Machine learning models are now capable of delivering coveted digital soil mapping (DSM) benefits (e.g., field capacity (FC) prediction); therefore, determining the optimal sample sites and sample size is essential to maximize the training efficacy. We solve this with a novel optimal sampling algorithm that allows the authentic augmentation of insufficient soil features using machine learning predictive uncertainty. Nine hundred and fifty-three forest soil samples and geographically referenced forest information were used to develop predictive models, and FCs in South Korea were estimated with six predictor set hierarchies. Random forest and gradient boosting models were used for estimation since tree-based models had better predictive performance than other machine learning algorithms. There was a significant relationship between model predictive uncertainties and training data distribution, where higher uncertainties were distributed in the data scarcity area. Further, we confirmed that the predictive uncertainties decreased when additional sample sites were added to the training data. Environmental covariate information of each grid cell in South Korea was then used to select the sampling sites. Optimal sites were coordinated at the cell having the highest predictive uncertainty, and the sample size was determined using the predictable rate. This intuitive method can be generalized to improve global DSM.

Changes in water quality were examined during selected rainfall events in a headwater stream severely damaged by a forest fire on 21–23 February 2021. Seven water quality parameters were analyzed: pH, electrical conductivity, biochemical oxygen demand, chemical oxygen demand, total suspended solids (TSS), total nitrogen, and total phosphorous. First-flush effect and event mean concentration analyses were conducted in spring (dry season) and summer (wet season) immediately after the forest fire. In spring, the non-biodegradable organic matter concentrations increased along with pollutants related to ash-enhanced soil water repellency in water affected by first-flush effects. In summer, TSS and nutrient concentrations increased along with pollutants related to fire-induced soil surface disturbances after a series of rainfall events. First-flush analyses indicated that cumulative pollutant loads were greater in the spring than in the summer due to a higher concentration of forest fire ash in the early storm runoff immediately after the forest fire. The event mean concentrations revealed that pollutant loads were associated with both forest fire ash and storm events in the immediate post-fire period. Therefore, this study indicated that water quality along a headwater stream is affected by short-term effects of large and intense forest fires and rainfall seasonality.

The severity and incidence of flash floods are increasing in forested regions, causing significant harm to residents and the environment. Consequently, accurate estimation of flood peaks is crucial. As conventional physically based prediction models reflect the traits of only a small number of areas, applying them in ungauged catchments is challenging. The interrelationship between catchment characteristics and flood features to estimate flood peaks in ungauged areas remains underexplored, and evaluation standards for the appropriate number of flood events to include during data collection to ensure effective flood peak prediction have not been established. Therefore, we developed a machine-learning predictive model for flood peaks in ungauged areas and determined the minimum number of flood events required for effective prediction. We employed rainfall-runoff data and catchment characteristics for estimating flood peaks. The applicability of the machine learning model for ungauged areas was confirmed by the high predictive performance. Even with the addition of rainfall-runoff data from ungauged areas, the predictive performance did not significantly improve when sufficient flood data were used as input data. This criterion could facilitate the determination of the minimum number of flood events for developing adequate flood peak predictive models.

Accurate modeling of flood flow hydrographs for small forested catchments in steep mountainous terrain is challenging because of large errors in the estimation of response time using existing empirical equations. The time of concentration (T_C) for a catchment is a widely used time parameter for estimating peak discharges in hydrological designs. In this study, we developed an estimated T_C using readily available mountain catchment variables, a small catchment, steep slope, and narrow valley, using empirical equations. For our approach, we used directly measured data from 39 forested catchments (area: 0.02–9.69 km²) during 3648 observed rainfall events over a 10-year observation period. Based on the uncertainties inherent in the empirical equation, the estimated T_C values were compared and analyzed through multiple regression and two different modified empirical modelling equations using our observed catchment parameters. The mean T_C was significantly correlated with catchment size and stream length but negatively correlated with stream slope (p < 0.01). As a result, the mean T_C estimated using the three modelling equations with catchment variables was qualitatively similar and had relative differences ranging from −12.5 to 15.5 min (−49 to 56%). Therefore, the models (particularly modeling equations with multiple regression, a modified empirical formula, and modified SCS Lag) can efficiently determine the T_C and can be used in any small forested catchment in steep mountainous terrain.

Event flow characteristics were evaluated based on temperature and level of stream water in 22 forested catchments (area: 13.2–281.4 ha) to investigate sustainable flood management measures. Temperature and stream water levels were during 346 rainfall events in the summer season (July–September) from 2020 to 2022. Rising stream water levels responded to falling stream water temperature between ≤100 and >100 ha forested catchments in two types of time of concentration. Stream water temperature decreased by 3.0 °C when the stream water level increased by up to 0.9 m during rainfall events. Falling stream water temperature at two types of time of concentration was negatively correlated with total precipitation and rising stream water level. Based on the relatively high value of regression and cumulative frequency distribution, the estimated rising stream water level was appropriate in small catchments (≤100 ha) when the stream water temperature decreased, and the stream water level increased during rainfall events. Rising stream water levels and falling stream water temperatures are responses to catchment-scale effects, which are influenced by the nature and rapidity of the hydrological responses. Therefore, the results of the present study indicate that spatial and temporal differences in thermal responses of stream water temperature to water levels were controlled by catchment-scale effects under rapidly changing rainfall.