Ground truthing groundwater-recharge estimates derived from remotely sensed evapotranspiration: a case in South Australia

Using a water balance to estimate groundwater recharge through the use of remotely sensed evapotranspiration offers a spatial and temporal density of data that other techniques cannot match. However, the estimates are uncertain and therefore ground truthing of the recharge estimates is necessary. This study, conducted in the south-east of South Australia, demonstrated that the raw water-balance estimates of recharge had a negative bias of 45 mm/yr when compared to 190 recharge estimates using the water-table fluctuation method over a 10-year period (2001-2010). As this bias was not related to the magnitude of the recharge estimated using the water-table fluctuation method, a simple offset was used to bias-correct the water-balance recharge estimates. The bias-corrected recharge estimates had a mean residual that was not significantly different from an independent set of 99 historical recharge estimates but did have a large mean absolute residual indicating a lack of precision. The value in this technique is the density of the data (250-m grid over 29,000 km(2)). The relationship between the water-table depth and net recharge under different vegetation types was investigated. Under pastures, there was no relationship with water-table depth, as the shallow roots do not intercept groundwater. However, under plantation forestry, there was a relationship between net recharge and water-table depth. Net recharge under plantation forestry growing on sandy soils was independent of the water table at around 6 m depth but, under heavier textured soils, the trees were using groundwater from depths of more than 20 m.