A new method to improve the accuracy of remotely sensed data for wetland water balance estimates

Abstract

Study region: Thirlmere Lakes National Park, New South Wales, Australia Study focus: Uncertainties in water balance calculations can arise from errors associated with each of the budget input terms: precipitation, evapotranspiration, and inflows/outflows. However, uncertainties associated with the accuracy of the surface storage calculation have seldom been the focus of previous water balance studies. Digital elevation models (DEMs) used in water balance studies typically rely on bathymetric/topographic surveys, with remote-sensing techniques including satellite imaging processing, Light-Detection-and-Ranging (LiDAR), and unmanned-aerial-vehicle photogrammetry. This study investigates the vertical errors in bathymetric DEMs obtained from various remote-sensing techniques and its implication on water balance estimates in an intermittent wetland under drying conditions with vegetation encroachment. New hydrological insights: When bathymetry data obtained from different remote-sensing survey methods were adopted to calculate the water balance of a lake, variations in the model-predicted levels were attributed to the poor quality of photogrammetric DEMs. To improve the photogrammetric data, a new ground-filtering approach is developed, which reduces vertical errors induced by vegetation interference. The correlation (R2) of the DEMs, as compared to ground-truthed elevations, was improved from 0.5 before ground filtering to 0.9 after ground filtering. Using the ground-filtered DEM in the water balance calculation, a 70 % improvement was achieved in the water balance residuals. As such, uncertainties in lake and wetland bathymetry should be assessed in future water balance studies.

Keywords

• Remote sensing surveys
• Water balance model
• Digital elevation models
• Intermittent wetlands