Comparing GPS-Constrained Forward and Inverse Models of Volume Change at The Geysers , CA

The Geysers geothermal field in northern California has seen subsidence, attributed to net volume loss during power production, since at least the 1960s. Over the last three decades this has been accompanied by reductions in reservoir steam pressure and power generation. To combat these effects, wastewater has been injected in the field since 1997. In order to better understand the effects of variations in production and wastewater injection on geothermal reservoir volume and surface subsidence over time, we installed two continuously-recording GPS stations (TG01 and TG02) in the northern Geysers in 2012 and one in the southern Geysers (TG03) in 2013. We present here our first analyses of the continuous GPS data. Combining these data with data from seven PBO GPS stations in the region, we first use common-mode filtering to remove any regionally-correlated seasonal noise from our GPS time series. Both TG01 and TG02 show early periods of uplift and later subsidence while TG03 shows ongoing subsidence. Next, we downsample steam extraction and injection data onto a rectangular grid and calculate ‘observed’ monthly volume changes as a function of position. We then use these to drive a forward elastic dislocation model to predict surface deformation changes each month in The Geysers field. This forward model overpredicts subsidence at all three Geysers GPS locations. We then compare the observed volume changes with inverse elastic dislocation models of the volume changes required to reproduce the GPS time series. We find that our GPS data do not have sufficient spatial resolution to reproduce the variability in reservoir volume change from reported well data. However, the estimates of total volume change from the inverse models and