In this study a method for filling in missing data of river water temperature using a pre-constructed mathematical relationship between air and water temperatures is presented. A regression between water temperatures at individual stations and ambient air temperatures at nearby weather stations can provide a practical method for representing missing water temperature data for an entire region. Air and water temperature data that were collected from two test sites (one coastal and, one inland) were individually fitted to a nonlinear regression model. To consider seasonal hysteresis effects, separate functions were fitted to the data in the rising and falling limbs. A single-criterion, multi-parameter optimization technique was used to determine the optimal parameter sets. This method minimizes the differences between the time series of the measured and estimated data. The constructed air-water temperature relationship was subsequently applied to represent missing water temperature data. It was found that the RMSEs(MBEs) were in the range of 1.843―1.976℃(-0.329―0.201℃) and the coefficient of determination were in the range of 0.92―0.96. The results demonstrate that the predicted water temperatures using the regression equations were reasonably accurate.

Abstract
 1. Introduction
 2. Study region and data
 3. Mathematical models
 4. Results
  4.1. Validation of test site A
  4.2. Validation of test site B
 5. Summary and conclusions
 REFERENCES

• Khil-Ha Lee(Department of Civil Engineering, Daegu University) Corresponding author