Uncertainty assessment of spatial-scale groundwater recharge estimated from unsaturated flow modelling | Evaluation de l’incertitude de la recharge des eaux souterraines à l’échelle spatiale estimée à partir de la modélisation de l’écoulement en zone non saturée Evaluación de la incertidumbre de la recarga de agua subterránea a escala espacial estimada a partir del modelado del flujo no saturado 根据非饱和水流模拟估算的空间尺度地下水补给不确定性评价 Avaliação de incerteza da recarga de águas subterrâneas em escala espacial estimadas por modelagem de fluxo em zona não saturada

Parameterisation of unsaturated flow models for estimating spatial-scale groundwater recharge is usually reliant on expert knowledge or best-estimated parameters rather than robust uncertainty analysis. This study chose the Campaspe catchment in southeastern Australia as a field example and examined the uncertainty of spatial groundwater recharge by performing uncertainty analysis. The study area was first divided into 13 zones according to different vegetation types, soil groups and precipitation. Individual models were then established for these zones using the biophysically based modelling code WAVES (Water Atmosphere Vegetation Energy and Solutes), which is capable of simulating unsaturated flow. The Monte Carlo method, together with the Latin-Hypercube sampling technique, was employed to perform uncertainty analysis by comparing modelled monthly evapotranspiration (ET) to MODIS ET. The results show that the common one-estimate-per-site approach can still identify the spatial pattern of groundwater recharge in the study area due to the presence of a precipitation pattern. In comparison, the uncertainty analysis not only identifies the spatial pattern, but also provides confidence levels in groundwater recharge that are critical for water resources management. The results also show that recharge absolute uncertainty is directly proportional to the amount of water input, but relative uncertainty in recharge is not. This study indicates that spatial recharge estimation without model calibration or knowledge of model uncertainty could be highly uncertain. MODIS ET can be used to reduce recharge uncertainty, but it is unlikely to lower the recharge uncertainty by a large extent because of the MODIS ET estimation error.