John Jairo Ramirez-Avila

The operation of multipurpose reservoirs for which flood risk management (FRM) is a primary or key purpose may be improved by the use of Forecast-Informed Reservoir Operations (FIRO). The ability to screen a portfolio of dams to assess FIRO viability allows for more-efficient use of time and funding than individual detailed analyses for several dams. In this study, screening criteria were used to evaluate the potential effectiveness of FIRO prestore and prerelease (PS-PR) and basic operation prerelease (BO-PR) methods using a large data set (58 dams) and along three lines of inquiry. A regression model was used to assess system characteristics and identify the reservoirs for which reservoir filling is most sensitive to changes in flood storage capacity. Historic flood events provided insight into the feasibility of FIRO based on seasonality and storm characteristics. Physical and policy constraints were studied to assess the feasibility of FIRO at or near top of conservation pool and to provide an initial input to the screening framework. A four-parameter equation to predict reservoir-filling frequency had an R 2 value of 0.67, and flood storage capacity was significant at a 99% confidence level. Review of historic reservoir-filling events showed a median 3-day flood storage of 23% and a 7-day flood storage of 38%. Policy and physical release constraints yielded maximum daily drawdown estimates for each reservoir and indicated a median of 2% of average annual inflow evacuation per day. A case study of five dams was used to assess viability of using the storage-based regression as a criterion for prerelease screening. The presented screening approach utilizes an estimate of maximum prerelease drawdown and the regression equations to provide high estimates for potential FIRO effectiveness (with uncertainty); actual applicability depends on forecast skill and available lead time, which are not evaluated in this study.

There has been a longstanding debate about the initial abstraction ratio ( λ = I a / S ), the ratio of initial abstraction ( I a ) and potential maximum retention ( S ). It remains unclear what value of λ best fits the National Resources Conservation Service (NRCS) rainfall runoff equation. In this study, we reviewed the λ value based on the observed data of 36 watersheds in the US to determine the range of empirically based λ values. We also conducted a sensitivity analysis of the change in runoff per unit initial runoff ( d Q / Q ), considering the ratio of various λ values (0.05, 0.1, 0.2, and 0.3). Because there was no noticeable functional relationship of change in runoff with standalone factors such as total rainfall depth ( P ) and S , we introduced a dimensionless precipitation ratio ( P ∗ = P / S ). Our analysis suggested that Q is more sensitive at the lower ranges of P ∗ regardless of λ . In general, a change in d Q / Q for every change in P ∗ would be more sensitive to higher λ . However, for a higher P ∗ , it was not sensitive to λ . Similarly, we investigated the ratio of change in Q with curve number C N ( d Q / d C N ) for various P ∗ and λ . Our analysis suggested that the d Q / d C N was sensitive to the lower values of P ∗ and higher values of λ . However, d Q / d C N was less sensitive to higher P ∗ and generally true regardless of λ . We also investigated the change in Q with the change in P ( d Q / d P ) for various P ∗ . This was also more sensitive for smaller P ∗ , but, in contrast to d Q / Q , it was more sensitive to smaller λ . In general, results indicate that runoff sensitivity is highest under low precipitation and low C N conditions, and that λ selection becomes less critical when C N values are high, especially after P ∗ > 3.5 . Also, the selection of λ is more critical for smaller C N than for smaller P . The λ selection is least sensitive to fully developed areas and regions with higher P .

Tropical coastal regions exhibit pronounced spatial and temporal variability in rainfall driven by seasonal atmospheric circulation and coastal–orographic interactions. Accurate representation of the temporal distribution of rainfall is essential for hydrologic modeling and infrastructure design. This study develops regionalized Huff curves for the Department of Magdalena, Colombia, addressing a critical gap in the characterization of rainfall temporal patterns in tropical coastal regions. A total of 270 short-duration (5–6 h) rainfall events from automatic stations were converted into normalized cumulative mass curves. The resulting curves were grouped into homogeneous temporal patterns using clustering algorithms. Three dominant storm types were identified: early-peak (Curve 1), intermediate (Curve 2), and uniform (Curve 3), reflecting the region’s coastal, lowland, and orographic influences. Probability envelopes and representative design hyetographs were derived to quantify intra-event variability. Rainfall–runoff simulations for a 100-km² watershed showed peak-flow differences of up to 132% between storm types, highlighting the sensitivity of hydrologic response to rainfall temporal distributions. The resulting regionalized Huff curves provide a practical and transferable framework for hydrologic modeling, flood-risk assessment, and infrastructure planning in tropical regions with limited high-resolution rainfall data.

The analysis of the hydrological response of watersheds has made the use of the Curve Number methodology relevant. However, the global application of this method has presented limitations mainly associated with the availability of land use and hydrology information for the study areas. The Global Curve Number map (GNC) is a tool that defines the value of the curve number of a region with a low spatial resolution (250 m) and its validation is necessary in the development of research related to water resource management in watersheds with limited information. The present study determined the representative curve number value in three micro-watersheds contributing to the Pisque River in North Ecuador. Curve number values proposed by the GNC for these micro-watersheds were compared with curve number values obtained for land use scenarios for the years 1986 and 2013, generated through the analysis of maps derived from satellite images. The geographic land use and vegetation cover was obtained through principal component analysis and validated using the confusion matrix method and the kappa index. The proposed land cover classification had a high accuracy (82 - 84 %) when compared with secondary reference information, which represents high reliability in the curve number maps generated for the study area. The models developed for the three micro-watersheds present greater detail and spatial resolution, maintaining an excellent fit with the CNG which is reflected in the goodness of fit statistics (NSE = 0.96 - 0.98, PBIAS = 13.06 - 19.03 and RMSE = 18.94 - 22.21).

The spatial distribution and data quality of curve number (CN) values determine the performance of hydrological estimations. However, existing CN datasets are constrained by universal-applicability hypothesis, medium resolution, and imbalance between specificity CN tables to generalized land use/land cover (LULC) maps, which hinder their applicability and predictive accuracy. A new annual CN dataset named CUSCN30, featuring an enhanced resolution of 30 meters and accounting for temporal variations in climate and LULC in the continental United States (CONUS) between 2008 and 2021, was developed in this study. CUSCN30 demonstrated good performance in surface runoff estimation using CN method when compared to observed surface runoff for the selected watersheds. Compared with existing CN datasets, CUSCN30 exhibits the highest accuracy in runoff estimation for both normal and extreme rainfall events. In addition, CUSCN30, with its high spatial resolution, better captures the spatial heterogeneity of watersheds. This developed CN dataset can be used as input for hydrological models or machine learning algorithms to simulate rainfall-runoff across multiple spatiotemporal scales.