Terrain modelling to derive a groundwater surface for NSW upland areas using the SRTM-S adaptively smoothed vegetation offset DEM

Abstract

Groundwater is a critical component of the hydrological cycle supporting and maintaining ecosystems, whether through direct access or indirectly through stream baseflow contributions. With growing pressures on water resources in NSW, more knowledge is needed in terms of depth of the watertable in the landscape. Currently NSW has large scale interpreted contour groundwater maps and spatially modelled surfaces based on dryland salinity data. This paper aims to describe the first stages in developing a more spatially explicit surface for groundwater, with improvements on predictions of groundwater depth. The study will encompass the coast, tablelands and ranges of NSW. The SRTM data used was the adaptively smoothed DEM with vegetation offsets removed (DEM-S) that was resampled to 50m for the purposes of computational efficiencies (DEM-S 50). The Multi resolution Valley Bottom Floor index was derived for the study area on a region by region basis. This layer was normalized, inverted and combined with the DEM-S 50. Both layers were then combined through direct multiplication. The result was a derived surface where MRVBF dominates in the large flat areas and the elevation dominates the hillslopes. The logic being that hillslope and alluvial hydrological processes are represented within the one surface. To develop a groundwater surface all groundwater bores with construction and water level information from each major river system catchment were used in a linear regression of a modelled surface fit to observed groundwater levels. The groundwater data available was from different years and tended to be biased towards alluvial water extraction areas. Given the large spatial area being involved the groundwater data was used to get the widest possible spatial coverage rather than selectively filtering the data to one water level sampling period as the number of samples than become to small. Once a fit to the groundwater data was created this was applied back to the modelled surface. To validate the model four sub catchments (∼ 400 - 2000km2) spread across NSW were used. They had groundwater data throughout the catchment, captured within one sampling period. The validations showed R 2 = 0.27 - 0.44. The FLAG UPNESS index is a published method used for determining implied groundwater surfaces and was used as a comparison. The UPNESS index achieved R 2 = 0.08 - 0.12 for the validation of large regional catchments. To improve the R 2 further, geology, landuse and regolith cover relating to hydrology characteristics are planned as the next logical step. This study is critical in terms of the issues it aims to address but the groundwater data available to calibrate a modelled surface is the most restricting factor that not only impacts calibration but also validation confidence. Presently remote sensing products are being developed at the national and state scale to define areas in the landscape where vegetation is using stored water either from the soil profile or the water table. These products could provide valuable surrogate information in the future for calibration and validation of groundwater models to complement the sparse groundwater level information that exists.