Comparison of Filtering Techniques for Water Level Estimation Using GEDI

Abstract. Global Ecosystem Dynamics Investigation (GEDI) satellite-based LiDAR altimeter data can used to determine water levels. However, GEDI LiDAR beams may contain errors due to factors such as atmospheric conditions and topographic features. Therefore, it is essential to filter out data that contains errors and does not meet the required conditions in water level estimation. In this study, two different filtering techniques have been used for estimation of water levels. The first technique utilizes auxiliary data such as quality flag, sensitivity, and solar elevation, which are provided together with the GEDI data. On the other hand, the second technique utilizes the Interquartile Range (IQR) method. The Burdur Lake, a Ramsar site located in the Southwest of the Turkey, was selected as a test area. Daytime data from 18 October 2019 and nighttime data from 14 January 2020 were downloaded for the Burdur Lake. The Root Mean Square Error (RMSE) for the daytime data was calculated as 19.295 m. After filtering the data with auxiliary attributes, the RMSE value is decreased to 0.258 m. After applying the IQR filtering technique, the RMSE value was obtained as 0.317 m. For the nighttime data, the RMSE value was calculated as 0.292 m before filtering. The auxiliary data filtering was decreased the RMSE value into 0.118 meters. As a result of filtering with the IQR method, the RMSE value was obtained as 0.266 m. It was concluded that filtering is necessary to estimate water levels with GEDI data, especially for daytime data.