Toward a better understanding of curve number and initial abstraction ratio values from a large sample of watersheds perspective
Brandão, Abderraman R.Amorim ; Schwamback, Dimaghi LU
; Ballarin, André Simões
; Ramirez-Avila, John J.
; Vasconcelos Neto, José Goes
and Oliveira, Paulo Tarso S.
(2025)
In Journal of Hydrology
655.
- Abstract
The Natural Resources Conservation Service Curve Number method (NRCS-CN) is the most widely used approach to estimate runoff from rainfall events. However, some uncertainties in the method remain linked to the value of the standard initial abstraction ratio (λ) and discrepancies between computed and standard tabulated Curve Number (CN) values. Here, we compute CN values and investigate the effects of λ on runoff estimation performance using a large sample of 3,578 watersheds distributed across the globe. We evaluate the impact of the default λ value of 0.2 and the proposed value of 0.05 across three methods and examine two rainfall event thresholds for CN calibration. We further investigated the association between the parameters of the... (More)
The Natural Resources Conservation Service Curve Number method (NRCS-CN) is the most widely used approach to estimate runoff from rainfall events. However, some uncertainties in the method remain linked to the value of the standard initial abstraction ratio (λ) and discrepancies between computed and standard tabulated Curve Number (CN) values. Here, we compute CN values and investigate the effects of λ on runoff estimation performance using a large sample of 3,578 watersheds distributed across the globe. We evaluate the impact of the default λ value of 0.2 and the proposed value of 0.05 across three methods and examine two rainfall event thresholds for CN calibration. We further investigated the association between the parameters of the NRCS-CN method and the catchment and climatic characteristics of the watersheds using machine learning techniques. Our findings indicated that the Least-Squares (LS) method better calibrates CN values and performs more accurately using λ = 0.05 compared to the default λ = 0.2 and that the 25.4 mm precipitation threshold showed better performance for calibrating the methods. The CN spatial variability revealed that high values of CN are controlled by the spatial variation of slope, precipitation, and soil silt content, while lower CNs aligned with forest lands, and strongly correlated to regions of sandy soils, down to the aridity index. Land cover emerges as the most influential characteristic in determining the λ, with cropland percentage exhibiting the greater influence. Arid regions, increases cropland, urban areas, and soil sand content are associated with λ = 0.05, whereas higher pasture percentages correspond to λ = 0.2. We also provide equations for converting parameters from λ = 0.2 to λ = 0.05. Several watersheds worldwide are ungauged, and this study emphasizes the non-linear and complex nature of hydrological processes influencing the NRCS-CN method parameters. Our study provides a better understanding of the NRCS-CN method, bringing significant practical implications for various applications, including hydrological processes, stormwater management, flood forecasting, sediment management, hydrological modeling, and water resources engineering.
(Less)
- author
- Brandão, Abderraman R.Amorim
; Schwamback, Dimaghi
LU
; Ballarin, André Simões
; Ramirez-Avila, John J.
; Vasconcelos Neto, José Goes
and Oliveira, Paulo Tarso S.
- organization
- publishing date
- 2025-07
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- Dynamic runoff, Land cover, Machine Learning, Rainfall-runoff, SCS-CN (NRCS-CN), SHAP
- in
- Journal of Hydrology
- volume
- 655
- article number
- 132941
- publisher
- Elsevier
- external identifiers
-
- scopus:85218863012
- ISSN
- 0022-1694
- DOI
- 10.1016/j.jhydrol.2025.132941
- language
- English
- LU publication?
- yes
- id
- 6795fa50-669e-4d0d-9b2a-1d6e3deeb329
- date added to LUP
- 2025-06-09 11:33:00
- date last changed
- 2025-06-09 11:34:02
@article{6795fa50-669e-4d0d-9b2a-1d6e3deeb329,
abstract = {{<p>The Natural Resources Conservation Service Curve Number method (NRCS-CN) is the most widely used approach to estimate runoff from rainfall events. However, some uncertainties in the method remain linked to the value of the standard initial abstraction ratio (λ) and discrepancies between computed and standard tabulated Curve Number (CN) values. Here, we compute CN values and investigate the effects of λ on runoff estimation performance using a large sample of 3,578 watersheds distributed across the globe. We evaluate the impact of the default λ value of 0.2 and the proposed value of 0.05 across three methods and examine two rainfall event thresholds for CN calibration. We further investigated the association between the parameters of the NRCS-CN method and the catchment and climatic characteristics of the watersheds using machine learning techniques. Our findings indicated that the Least-Squares (LS) method better calibrates CN values and performs more accurately using λ = 0.05 compared to the default λ = 0.2 and that the 25.4 mm precipitation threshold showed better performance for calibrating the methods. The CN spatial variability revealed that high values of CN are controlled by the spatial variation of slope, precipitation, and soil silt content, while lower CNs aligned with forest lands, and strongly correlated to regions of sandy soils, down to the aridity index. Land cover emerges as the most influential characteristic in determining the λ, with cropland percentage exhibiting the greater influence. Arid regions, increases cropland, urban areas, and soil sand content are associated with λ = 0.05, whereas higher pasture percentages correspond to λ = 0.2. We also provide equations for converting parameters from λ = 0.2 to λ = 0.05. Several watersheds worldwide are ungauged, and this study emphasizes the non-linear and complex nature of hydrological processes influencing the NRCS-CN method parameters. Our study provides a better understanding of the NRCS-CN method, bringing significant practical implications for various applications, including hydrological processes, stormwater management, flood forecasting, sediment management, hydrological modeling, and water resources engineering.</p>}},
author = {{Brandão, Abderraman R.Amorim and Schwamback, Dimaghi and Ballarin, André Simões and Ramirez-Avila, John J. and Vasconcelos Neto, José Goes and Oliveira, Paulo Tarso S.}},
issn = {{0022-1694}},
keywords = {{Dynamic runoff; Land cover; Machine Learning; Rainfall-runoff; SCS-CN (NRCS-CN); SHAP}},
language = {{eng}},
publisher = {{Elsevier}},
series = {{Journal of Hydrology}},
title = {{Toward a better understanding of curve number and initial abstraction ratio values from a large sample of watersheds perspective}},
url = {{http://dx.doi.org/10.1016/j.jhydrol.2025.132941}},
doi = {{10.1016/j.jhydrol.2025.132941}},
volume = {{655}},
year = {{2025}},
}