Reconstruction of bore hydrograph trends in fractured rock aquifers using data minig techniques

Abstract

The lack of recorded groundwater hydrographs presents a problem for many ground-water studies. This is especially the case in the areas of fractured rock basement. In the New South Wales (NSW) portion of the Murray-Darling Basin (MDB), Australia, long-term hydrograph trends were required to separate groundwater components of climate forcing from the effects of deforestation on groundwater rise. To overcome the problem of missing long-term hydrographs, the data-mining technique was developed to enable reconstruction of aquifer behaviour from independent point data acquired at different times. The multi-decadal descriptive data-mining methodology described here uses independent Standing Water Level data (SWL – depth to water table), recorded in the groundwater database after borehole construction. Data are restricted to the Palaeozoic and early Mesozoic fractured rock aquifers closest to the land-surface. Catchments with pronounced rainfall gradients are divided to preserve homogeneity and avoid spatial-temporal confounding. Annual SWL time-series are created using a median function. The noise is filtered by applying 21-year moving average. The methodology is demonstrated on the Namoi Catchment. Results are validated using existing short-term hydrographs and long-term proxy datasets: rainfall trends. Evaluation is assessed using Root Mean Square Error method. It was validated on the Lachlan and Murrumbidgee catchments, where historic hydrographs exist. The approach described here has been successful in recreating past multi-decadal groundwater level trends in the fractured rock areas in eastern NSW. Aquifers rapidly responded to wetter climatic conditions and equilibrated within two to seventeen years. The influences of changes in land-use were detected in the Lachlan West section, where wheat-growing replaced pastoralism, and was separated from climatic drivers. Everywhere else the climate was the main driver of groundwater rise and was directly implicated in triggering dryland salinity. This observation helps clarify the mechanisms of dryland salinity occurrence. The methodology for reconstruction of aquifer behaviour has obvious applicability in many other parts of the world where only basic bore completion data exist,but long-term monitoring has not been carried out. In particular it can be applied to the basement areas in Africa, Australia and India.