Groundwater Model Development

GMS, a groundwater modeling system developed by the department of defense couples a GIS (geographic information system) pre and post processor with MODFLOW.  To use this package data is transformed into spatially based coverages.  Individual coverages were created for each input or output data type.  Surface elevations for the San Juan watershed were determined from digital elevation models (DEMs) obtained from Instituto Nacional De Estadistica Geografia E Informatica (Figure 2).  This electronic raster data was at 30 meter resolution.  Cell size is a uniform 500 x 500 meters.

For the purposes of this study, the San Juan watershed was assumed to be underlain by a single unconfined aquifer of relative uniform thickness.  The 1986 UNAM study consider the aquifer to be confined, but continual withdrawal by municipal and agricultural wells was assumed to covert the aquifer from confined to unconfined.  Based on the transmissivity values estimated in the UNAM Study and the assumption of a uniform thickness of 300 meters, an initial distribution of hydraulic conductivity was estimated, and later modified by the model calibration process.  The calibrated distribution of hydraulic conductivity is shown in Figure 6.  Specific yields were 0.22 for the northern portion of the basin and 0.2 for the southern portion.

Underlying the San Juan aquifer is a series of geological faults (Umhoefer et al. 2001).  Current research by University of California, Berkeley and Niparaja (Personal communication, 2005) indicates that a fault dissecting the northern and southern portion of the basin may be acting as a barrier to groundwater flow (Figure 7).  At this time there is not enough data to incorporate a no-flow barrier into the model; however, this area was modeled with a lower hydraulic conductance to impede groundwater flow through this region.  Further research is required to resolve this issue. Well locations and types (agriculture, private or municipal) were determined and translated into an electronic coverage (UNAM, 1986).  Annual pumping rates associated with each well were estimated.  The total annual pumping determined from our model matched the agricultural and municipal pumping totals developed by CNA for the basin.   Climate data was obtained from Comision Nacional Del Agua. Servicio Meteorologic Nacional.  

To calculate the amount of recharge entering the aquifer, the sub-basins within the watershed were delineated from the digital elevation model and the total volume of water that fell within each sub-basin was determined for annual rainfall totals and storm data (Figure 8).  In semi-arid areas such as Loreto, only a small portion of this volume becomes recharge; most rainfall is lost to run-off and evaporation.  There are several well established methods of calculating the percent of rainfall that becomes recharge: the Anderson equation (Anderson et al. 1992) and the Maxey-Eakin (Maxey and Eakin, 1949) are both appropriate for this area and climate.  Both methods were used for comparison.  The Maxey-Eakin method consistently yielded higher values so was used to calculate recharge.   

In average or dry years - years with an average annual rainfall of 11.5 cm or less - rainfall is not sufficient to produce aquifer recharge.  Potential recharge in wet years was estimated by including the data from the larger 2-year, 5-year, 25-year, and 50-year storm events.  Since the model is annual the storm volume was then distributed over the event time frame (i.e. the storm volume from the 10 year storm was divided by 10 to achieve recharge per year).  The 2-year storm event yielded the highest possible amount of annual recharge at approximately 2.1 Mm³/yr. 

A basic water balance equation and rudimentary flow net analysis was used to estimate recharge in the 1986 UNAM report.  Given the sparse data available and high level of uncertainty, they estimated a recharge rate of 10 Mm³ per year, plus or minus an order of magnitude. This translates into a recharge rate that falls within a range of 1 Mm³ to 100 Mm³ per year.  The updated analysis carried out for this study uses improved data analysis tools and data that were not available at the time of the prior study and indicates that given the level of rainfall and the aridity of the area, a recharge of 10 Mm³ per year is not possible. 

To ascertain the effects of increased well pumping on the aquifer from the proposed urban development, projected water consumption rates for each of the growth scenarios were developed (Table 1).  Five projected populations for the year 2025 and four different scenarios were modeled.  For political reasons, both the high and low recharge estimates (10 and 2.1 Mm³ per year) were tested in different runs of the models. 

Model calibration depends upon comparing simulated groundwater elevations with measured groundwater elevations.  Informal measurement of current groundwater levels were obtained from local sources.  However, all requests for measured groundwater levels from formal sources were unavailable.  This is especially pertinent in the area of the municipal wells.  Better measurements of existing groundwater elevation would have improved model accuracy and resulted in more conclusive results.  

Initially, lower groundwater levels and more drawdown were simulated to match the limited number of recorded water measurements and the 2.5 meter decline reported in the 1986 study.  However, a presentation by CNA in Loreto on October 20, 2005 suggested higher existing water levels and less groundwater decline.  The models were subsequently adjusted.  It should be noted that the water levels being reported from field researchers and those being reported by CNA do not match.  While we have adjusted the groundwater model to more closely match CNA’s estimates, there is a high probability that it understates the amount of groundwater decline. This discrepancy needs to be resolved in future modeling efforts.