# Trend analysis of dam inflow data using the trend accuracy index and the potential-evapotranspiration correction factor

## TL;DR

This paper introduces a new method for simulating dam inflow trends using a trend accuracy index and a potential-evapotranspiration correction factor to improve long-term water resource management.

## Contribution

The novel contribution is the use of the trend accuracy index (TAI) and potential-evapotranspiration correction factor (PET-CF) for optimizing dam inflow simulations.

## Key findings

- TAI + PET-CF achieves 76.4% concordance with observed inflow trends, outperforming NSE-based calibration.
- Rainfall and inflow trends are not consistently aligned across time scales in historical and future periods.
- Three clusters capture spatial differences in rainfall linked to station location and elevation.

## Abstract

Climate change intensifies rainfall and runoff variability, complicating long-term dam inflow simulation. Long-term runoff models often optimize parameters using error-based metrics such as NSE or RMSE, which reduce deviations but may fail to reproduce long-term trends. We optimize the four-tank model using the trend accuracy index (TAI) and the potential-evapotranspiration correction factor (PET-CF) to better capture inflow trends. We validate the approach by assessing trends in simulated inflows. We evaluate monthly, seasonal, and annual rainfall trends at 101 stations, examine their consistency with inflow trends, and apply K-means + + clustering. Using rainfall in the Nakdong river basin and inflow at Hapcheon Dam, we conduct clustering and trend analyses for 2000–2019 and two future periods (2021–2050, 2051–2100) under SSP2-4.5 and SSP3-7.0. Three clusters capture spatial differences in rainfall linked to station location and elevation. Across time scales, rainfall and inflow trends are not consistently aligned in either the historical or future periods. In dam inflow simulation, TAI + PET-CF achieves the highest concordance with observed inflow trends (76.4%), outperforming NSE-based calibration. Combined with rainfall trend and clustering information, the framework can support efficient water-resources management.

The online version contains supplementary material available at 10.1038/s41598-026-40225-y.