Simultaneous fitting of drawdowns induced by pumping at multiple sites in an area of complex geology revealed important patterns of connected permeability
Over a six month period in July 2014, BHP conducted six pumping tests of between 5 and 11 days duration near the site of its OB31 pit in the Pilbara region of Western Australia. The pit was not operating at the time. Drawdowns and recoveries were monitored in 21 observation bores. Through interpretation of data acquired during these tests, it was hoped that insights could be gained into stratigraphic and structural controls on flow of groundwater in the area. Of particular importance at the time were assessments of dewatering rate requirements for the proposed OB31 iron ore pit.
The geology of the area is complex. It is comprised of steeply dipping beds of very different hydraulic properties. These are intersected by faults of unknown hydraulic significance. Despite their short duration, the complex patterns of drawdown induced by these tests reflected some of this heterogeneity. Could detailed fitting of temporal and spatial nuances of these drawdowns and recoveries using a model yield a map of subsurface hydraulic properties?
The answer is, of course, “no”. However, it was hoped that data acquired through these tests could nevertheless reveal something about the nature and disposition of aspects of the subsurface that are of relevance to future pumping. In particular, it was hoped that they would support workable estimates of pumping rates required to maintain dryness of an expanding and deepening pit.
This worked example report describes the process through which modelling was employed to extract information from the pumping test dataset. MODFLOW-USG was used for simulation. Efficient assimilation of pumping-test data was effected using randomized Jacobian functionality of PEST_HP. Post-history-matching linear analysis was used to explore uncertainties associated with inferences of subsurface hydraulic property heterogeneity.
Design and implementation of the modelling process recognised data assimilation as its primary task. This required the assignment of many parameters to a simplified, one-layer model. Interpretation of the patterns of hydraulic property heterogeneity that emerged in this layer through good model replication of observed drawdowns and recoveries required recognition of the somewhat abstract nature of these patterns. However this did not detract from their importance, nor from their relevance to predictions of future groundwater movement when the system will be subjected to much greater stresses than those that prevailed during the six pumping tests.