Streamflow forecasting using LSTM and weighted curve number method in Tawi watershed, Western Himalaya

Runoff is one of the significant hydrologic variables used in generally of the water resources applications. The Soil Conservation Service–Curve Number (SCS–CN) method is adopted for the evaluation of surface runoff in the Karur District, Tamil Nadu, India using multispectral remote sensing data, rainfall data and curve number approach. The weighted curve number is determined based on antecedent moisture condition (AMC)-II with an integration of Hydrologic Soil Groups(HSGs) and land use/ land cover categories. The daily runoff was estimated for rainfall Period September 2018 to November 2018. The results of the present study shows that the runoff depth for the study area are ranging in between 106.57 and 713.65 mm and runoff volume are ranging in between 2.75 and 126.21 mcm. In the present study, the methodology for determination of runoff for study area using remote sensing, GIS and SCS–CN method was described. Keywords: Surface runoff estimation, sub-watershed, Remote Sensing, GIS, Curve number.

Investigating practical alternatives to the NRCS-CN method for direct runoff estimation using slope-adjusted curve numbers

The Soil Conservation Service curve number (CN) method is widely used for predicting direct runoff from rainfall. However, despite the extent of cultivation on hillslope areas, very few attempts have been made to incorporate a slope factor into the CN method. The objectives of this study were (1) to evaluate existing approaches integrating slope in the CN method, and (2) to develop an equation incorporating a slope factor into the CN method for application in the steep slope areas of the Loess Plateau of China. The dataset consisted of 11 years of rainfall and runoff measurements from two experimental sites with slopes ranging from 14 to 140%. The results indicated that the standard CN method underestimated large runoff events and overestimated small events. For our experimental conditions, the optimized and non‐optimized forms of the slope‐modified CN method of the Erosion Productivity Impact Calculator model improved runoff prediction for steep slopes, but large runoff events were still underestimated and small ones overpredicted. Based on relationships between slope and the observed and theoretical CN values, an equation was developed that better predicted runoff depths with an R2 of 0·822 and a linear regression slope of 0·807. This slope‐adjusted CN equation appears to be the most appropriate for runoff prediction in the steep areas of the Loess Plateau of China. Copyright © 2005 John Wiley & Sons, Ltd.

The present work proposes to improve estimates of snowpack and snowmelt and their assessment in the steep Himalayan ranges at the sub-catchment scale. Temporal variability of streamflow and the associated distribution of accumulated snow in catchments with glacier presence in the Himalayas illustrates how changes in snowpack and snowmelt can affect the water supply for local water management. The primary objective of this study is to assess the role of elevation, temperature lapse rate (TLR), and precipitation lapse rate (PLR) in the computation of snowpack (or snowfall) and snowmelt in sub-catchments of the Satluj River basin. Modeling of snowpack and snowmelt was constructed using the Soil Water Assessment Tool (SWAT) in both historical (1991–2008) and near-time scenarios (2011–2030) by implementing real-time hydrometeorological, snow-hydrological parameters, and Global Circulation Model (GCM) datasets. The modeled snowmelt-induced streamflow showed a good agreement with the observed streamflow (~60%), calibrated and validated at three gauges. A Sequential Uncertainty Parameter Fitting (SUFI2) method (SUFI2) resulted that the curve number (CN2) was found to be significantly sensitive during calibration. The snowmelt hydrological parameters such as snowmelt factor maximum (SMFMX) and snow coverage (SNO50COV) significantly affected objective functions, such as R2 and NSE, during the model optimization. For the validation of snowpack and snowmelt, the results have been contrasted with previous studies and found comparable. The computed snowpack and snowmelt were found highly variable over the Himalayan sub-catchments, as also reported by previous researchers. The magnitude of snowpack change consistently decreases across all the sub-catchments of the Satluj river catchment (varying between 4% and 42%). The highest percentage of changes in the snowpack was observed over high-elevation sub-catchments.

ABSTRACTThe Natural Resources Conservation Service (NRCS)‐curve number (CN) method was originally proposed to predict runoff on small and midsize catchments, but it has also been used at the scale of erosion plots. In this case, uncertainties exist with reference to the factors, for example, scale effects, affecting the experimental CN values. In this study, the reliability of the CN method in reproducing plot runoff is analysed by using data collected at the Sparacia erosion plots (Sicily, Southern Italy), which are characterised by different sizes and steepness. This investigation aimed to test the possibility of using simulated runoff within universal soil loss equation (USLE)‐type models, including runoff as a term in the erosivity factor. This analysis pointed out that the experimentally determined value of the initial abstraction ratio of the CN method was very low (0.0001). For each plot type (i.e., fixed length and steepness), the calibration was performed for 18 combinations of three rainfall ranges (all data, rainfall depth less than the median, and exceeding the median), two calibration approaches (least‐squares and median value) and three datasets (all data, interrill, and rill). The best CN model fit was systematically produced for data with rainfall depth less than the median. The least‐squares calibration approach generally performed slightly better than the median value one. Results showed that the CN method can be considered effective only for events producing rills. The CN values generally increased with plot steepness and decreased as plot length increased. For each plot type, CN tendentially increased for increasing soil moisture before the rainfall event, but moisture and rainfall depth were able to explain a minor part (from 19.5% to 41%) of CN variance. Finally, the USLE‐MB that incorporates runoff simulated by the CN method was found to satisfactorily predict (relative standard error = 0.69, Nash and Sutcliffe Efficiency Index = 0.54) event soil loss caused by simultaneous interrill and rill erosion due to the higher rainfall depths recorded at the Sparacia station.

ABSTRACT: The U.S. Department of Agriculture Curve Number (CN) method is one of the most common and widely used techniques for estimating surface runoff and has been incorporated into a number of popular hydrologic models. The CN method has traditionally been applied using compositing techniques in which the area weighted average of all curve numbers is calculated for a watershed or a small number of sub‐watersheds. CN compositing was originally developed as a time saving procedure, reducing the number of runoff calculations required. However, with the proliferation of high speed computers and geographic information systems, it is now feasible to use distributed CNs when applying the CN method. To determine the effect of using composited versus distributed CNs on runoff estimates, two simulations of idealized watersheds were developed to compare runoff depths using composite and distributed CNs. The results of these simulations were compared to the results of similar analyses performed on an urbanizing watershed located in central Indiana and show that runoff depth estimates using distributed CNs are as much as 100 percent higher than when composited CNs are used. Underestimation of runoff due to CN compositing is a result of the curvilinear relationship between CN and runoff depth and is most severe for wide CN ranges, low CN values, and low precipitation depths. For larger design storms, however, the difference in runoff computed using composite and distributed CNs is minimal.

In the 1950s, the United States Department of Agriculture (USDA)–Soil Conservation Service (SCS) developed the empirical runoff curve number (CN) method, providing modelers, planners, and designers a tool for estimating runoff from rainfall accounting for losses such as evaporation, transpiration, infiltration, and surface storage. SCS originally developed the CN method to analyze primarily agricultural watersheds for watershed protection and flood prevention operations Act of 1954 (PL-83-566) projects. Today modelers world-wide use the CN method, or methods based on it, to model agricultural and urban hydrology. For more than 80 years, SCS, and its successor, the Natural Resources Conservation Service (NRCS), worked and continue to work with farmers; ranchers, State, Tribal, and local governments; and other Federal agencies to maintain healthy and productive working landscapes. In 2015, NRCS partnered with the American Society of Civil Engineers (ASCE) to develop updates to four chapters for the NRCS National Engineering Handbook, Part 630, Hydrology (NEH-630), to incorporate the latest research recommendations to revise the Ia/S (initial abstraction/maximum potential storage) ratio. This represents the first significant update to the CN method since its original development. The Curve Number (CN) Task Committee, under the ASCE–Environmental and Water Resources Institute’s Watershed Management Technical Committee was instrumental in this effort. Through the internal review process, NRCS received over 1,000 individual comments and identified items needing further review. NRCS continues to work to incorporate the internal review comments and additional update recommendations from the CN Task Committee. This paper discusses the purpose of the NEH-630, the process of developing the NEH updates, and actions taken to finalize the NEH-630 CN chapters.

The application of hydrologic modeling tools to represent urban watersheds is widespread, and calculation of infiltration losses is an essential component of these models. The curve number (CN) method is widely used in such models and is implemented in US EPA’s Storm Water Management Model (SWMM 5). SWMM 5 models can be created either using CN values computed only for the pervious fraction of subcatchments, or using the entire subcatchment area, but choice is not clearly understood. The present work evaluates the differences between these approaches in CN computation within SWMM through a comparison with field data collected in an urban watershed in Alabama and with WinTR-55. Four approaches to computing CN were considered in which the impervious fractions varied according to a threshold CN value. Results indicated that a Fully Composite approach, which computed CN from all subcatchment areas, yielded the best results for the sub-watershed with higher average CN. It was also observed that results from the approaches using CN Cut-off values of 90 and 93 were better for subcatchments with lower average CN. The comparison between SWMM 5 and WinTR-55 indicated that SWMM 5 hydrographs had larger peak flow rates, but these differences decreased with larger intensity rain events. Research findings are useful to hydrologic modelers, and in particular for setting up SWMM 5 models using CN method.

Warner, Richard C., Carmen T. Agouridis, Page T. Vingralek, and Alex W. Fogle, 2010. Reclaimed Mineland Curve Number Response to Temporal Distribution of Rainfall. Journal of the American Water Resources Association (JAWRA) 46(4): 724‐732. DOI: 10.1111/j.1752‐1688.2010.00444.xAbstract: The curve number (CN) method is a common technique to estimate runoff volume, and it is widely used in coal mining operations such as those in the Appalachian region of Kentucky. However, very little CN data are available for watersheds disturbed by surface mining and then reclaimed using traditional techniques. Furthermore, as the CN method does not readily account for variations in infiltration rates due to varying rainfall distributions, the selection of a single CN value to encompass all temporal rainfall distributions could lead engineers to substantially under‐ or over‐size water detention structures used in mining operations or other land uses such as development. Using rainfall and runoff data from a surface coal mine located in the Cumberland Plateau of eastern Kentucky, CNs were computed for conventionally reclaimed lands. The effects of temporal rainfall distributions on CNs was also examined by classifying storms as intense, steady, multi‐interval intense, or multi‐interval steady. Results indicate that CNs for such reclaimed lands ranged from 62 to 94 with a mean value of 85. Temporal rainfall distributions were also shown to significantly affect CN values with intense storms having significantly higher CNs than multi‐interval storms. These results indicate that a period of recovery is present between rainfall bursts of a multi‐interval storm that allows depressional storage and infiltration rates to rebound.

Abstract. The curve number (CN) method was developed more than half a century ago and is still used in many watershed and water-quality models to estimate direct runoff from a rainfall event. Despite its popularity, the method is plagued by a conceptual problem where CN is assumed to be constant for a given set of watershed conditions, but many field observations show that CN decreases with event rainfall (P). Recent studies indicate that heterogeneity within the watershed is the cause of this behavior, but the governing mechanism remains poorly understood. This study shows that heterogeneity in initial abstraction, Ia, can be used to explain how CN varies with P. By conventional definition, Ia is equal to the cumulative rainfall before the onset of runoff and is assumed to be constant for a given set of watershed conditions. Our analysis shows that the total storage in Ia (IaT) is constant, but the effective Ia varies with P, and is equal to the filled portion ofIaT, which we call IaF. CN calculated using IaF varies with P similar to published field observations. This motivated modifications to the CN method, called variable Ia models (VIMs), which replace Ia with IaF. VIMs were evaluated against conventional models CM0.2 (λ = 0.2) and CMλ (calibrated λ) in their ability to predict runoff data generated using a distributed parameter CN model. The performance of CM0.2 was the poorest, whereas those of the VIMs were the best in predicting overall runoff and watershed heterogeneity. VIMs also predicted the runoff from smaller events better than the CMs and eliminated the false prediction of zero-runoffs, which is a common shortcoming of the CMs. We conclude that including variable Ia accounts for heterogeneity and improves the performance of the CN method while retaining its simplicity.

A popular way for describing the link between storm rainfall depth and direct runoff is the curve number (CN) method. It is a straightforward approach that has been extensively studied and widely adopted. However, there has been less focus on the impact of slope and the initial abstraction ratio, which is a crucial factor for accurately estimating direct runoff when utilizing the soil conservation service- Curve Number (SCS-CN) method. The initial abstraction ratio is typically assumed to be 0.20, as initially proposed by the method's developers. In this study, we analyzed daily rainfall data from seventeen watersheds in different physiographic locations in the Kurdistan region of Iraq, recorded between 2022 and 2023. Our aim was to assess the effect of slope adjusted curve number and modified the initial abstraction ratio (0.1) on estimation of direct runoff. The results demonstrated that adjusting the CN for slope and using a modified initial abstraction ratio increased the estimated runoff compared to the original method (without adjustment for slope and initial abstraction ratio=0.2). Therefore, when applying the SCS-CN method, it is crucial to correct the CN for slope in steeper areas and consider the initial abstraction ratio rather than relying on the suggested value of 0.2. this study highlights the importance of considering local conditions and estimating the initial abstraction ratio based on specific watershed characteristics to enhance the accuracy of direct runoff estimation using the CN method.

Highlights The revised CN equation uses decreased CN values, causing decreased initial abstraction values and increased runoff. SWAT’s water balance approach simulated increased runoff to decrease groundwater flow and associated baseflow with the revised SCS CN method. Landscape analysis showed minor changes in runoff and nutrient loadings, except in row cropping areas in high runoff potential soils. Reduced mitigation occurs if post-development runoff increases by a greater margin than the pre-development runoff. Abstract. The Soil Conservation Service (SCS) Curve Number (CN) method is a widely used model developed by the Natural Resources Conservation Service (NRCS) that estimates surface runoff generated from a land area using factors such as land cover, soil type, and precipitation depth. However, because this empirical model was developed over 60 years ago with limited data, the NRCS is considering revising it with the help of an ASCE-ASABE task group. The proposed revisions include updating the initial abstraction (Ia) ratio in the CN equation and modifying the curve number (CN) for all land use and soil hydrologic group combinations. The proposed revisions are expected to alter the hydrologic response estimation, especially for low flow events. Since surface runoff is driving sediment, nutrients, and the transport of chemicals in natural systems, this study aims to understand how predicted hydrologic and water quality variables vary between the original and the proposed revised versions of the SCS CN method using the Soil and Water Assessment Tool (SWAT). Additionally, a case study site was selected to understand how the revisions alter the design storm runoff analysis. Results from the simulation case study on a 36,900 ha mixed land-use watershed located in central Pennsylvania, USA indicate a 5% to 14% increase in annual runoff predictions with the revised equation compared to the original equation. Though most average monthly runoff depths increased by 1% to 25% using the revised equation, winter months generally simulated runoff depths that were 1% to 5% lower than the original equation. Average daily loadings at the watershed outlet increased by 11% to 19% and 4% to 11% for nitrate and mineral phosphorus, respectively. The SWAT simulation results suggest that current agricultural and urban best management practices (BMPs) will experience higher runoff volumes on a yearly time scale due to the decrease in the Ia term, while the design storm case study results suggest that urban BMPs can be smaller in size when rural lands are urbanized due to the reduction in runoff estimation from the initial land use with the revised equation. Additional detailed evaluations of the proposed SCS CN method are required for water quality simulations, stormwater management, and BMP designs. Keywords: Hydrology, Revised curve number method, SCS Curve Number Method, Soil and Water Assessment Tool, Water quality.

The USDA Soil Conservation Service (SCS) curve number (CN) method has been the foundation of the hydrology algorithms in commonly used continuous simulation models, including the Soil and Water Assessment Tool (SWAT). This expanded use of the SCS-CN method has proven successful for many applications. However, because the SCS-CN method was originally developed to determine design peak discharges of synthetic storm events under an average antecedent moisture condition, research is needed to address the controversy over the use of this method to represent continuous precipitation runoff processes. In addition, poor results obtained for some conditions indicate the necessity to improve the method to provide a more realistic and accurate representation of water flow amounts, paths, and source areas upon which erosion and water quality predictions depend. Thus, the objectives of this study were to: (1) propose a modified curve number (MCN) method, and (2) assess the MCN method relative to the existing SWAT method with an Ia/S value either equal to 0.2 or 0.05. The equations that formulate the MCN method were coded into the SWAT 2005 framework. A SWAT model implementing the MCN method was evaluated along with the models implementing the existing SWAT method with Ia/S values of 0.2 and 0.05. The evaluation was conducted in the 870 km2 upper portion of the Forest River watershed located in northeastern North Dakota. The results revealed that the total streamflows predicted by the three models were comparable, as indicated by similar values for the Nash-Sutcliffe coefficient. However, the MCN approach resulted in the most accurate prediction of the streamflow components (i.e., baseflow versus direct flow) as well as water yields. For the study area, the MCN method was judged to be superior to the existing commonly used curve number methods in terms of mimicking the precipitation runoff processes.

Despite gullies significantly contributing to land degradation happening globally, predicting their spatial patterns in relation to climate, land use, and other factors remains challenging, especially in a process‐oriented manner. Nevertheless, such models appear crucial for developing effective land management strategies. Over the past years, several studies have proposed using the curve number (CN) method to estimate potential runoff discharges. However, although promising, the actual performance of such a CN‐based approach remains poorly tested, especially at large scales. Here we address this gap by evaluating the CN method's ability to predict gully head occurrence at different spatial scales in a process‐oriented way. We propose a gully head initiation (GHI) index, reflecting the ratio between a shear stress index (SSI) and a critical shear stress index (CSI). On the one hand, the SSI is determined by a pixel's contributing area, slope and a CN‐derived runoff depth estimate based on land use and soil type. On the other hand, the CSI is based on the estimated pixel soil clay content. We applied the GHI index at both the continental scale of Africa and at the local scale in two small (<10 km2) catchments in the Ethiopian highlands, using state‐of‐the‐art, high‐resolution Geographic Information System (GIS) data layers, and tested the ability of the GHI index to distinguish gully heads from non‐gully heads based on extensive datasets of mapped gully locations. Results show that the GHI index reasonably distinguishes pixels with and without gully heads across different scales, with area under the curve (AUC) values of 0.67 and 0.65 for the continental and local scale, respectively. The GHI index offers a conceptually sound description of gully initiation conditions, has low data requirements and requires no calibration, suggesting its potential to simulate gully erosion in more process‐oriented ways. However, its performance is clearly lower than data‐driven approaches that empirically relate gully occurrence to environmental variables derived from similar GIS layers (AUC of 0.83 at continental scale, 0.73 at local scale). We discuss possible reasons for this performance gap, such as the limited ability of the CN method to accurately simulate contrasts in runoff production and the high sensitivity to error propagation inherent to such a process‐oriented approach, and explore future improvement avenues.

Infiltration is an essential process in watershed hydrology. The curve number (CN) method has been widely used to calculate watershed infiltration for a given rainfall input but does not consider steady infiltration. This study develops a modified CN (MCN) method that included a term for the steady infiltration amount (Fc). Observed rainfall–runoff data for 14 rainfall events from a typical small watershed on the Loess Plateau of China were used to derive the watershed final infiltration rate (fc). Watershed infiltration after runoff initiation was then calculated by both the MCN and CN methods using initial abstraction values that were either observed (Ia−obs) or calculated (Ia−0.2 S) based on the calibrated fc. Three criteria [relative error (Er), model efficiency coefficient (E), and root mean square error (RMSE)] and visual assessments of graphed results were used to evaluate the methods’ simulation performances. The MCN method generally outperformed the CN method when using either of the ini...