A Modified Nearest Neighbors Approach for Estimating Missing Wind Data

Wind data is important for hydrodynamics and sediment transport modeling in a lake system because wind is the major driving force for most lakes. However, there is missing data caused by instrumental failure due to birds, thunderstorms, or other unexpected events. An estimation of these missing values becomes an important issue, since missing data affect the accuracy of the modeling results. Wind data collected in Lake Okeechobee is used as a test database for this study. This paper introduces an estimation model that includes statistical approaches coupled with multi-variable time lag analysis and a modified nearest neighbors approach to solve the missing data problems. The synthesized wind speeds for incomplete data are smoothed by a moving average and then compared to the actual measured values. The developed model demonstrates its ability to reproduce accurate wind speed for the years 1996 to 1999.

Keywords:

• Missing Data
• Nearest Neighbors
• Wind
• Time Series
• Estimating
• Data Transformation