimation of Nonpoint Source Nitrate Concentrations in Indiana Rivers Based on Agricultural Drainage in the Watershed

AbstractSubsurface tile‐drained agricultural fields are known to be important contributors to nitrate in surface water in the Midwest, but the effect of these fields on nitrate at the watershed scale is difficult to quantify. Data for 25 watersheds monitored by the Indiana Department of Environmental Management and located near a U.S. Geological Survey stream gage were used to investigate the relationship between flow‐weighted mean concentration (FWMC) of nitrate‐N and the subsurface tile‐drained area (DA) of the watershed. The tile DA was estimated from soil drainage class, land use, and slope. Nitrate loads from point sources were estimated based on reported flows of major permitted facilities with mean nitrate‐N concentrations from published sources. Linear regression models exhibited a statistically significant relationship between annual/monthly nonpoint source (NPS) nitrate‐N and DA percentage. The annual model explained 71% of the variation in FWMC of nitrate‐N. The annual and monthly models were tested in 10 additional watersheds, most with absolute errors within 1 mg/l in the predicted FWMC. These models can be used to estimate NPS nitrate for unmonitored watersheds in similar areas, especially for drained agricultural areas where model performance was strongest, and to predict the nitrate reduction when various tile drainage management techniques are employed.

“Nutrient Inputs to the Laurentian Great Lakes by Source and Watershed Estimated Using SPARROW Watershed Models” by Dale M. Robertson and David A. Saad2

Richards, R. Peter, Ibrahim Alameddine, J. David Allan, David B. Baker, Nathan S. Bosch, Remegio Confesor, Joseph V. DePinto, David M. Dolan, Jeffrey M. Reutter, and Donald Scavia, 2012. Discussion –“Nutrient Inputs to the Laurentian Great Lakes by Source and Watershed Estimated Using SPARROW Watershed Models” by Dale M. Robertson and David A. Saad. Journal of the American Water Resources Association (JAWRA) 1‐10. DOI: 10.1111/jawr.12006Abstract: Results from the Upper Midwest Major River Basin (MRB3) SPARROW model and underlying Fluxmaster load estimates were compared with detailed data available in the Lake Erie and Ohio River watersheds. Fluxmaster and SPARROW estimates of tributary loads tend to be biased low for total phosphorus and high for total nitrogen. These and other limitations of the application led to an overestimation of the relative contribution of point sources vs. nonpoint sources of phosphorus to eutrophication conditions in Lake Erie, when compared with direct estimates for data‐rich Ohio tributaries. These limitations include the use of a decade‐old reference point (2002), lack of modeling of dissolved phosphorus, lack of inclusion of inputs from the Canadian Lake Erie watersheds and from Lake Huron, and the choice to summarize results for the entire United States Lake Erie watershed, as opposed to the key Western and Central Basin watersheds that drive Lake Erie’s eutrophication processes. Although the MRB3 SPARROW model helps to meet a critical need by modeling unmonitored watersheds and ranking rivers by their estimated relative contributions, we recommend caution in use of the MRB3 SPARRROW model for Lake Erie management, and argue that the management of agricultural nonpoint sources should continue to be the primary focus for the Western and Central Basins of Lake Erie.

Controls on stream water dissolved mercury in three mid‐Appalachian forested headwater catchments

Determining the controls on dissolved mercury (HgD) transport is necessary to improve estimations of export from unmonitored watersheds and to forecast responses to changes in deposition and other environmental forcings. Stream water HgD and dissolved organic carbon (DOC) were evaluated over a range of discharge conditions in three streams within Shenandoah National Park, VA. Watersheds are distinguished by stream water pH (ranging from neutral to acidic) and soil size fractioning (ranging from clays to sands). At all sites, discharge was a significant but poor predictor of HgD concentrations (r2 from 0.13–0.52). HgD was strongly coupled with DOC at all sites (r2 from 0.74–0.89). UV absorbance at 254 nm (UV254), a proxy for DOC quantity and quality, slightly improved the predictions of HgD. Mean DOC quality differed between streams, with less aromatic DOC mobilized from the more acidic watershed. The site with less aromatic DOC and sandy soils mobilized more Hg to the stream for the same quantity and quality of DOC, likely due to the reduced capacity of the larger‐grained soils to retain Hg, leaving a greater fraction associated with the organic matter. A similar amount of 0.54 ng HgD/mg DOC is transported at all sites, suggesting the less aromatic DOC transports less Hg per unit DOC, offsetting the effects of soil type. This research demonstrates that soil composition and DOC quality influence HgDexport. We also provide evidence that soil organic carbon is a primary control on Hg‐DOC ratios (0.12–1.4 ng mg−1) observed across the U.S. and Sweden.

Detecting the Dynamic Linkage between Landscape Characteristics and Water Quality in a Subtropical Coastal Watershed, Southeast China

Geospatial analysis and statistical analysis are coupled in this study to determine the dynamic linkage between landscape characteristics and water quality for the years 1996, 2002, and 2007 in a subtropical coastal watershed of Southeast China. The landscape characteristics include Percent of Built (%BL), Percent of Agriculture, Percent of Natural, Patch Density and Shannon's Diversity Index (SHDI), with water quality expressed in terms of COD(Mn) and NH(4)(+)-N. The %BL was consistently positively correlated with NH(4)(+)-N and COD(Mn) at time three points. SHDI is significantly positively correlated with COD(Mn) in 2002. The relationship between NH(4)(+)-N, COD(Mn) and landscape variables in the wet precipitation year 2007 is stronger, with R(2) = 0.892, than that in the dry precipitation years 1996 and 2002, which had R(2) values of 0.712 and 0.455, respectively. Two empirical regression models constructed in this study proved more suitable for predicting COD(Mn) than for predicting NH(4)(+)-N concentration in the unmonitored watersheds that do not have wastewater treatment plants. The calibrated regression equations have a better predictive ability over space within the wet precipitation year of 2007 than over time during the dry precipitation years from 1996 to 2002. Results show clearly that climatic variability influences the linkage of water quality-landscape characteristics and the fit of empirical regression models.

Calibração e validação da equação universal de perda de solos modificada (MUSLE) utilizando dados hidrossedimentológicos locais

Modelos hidrossedimentológicos são úteis na análise ambiental de bacias hidrográficas não monitoradas. Entretanto, para que as suas predições sejam confiáveis, é necessário que os modelos sejam adequadamente calibrados e validados para as condições locais. O objetivo do presente trabalho foi calibrar o coeficiente a da MUSLE para as condições da bacia do ribeirão Pipiripau, usando dados hidrossedimentológicos locais, bem como validar a equação calibrada, com uma série de dados diferente da usada na calibração. O coeficiente a da MUSLE foi calibrado por meio do ajuste entre os valores observados e calculados de aporte de sedimento de eventos individuais, correspondentes ao período entre 1999 e 2005, usando dados hidrológicos (Q e q p) observados. Para validação do modelo calibrado, a série usada correspondeu aos anos de 1998 e 2006-2009 (N.A.: dados corretos, ver p. 12), onde apenas os dados pluviométricos locais e os fatores CN, K, L, S, C e P da bacia, previamente obtidos, foram usados. Os resultados indicam que a MUSLE calibrada apresenta melhor acurácia na estimativa do aporte de sedimento total anual (R² = 0,68 e E = 0,61) do que em nível mensal (R² = 0,44 e E = 0,43). Como o coeficiente de ajuste a da MUSLE foi proporcional ao volume de precipitação pluvial anual, este poderá ser usado para melhorar as predições do modelo.

Predicting Water Quality in Unmonitored Watersheds Using Artificial Neural Networks

Land use and land cover (LULC) play a central role in fate and transport of water quality (WQ) parameters in watersheds. Developing relationships between LULC and WQ parameters is essential for evaluating the quality of water resources. In this paper, we present an artificial neural network (ANN)-based methodology to predict WQ parameters in watersheds with no prior WQ data. The model relies on LULC percentages, temperature, and stream discharge as inputs. The approach is applied to 18 watersheds in west Georgia, United States, having a LULC gradient and varying in size from 2.96 to 26.59 km2. Out of 18 watersheds, 12 were used for training, 3 for validation, and 3 for testing the ANN model. The WQ parameters tested are total dissolved solids (TDS), total suspended solids (TSS), chlorine (Cl), nitrate (NO3), sulfate (SO4), sodium (Na), potassium (K), total phosphorus (TP), and dissolved organic carbon (DOC). Model performances are evaluated on the basis of a performance rating system whereby performances are categorized as unsatisfactory, satisfactory, good, or very good. Overall, the ANN models developed using the training data performed quite well in the independent test watersheds. Based on the rating system TDS, Cl, NO3, SO4, Na, K, and DOC had a performance of at least "good" in all three test watersheds. The average performance for TSS and TP in the three test watersheds were "good." Overall the model performed better in the pastoral and forested watersheds with an average rating of "very good." The average model performance at the urban watershed was "good." This study showed that if WQ and LULC data are available from multiple watersheds in an area with relatively similar physiographic properties, then one can successfully predict the impact of LULC changes on WQ in any nearby watershed.

Neural networks modelling of nitrogen export: model development and application to unmonitored boreal forest watersheds

In remotely located boreal forest watersheds, monitoring nitrogen (N) export in stream discharge often is not feasible because of high costs and site inaccessibility. Therefore, modelling tools that can predict N export in unmonitored watersheds are urgently needed to support management decisions for these watersheds. The hydrological and biogeochemical processes that regulate N export in streams draining watersheds are complex and not fully understood, which makes artificial neural network (ANN) modelling suitable for such an application. This study developed ANN models to predict N export from watersheds relying only on easily accessible climate data and remote sensing (RS) data from the public domain. The models were able to predict the daily N export (g/km2/d) in five watersheds ranging in size from 5–130 km2 with reasonable accuracy. Similarity indices were developed between any two studied watersheds to quantify watershed similarity and guide the transferability of models from monitored watersheds to unmonitored ones. To demonstrate the applicability of the ANN models to unmonitored watersheds, the calibrated ANN models were used to predict N export in different watersheds (unmonitored watersheds in this perspective) without further calibration. The similarity index based upon a rainfall index, a peatland index and a RS normalized difference water index showed the best correlation with the transferability of the models. This study represents an important first step towards transferring ANN models developed for one watershed to unmonitored watersheds using similarity indices that rely on freely available climate and RS data.

Fine‐Sediment Loadings to Lake Tahoe1

Abstract: Over the past 35 years, a trend of decreasing water clarity has been documented in Lake Tahoe, attributable in part to the delivery of fine‐grained sediments emanating from upland and channel sources. The overall objective of the research reported here was to determine the amount of fine sediment delivered to Lake Tahoe from each of the 63 contributing watersheds. The research described in this report used combinations of field‐based observations of channel and bank stability with measured and simulated data on fine‐sediment loadings to estimate fine‐sediment loadings from unmonitored basins throughout the Lake Tahoe Basin. Loadings were expressed in the conventional format of mass per unit time but also in the number of particles finer than 20 μm, the latter for future use in a lake‐clarity model. The greatest contributors of fine sediment happened to be those with measured data, not requiring extrapolation. In descending order, they are as follows: Upper Truckee River [1,010 tonnes per year (T/year)], Blackwood Creek (846 T/year), Trout Creek (462 T/year), and Ward Creek (412 T/year). Summing estimated values from the contributing watersheds provided an average, annual estimate of fine‐sediment (<0.063 mm) loadings to the lake of 5,206 T/year. A total of 7.79E + 19 particles in the 5‐20 μm fraction were calculated to enter Lake Tahoe in an average year with the Upper Truckee River accounting for almost 25% of the total. Contributions from Blackwood, Ward, Trout, and Third creeks account for another 23% of these very fine particles. Thus, these five streams making up about 40% of the basin area, account for almost 50% of all fine‐sediment loadings to the lake. Contribution of fine sediment from streambank erosion were estimated by developing empirical relations between measured or simulated bank‐erosion rates with a field‐based measure of the extent of bank instability along given streams. An average, annual fine‐sediment loading from streambank erosion of 1,305 T/year was calculated. This represents about 25% of the average, annual fine‐sediment load delivered to the lake from all sources. The two largest contributors, the Upper Truckee River (639 T/year) and Blackwood Creek (431 T/year), account for slightly more than 80% of all fines emanating from streambanks, representing about 20% of the fine sediment delivered to Lake Tahoe from all sources. Extrapolations of fine‐sediment loadings to the unmonitored watersheds are based on documented empirical relations, yet contain a significant amount of uncertainty. Except for those values derived directly from measured data, reported results should be considered as estimates.

MULTIVARIATE ANALYSIS OF WATER QUALITY AND PHYSICAL CHARACTERISTICS OF SELECTED WATERSHEDS IN PUERTO RICO1

ABSTRACT: Multivariate analyses were used to develop equations that could predict certain water quality (WQ) conditions for unmonitored watersheds in Puerto Rico based on their physical characteristics. Long term WQ data were used to represent the WQ of 15 watersheds in Puerto Rico. A factor analysis (FA) was performed to reduce the number of chemical constituents. Cluster analysis (CA) was used to group watersheds with similar WQ characteristics. Finally, a discriminant analysis (DA) was performed to relate the WQ clusters to different physical parameters and generate predicting equations. The FA identified six factors (77 percent of variation explained): nutrients, dissolved ions, sodium and chloride, silicacious geology, red ox conditions, and discharge. From the FA, specific conductance, sodium, phosphorous, silica, and dissolved oxygen were selected to represent the WQ characteristics in the CA. The CA determined five groups of watersheds (forested, urban polluted, mixed urban/rural, forested plutonic, and limestone) with similar WQ properties. From the five WQ clusters, two categories can be observed: forested and urban watersheds. The DA found that changes in forest cover, percent of limestone, mean annual rainfall, and watershed shape factor were the most important physical features affecting the WQ of watersheds in Puerto Rico.

Dendrohydrological Evidence of Flooding Along a Small Appalachian Stream

Tree-ring records from flood-damaged trees have been most frequently used for reconstructing flood histories along large streams. Rarely has this approach been taken to gain hydrological information about small headwater channels. In this study, I analyze dendrohydrological evidence of flooding along two headwater streams in Virginia to determine its usefulness for hydrologic and geomorphic investigations of small, unmonitored watersheds. Easily obtained tree-ring dates for corrasion scars, adventitious sprouts and thin-ring sequences comprise the fundamental data set. The most noteworthy aspects of the data set are the youthfulness of most samples and a large corrasion scar peak in 1989. Elevations of corrasion scars above the channel bed indicate higher peak 1989 discharge than that expected based on field observations. These observations, coupled with hydrological monitoring data from closest stations and HEC-1 modeling results, emphasize the importance of wet antecedent conditions and repeated moderate events in scar creation. Earlier studies of lake sediment do not resolve sediment yield from year-1989 flooding, and average sediment yield was generally low at this time. Geomorphic work accomplished and/or sediment delivery may have been lower during 1989 than those associated with the Hurricane Agnes event in 1972. Although botanical flood evidence from small streams, and in particular corrasion scarring, are demonstrated to be useful for some hydrologic analyses, their use in reconstructing flood chronology is limited to recent decades by rapid tree healing.

AN AGNPS-BASED RUNOFF AND SEDIMENT YIELD MODEL FOR TWO SMALLWATERSHEDS IN GERMANY

The event-based Agricultural Non-Point Source (AGNPS) pollution model is used extensively to simulatesurface runoff, sediment yield and nutrient transport in unmonitored watersheds. Investigation, that compare AGNPSpredictions to measured data are rare. The objective of the present study was to compare surface runoff and sedimentyield predictions from AGNPS water quality simulation model and modified versions to measured data. Shortcomings ofthe AGNPS model were examined. The study was carried out using 52 rainfall-runoff events, 22 for calibration and 30 forvalidation, from two small watersheds (G1 and G2) in Bavaria, Germany. Evaluation of model outputs was based onstatistical comparisons between measured and predicted values for each rainfall-runoff event. We compared threedifferent surface runoff prediction methods: uncalibrated curve number (Q1), calibrated curve number (Q2), and Lutz(Q3). The modifications made to sediment yield calculations encompassed: (i) replacement of the Universal Soil LossEquation LS factor algorithm (S1) by one based on stream power theory (S2), and (ii) linkage of channel erosion byindividual categories of particle size to runoff velocity (S3). Measured median for surface runoff was under-predicted by55.5% using Q1, overpredicted by 36.8% using Q2 and over-predicted by 13.1% using Q3 in G1. The largest coefficientof efficiency (E) was calculated for Q3 with 0.96 followed by 0.93 for Q2 and 0.25 for Q1 in G1. In G2, measured medianfor surface runoff was underpredicted by 80.0% using Q1, overpredicted by 45.0% using Q2, and overpredicted by 35.0%using Q3 in G2. Best performance in terms of E was calculated by Q3 (0.83) followed by 0.76 for Q2 and 0.24 for Q1 inG2. Median sediment yield measurement was underpredicted by 57.2% using S1, underpredicted by 47.6% using S2 andunderpredicted by 4.8% using S3 in G1. The largest E was calculated with 0.90 for S3 followed by 0.57 for S2 and 0.26for S1 in G1. Measured median for sediment yield was underpredicted by 53.9% using S1, underpredicted by 38.5% usingS2 and overpredicted by 3.3% using S3 in G2. E was largest with 0.72 (S3) followed by 0.60 (S2) and 0.57 (S1) in G2.Results of this study illustrated that a calibration of CN and Lutz method for surface runoff calculations and the use ofvariant S3 for sediment yield calculations with AGNPS model showed the highest merit to match measurements withpredictions at the drainage outlet.