In 2014, the U.S. Geological Survey, in cooperation with the U.S. Department of Defense’s Strategic Environmental Research and Development Program, initiated a project to evaluate the potential impacts of projected climate-change on Department of Defense installations that rely on Guam’s water resources. A major task of that project was to develop a watershed model of southern Guam and a water-balance model for the Fena Valley Reservoir. The southern Guam watershed model provides a physically based tool to estimate surface-water availability in southern Guam. The U.S. Geological Survey’s Precipitation Runoff Modeling System, PRMS-IV, was used to construct the watershed model. The PRMS-IV code simulates different parts of the hydrologic cycle based on a set of user-defined modules. The southern Guam watershed model was constructed by updating a watershed model for the Fena Valley watersheds, and expanding the modeled area to include all of southern Guam. The Fena Valley watershed model was combined with a previously developed, but recently updated and recalibrated Fena Valley Reservoir water-balance model.

Two important surface-water resources for the U.S. Navy and the citizens of Guam were modeled in this study; the extended model now includes the Ugum River watershed and improves upon the previous model of the Fena Valley watersheds. Surface water from the Ugum River watershed is diverted and treated for drinking water, and the Fena Valley watersheds feed the largest surface-water reservoir on Guam. The southern Guam watershed model performed “very good,” according to the criteria of Moriasi and others (2007), in the Ugum River watershed above Talofofo Falls with monthly Nash-Sutcliffe efficiency statistic values of 0.97 for the calibration period and 0.93 for the verification period (a value of 1.0 represents perfect model fit). In the Fena Valley watershed, monthly simulated streamflow volumes from the watershed model compared reasonably well with the measured values for the gaging stations on the Almagosa, Maulap, and Imong Rivers—tributaries to the Fena Valley Reservoir—with Nash-Sutcliffe efficiency values of 0.87 or higher. The southern Guam watershed model simulated the total volume of the critical dry season (January to May) streamflow for the entire simulation period within –0.54 percent at the Almagosa River, within 6.39 percent at the Maulap River, and within 6.06 percent at the Imong River.

The recalibrated water-balance model of the Fena Valley Reservoir generally simulated monthly reservoir storage volume with reasonable accuracy. For the calibration and verification periods, errors in end-of-month reservoir-storage volume ranged from 6.04 percent (284.6 acre-feet or 92.7 million gallons) to –5.70 percent (–240.8 acre-feet or –78.5 million gallons). Monthly simulation bias ranged from –0.48 percent for the calibration period to 0.87 percent for the verification period; relative error ranged from –0.60 to 0.88 percent for the calibration and verification periods, respectively. The small bias indicated that the model did not consistently overestimate or underestimate reservoir storage volume.

In the entirety of southern Guam, the watershed model has a “satisfactory” to “very good” rating when simulating monthly mean streamflow for all but one of the gaged watersheds during the verification period. The southern Guam watershed model uses a more sophisticated climate-distribution scheme than the older model to make use of the sparse climate data, as well as includes updated land-cover parameters and the capability to simulate closed depression areas.

The new Fena Valley Reservoir water-balance model is useful as an updated tool to forecast short-term changes in the surface-water resources of Guam. Furthermore, the now spatially complete southern Guam watershed model can be used to evaluate changes in streamflow and recharge owing to climate or land-cover changes. These are substantial improvements to the previous models of the Fena Valley watershed and Reservoir. Datasets associated with this report are available as a U.S. Geological Survey data release (Rosa and Hay, 2017; DOI:10.5066/F7HH6HV4).