2015 GSA Annual Meeting in Baltimore, Maryland, USA (1-4 November 2015)

Paper No. 113-1
Presentation Time: 9:00 AM-6:30 PM

WORKING WITH REAL DATA: GETTING ANALYTIC ELEMENT GROUNDWATER MODEL RESULTS TO FIT FIELD DATA


CONGDON, Roger D., USDA Forest Service, 333 Broadway Blvd SE, Albuquerque, NM 87102, rcongdon@fs.fed.us

Models of groundwater flow often work best when very little field data exist. In such cases, some knowledge of the depth to the water table, annual precipitation totals, and basic geological makeup is sufficient to produce a reasonable-looking and potentially useful model. However, in this case where a good deal of information is available regarding depth to bottom of a dune field aquifer, attempting to incorporate the data set into the model has variously resulted in convergence, failure to achieve target water level criteria, or complete failure to converge. The first model did not take the data set into consideration, but used general information that the aquifer was thinner in the north and thicker in the south. This model would run and produce apparently useful results. The first attempt at satisfying the data set; in this case 51 wells showing the bottom elevation of a Pacific coast sand dune aquifer, was to use the isopach interpretation of Robinson (OFR 73-241). Using inhomogeneities (areas of equal characteristics) delineated by Robinson’s isopach diagram did not enable an adequate fit to the water table lakes, and caused convergence problems when adding pumping wells. The second attempt was to use a Thiessen polygon approach, creating an aquifer thickness zone for each data point. The results for the non-pumping scenario were better, but run times were considerably greater. Also, there were frequent runs with non-convergence, especially when water supply wells were added. Non-convergence may be the result of the lake line-sinks crossing the polygon boundaries or proximity of pumping wells to inhomogeneity boundaries. The third approach was to merge adjacent polygons of similar depths; in this case within 5% of each other. The results and run times were better, but matching lake levels was not satisfactory. The fourth approach was to reduce the number of inhomogeneities to four, and to average the depth data over the inhomogeneity. The thicknesses were varied within 5% of the average until the lake levels were closely matched. This last methodology proved satisfactory and stable. The data were honored and the solver worked relatively quickly; thus preserving the simplicity and speed of the Analytic Element method; and various pumping scenarios were stable.
Handouts
  • 2015_GSA_poster.pdf (40.5 MB)