Overview

        The New Mexico / Texas Water Commission is undertaking planning studies of the Rio Grande Project water delivery system below Elephant Butte Dam to assess the feasibility of providing a year-round supply of surface water for municipal and industrial purposes. This effort is called the Las Cruces - El Paso Sustainable Water Project. Because changes in the timing or routing of surface water releases will change the flow regime of the Rio Grande, simulation models are necessary to assess the implications of changes on the study area’s hydrologic conditions. The Commission employed an engineering consultant to develop a computerized simulation model to assess both flow and water quality in the Rio Grande Project area. The consultant developed a model that links a stream simulation model with a ground water model to conduct these analyses. The report that follows describes the structure of the model, the development of some of its key inputs and the approach used to calibrate the model.

New Mexico / Texas Water Commission

        Growing demands and competition for the Paso del Norte region’s limited water resources have resulted in conflicts between competing interests. Beginning in the 1970s, the city of El Paso began acquiring property in southern Doña Ana County to gain access to water rights in that portion of the Mesilla Bolson. Subsequent actions by the New Mexico State Engineer and special legislation adopted by the New Mexico Legislature effectively barred El Paso from being able to use ground water from the New Mexico portion of the Mesilla Bolson. El Paso mounted a legal challenge, and years of costly litigation followed. On March 6, 1991, a landmark settlement was reached under the New Mexico Court of Appeals with the signing of a Joint Settlement Agreement by all parties to the litigation. The Settlement Agreement marked the beginning of a new relationship where conflict was replaced by cooperation. Key conditions of this agreement are:

• the formation of a Joint Commission;
• establishing water conservation as first priority;
• using the region’s surface water for all purposes; and,
• performing water resource studies applicable to the entire project area.

        The Commission has assumed the responsibility for coordinating planning activities of both the Rio Grande Project water delivery system below Elephant Butte Dam and regional ground water supplies in southern New Mexico and the El Paso area. The stated goal of the Commission is to provide a sustainable water supply to users in the Rio Grande Project area through the conjunctive use of all available water resources. The Commission is examining interstate water delivery options insofar as the Rio Grande Project holds water for both Texas and New Mexico entities. However, the Commission is not engaged in changing water deliveries that are already legally established between Texas and New Mexico.

        Currently, the Commission is assessing various water conveyance options to deliver project water to users in El Paso and Doña Ana Counties. Of particular importance is to provide a year-round supply of surface water of a sufficient quality to enable its treatment and use for municipal and industrial purposes. This overall effort is called the Las Cruces - El Paso Sustainable Water Project. To facilitate its evaluation of conveyance options, the Commission employed the Boyle/Parsons Consultant Team to develop a computerized simulation model to assess flow and water quality in the Rio Grande Project area.

Conjunctive Water Resource Management Plan

        The Commission has undertaken several conceptual-level studies to identify water resource management strategies which best serve the water resource interests of the region. These studies were summarized in the Summary Report-Conjunctive Water Resource Management Plan. Conclusions of the Plan include:

• critical water supply shortfalls must be addressed by surface water;
• year-round surface water accessibility is required;
• water quality must be protected to meet drinking water standards;
• water quantity must be maximized through conservation; and,
• any excess flows should be used to recharge the two regional aquifers (Mesilla and Hueco Basins).

        Commission-directed studies defined a specific project, the Las Cruces - El Paso Sustainable Water Project as a first step toward satisfying the fundamental goals outlined by the Conjunctive Water Resource Management Plan. The area included in the Las Cruces - El Paso Sustainable Water Project consists of the Rio Grande Basin of southern New Mexico and far west Texas. The study area extends along the Rio Grande from Elephant Butte Dam on the north to Riverside Diversion Dam on the south.

        The primary water supply sources within the study area are the Mesilla and Hueco ground water basins and the Rio Grande. Surface waters within the study area are delivered by a Bureau of Reclamation project known as the Rio Grande Project. Water from the Rio Grande Project is allocated in accordance with the 1906 treaty between the United States and Mexico, a tri-state agreement known as the Rio Grande Compact, and Reclamation Law. Water is distributed to Mexico, the Elephant Butte Irrigation District, and El Paso County Water Improvement District Number One.

        The initial phase of the Las Cruces - El Paso Sustainable Water Project involves improvements and enlargements of the existing Rio Grande Project distribution system from Mesilla Dam in New Mexico to the American Dam in El Paso, Texas. Improvements contemplated include lining of selected canals, installation of new control structures, improved management information systems, conveyance pipelines, and one or more new water treatment plants to meet potable water demands in both New Mexico and Texas. Improvements may also include improved management and reclamation of drain flows which currently degrade the quality of the river, and underground storage of water during periods when supplies exceed demands to restore vital ground water basins and provide a degree of protection against future droughts. Because the Project has several alternatives of major significance, the Commission further divided the Project into phases to permit the identification and evaluation of alternatives prior to the commitment of financial resources for implementation.

Rio Grande System Hydrologic Modeling

        Changes in the timing or routing of surface water supplies contemplated under many of the Project alternatives will change the flow regime of the Rio Grande and the some aspects of the overall operations of the Rio Grande Project. Because the Rio Grande Project system is hydrologically and operationally complex, computer simulation models are being used to assess the implications of Project alternatives and to evaluate the effectiveness of the alternatives in meeting Project objectives. A streamflow simulation program was linked with a ground water simulation program to model the Rio Grande from Elephant Butte Reservoir in New Mexico to the Riverside Canal in Texas. The simulation model was further modified to track certain water quality parameters throughout the model using a mass balance approach. Finally, institutional and legal constraints were incorporated in the form of water right priorities, operating rules and allocation requirements mandated by the Rio Grande Compact.

        Analysis of the Project and evaluation of changes in hydrologic conditions within the study area associated with Project alternatives, are being accomplished using the combined results from a surface water model and ground water model. The surface water model must be supplied with information on interactions between the surface water system and the ground water system. These interactions occur mainly in the form of river gains from ground water and river losses to ground water, drain discharges, return flows from irrigation pumping, and river losses associated with irrigation pumping. Likewise, the ground water model must be supplied with information related to surface water conditions, mainly river and canal hydrology, in order to properly simulate ground water conditions.

        Previously, Boyle Engineering Corporation developed a general-purpose, data-driven water accounting and allocation simulation model known as Boyle Engineering Stream Simulation Model (BESTSM). Application of the model to the Rio Grande Project system required the collection, analysis and formatting of specific hydrologic data defining the physical and operational characteristics of the basin. BESTSM starts with streamflow as input to the system. The flow is modeled using a linked-node representation the key physical features of the system, such as dams, reservoirs, diversions, canals, and pipelines. The surface water system that was modeled begins at the San Marcial Gauge upstream of Elephant Butte Reservoir and ends at the Riverside Diversion Dam just downstream of El Paso, Texas.

        The ground water model domain includes the entire Mesilla Basin from the Leasburg Canal in New Mexico to just above the American Diversion Dam in Texas, while excluding the Rincon Valley to the north, and the Hueco Basin to the south and east. Ground water conditions in the Rincon Valley and the Hueco Basin, and their interaction with surface water resources, are approximated using analytical solutions incorporated directly in the surface water model.

        A different numerical model known as Modular Three-dimensional Finite-Difference Ground-Water Flow Model (MODFLOW) is used to analyze the movement and storage of ground water. MODFLOW uses the finite difference numerical method to approximate a solution to the ground water flow equation. MODFLOW was developed by the United States Geological Survey (USGS) and is in widespread use throughout the United States. The model employed in this study is a version of the original USGS model modified specifically for application to the Mesilla Bolson (Hamilton and Maddock, 1993. Application of a Ground-Water Flow Model to the Mesilla Basin, New Mexico and Texas). The Mesilla Bolson is the principal ground water component within the study area.

        A daily time-step is used for BESTSM while a seasonal time-step is employed in MODFLOW. Two seasons were used for calibration, an irrigation season (March through October) and a non-irrigation season (November through February). BESTSM and MODFLOW are linked such that information produced by one is passed to the other. This arrangement is believed to provide the best estimate of surface water / ground water interactions in the Mesilla Basin part of the study area, a critical reach of river in terms of Project alternatives.

        Water quality is an important consideration in this project, and the conservative elements of Total Dissolved Solids (TDS), sodium, chloride, and sulfate are simulated. Conservative elements do not precipitate or react, to an appreciable degree, in the aqueous system. This allows a simple mass balance approach to be used to simulate these water quality elements.

System Represented

        BESTSM represents the surface water system from the San Marcial gauge upstream of Elephant Butte Reservoir to the Riverside Diversion Dam downstream of El Paso, Texas. A linked-node approach is used to represent the system, with nodes being reservoirs, diversions, spillways, stream gauges, or any location where information is known or needed from the model. Links are canals, pipelines, the river, or anything which conveys water from node to node.

        Elephant Butte and Caballo Reservoirs, all of the diversion structures, all of the major drains, and most of the major spillways are represented in the model. While data are virtually non-existent at the spillways and inflows are typically not predicted by the model at these locations, they were included primarily to allow flow and quality conditions in the river to be queried at numerous locations. The selection of node locations correspond to spillway return points, thus providing physically meaningful locations and providing the flexibility of allowing more detailed representation of the off-mainstem system should that be desired at some point in the future. A general schematic of the system showing the key features represented by the linked BESTSM/MODFLOW model is presented in Figure One.

Calibration of Rio Grande System Model

        To confirm that BESTSM/MODFLOW was applied appropriately and that it accurately portrays the hydrologic and operational characteristics of the system, calibration of the model was required. Through the calibration process, the credibility of a model was established and a sense of confidence in the results from the model was developed. Calibration of a model is considered complete when simulated results produced by the model match historically observed values as measured at various locations in the system to a satisfactory level. Calibration of both the quantity and quality portions of the simulations were required. Because the water quality parameters simulated are conservative, they simply move with the flow; therefore, water quality cannot be calibrated until the quantity terms are calibrated. The MODFLOW model was assumed to be previously calibrated, so no changes were made to it. All calibration adjustments were confined to BESTSM input.

        The period of study selected for use in evaluating alternatives is 1925 to 1995, using a water year basis (e.g. water year 1925 is the period from October 1924 to September 1925). A shorter period was selected to calibrate the model. Releases from Caballo Reservoir were used to select the calibration period. The period of 1966 to 1990 was selected for calibration because:

• the average annual flow for this period was very similar to that of the 1925 to 1995 study period;
• the dry periods of 1966-1968, 1971-1973, and 1977-1978 are contained within this period; and,
• the wet period of 1986-1989 is contained within this period.

Study Period Calibration/Validation Period Calibration Period Years 1925-1995 1966-1995 1966-1990 Number of Years 71 30 25 Mean Annual (ac-ft) 691,440 686,810 657,440 Standard Deviation (ac-ft) 551,380 247,110 251,380

Hydrology

San Marcial Inflows: Gauged flows at San Marcial obtained from the U.S. Geologic Survey were used as inflows to the system. Daily flow values were used. Potential inaccuracies in the flow values at San Marcial due to gauge shifts, multiple channels, gauge submergence, or other sources of error were accounted for through the incremental inflow term calculated for Elephant Butte as described below.

Precipitation and Evaporation: Historic values for precipitation (inches) and evaporation (inches) were input to BESTSM to estimate net reservoir evaporation. A few months of evaporation data were missing at both reservoirs. Missing data were estimated as the average value for that month.

Incremental Inflows at the Reservoirs: Incremental inflows were estimated for Elephant Butte and Caballo Reservoirs. These flows were estimated on a monthly basis as the difference between the change in end-of-month storage and all known or estimated inflows and outflows. Estimated outflows included evaporation and seepage to the river downstream. The remaining volume required to explain the observed change in storage was the incremental inflow. In some cases this term was positive, indicating an inflow to the reservoir which could be from ungauged tributary inflows along the reservoir perimeter or a gain from bank storage. In some cases this term was negative, indicating a loss from the reservoir into bank storage. The estimated incremental inflows, whether positive or negative, were added to the reservoirs. Monthly values were input and divided uniformly over the month by the model to obtain daily values.

Rincon Valley Gains and Losses: Unaccounted gains and losses between Percha and Leasburg Dam were calculated on a monthly basis using the following equation:

Rincon Valley Gain/Loss = Rio Grande below Leasburg - Rio Grande below Caballo + Arrey Canal - Garfield Drain - Hatch Drain - Angostura Drain - Rincon Drain + Leasburg Canal

Rincon Drain Flows: Analysis of drain flow data in the Rincon in relation to diversion data at Percha Diversion Dam produced the following results with respect to return flow amounts and timing which were used in the model. The efficiency of use for Percha Dam diversions was set at 80 percent with respect to the river, i.e. 80 percent of the diversion is either consumptively used by crops or percolated to ground water. Accordingly, only 20 percent of the diversions return to the river.

Mesilla Drain Flows: Inflows for Selden Drain, Pichaco Drain, Del Rio Drain, La Mesa Drain, East Drain, and Montoya Drain are passed to BESTSM from the MODFLOW model. Seasonal values are received from MODFLOW and distributed to monthly values using the distributions as shown in Tables 2 and 3 below which were developed from analysis of historical data. The monthly values are distributed uniformly over the month by BESTSM to obtain daily values.
 

Drain Month March April May June July August September October Selden 4 9 12 13 17 18 16 11 Picacho 9 11 12 12 15 16 14 11 Del Rio 8 11 12 13 15 16 14 11 La Mesa 6 10 12 14 16 17 15 10 East 7 11 12 13 15 17 15 10 Montoya 8 11 12 13 15 16 14 11
 

Drain Month November December January February Selden 20 16 37 27 Picacho 25 22 28 25 Del Rio 24 21 29 26 La Mesa 22 19 32 27 East 23 21 30 26 Montoya 24 21 29 26

Mesilla Gains/Losses to Ground Water: Gains and losses to the alluvial aquifer in the Mesilla Valley were provided by MODFLOW. The total gain/loss for the reach was input to BESTSM at the Rio Grande at the El Paso Gauge. Seasonal values were divided uniformly over the season to obtain daily values.

Mesilla Spillway Flows into the River and Drains: A comparison of the gain/loss to ground water terms predicted by MODFLOW to the total gain/loss terms calculated using measured inflows and outflows between the Leasburg Diversion Dam and the El Paso Gauge was used to identify and estimate surface runoff from precipitation events and unmeasured spillway flows back to the river. These flows were estimated as the positive difference between total calculated gain/loss and MODFLOW predicted gain/loss to ground water.

        Comparison of the MODFLOW predicted drain flows to the observed drain flows was used to identify and estimate spillway flows into drains. These flows were estimated in composite as the positive difference between total observed drain flows and total MODFLOW predicted drain flows.

        The data were summed and used to develop a relationship between operational spills and total diversions. Spillway flows were estimated on a monthly basis as ten percent (10%) of agricultural diversions. These flows were input to BESTSM as monthly values at the Eastside Canal. The monthly values were divided uniformly over the month to obtain daily values.

Hueco Basin Gains and Losses: Gains/losses to the Hueco Basin below American Dam were estimated based on flow in the river below American Dam using predictive equations developed from historical monthly data from the period of 1955 to 1991. These flows were input to BESTSM at Riverside Diversion Dam. Different equations were used for the irrigation versus non-irrigation season. A paucity of reliable data to develop these relationships and a general lack of sensitivity in simulation results led to the ultimate decision to disregard these terms.

Spillway # 1, American Canal: Flows to the river from Spillway # 1 of the American Canal were estimated based on diversions into the American Canal using a predictive equation developed from historical monthly data from the period 1989-1991. These flows were input to BESTSM at Spillway #1.

Ascarate Wasteway: Historic flows from Ascarate Wasteway were used as input to the model. Monthly values were input and divided uniformly over the month to obtain daily values.

Water Quality

        Changes in the schedule of releases from the upstream dams will alter both the flow regime and the annual variation in water quality. The simulation model was modified to track certain water quality parameters to be able to predict variations in water quality. Following is a description of the methodology and procedures used to simulate water quality parameters.

Calculation Methodology for Total Dissolved Solids (TDS) and Other Constituents: TDS, sodium, chloride, and sulfate are simulated in the model using a mass balance approach. The water quality mass balance calculations are performed after the water quantity allocations are completed and all flows are known at all points in the system. The quality calculations start at the head of the system and proceed downstream from node to node calculating water quality for each component at each node based on the inflows and outflows to the node. The loss calculations take into account the difference in the densities of evaporative water (essentially pure water with a density of 1 mg/l), versus water flowing in the river (which may be as high as 1.05 mg/l).

        The concentration of TDS and each constituent is estimated at San Marcial using logarithmic equations relating concentration to flow. Concentration and flow are converted to mass and routed downstream. At inflow points, mass is added to the system. At outflow points, mass is subtracted from the system.

        All other inflows (incremental inflows to reservoirs, gains from ground water, drain flows, and spillway flows) are assigned a concentration which is independent of flow. Each inflow can be assigned a different constant value, but any inflow from that source will always have the same concentration. The mass from that source will change from time step to time step and introduce dynamics to the system because the flow changes (mass = concentration x flow).

        All outflows from the system use the concentration of the constituent in the water of the river at the point of outflow to determine the mass of the constituent leaving the system. The exception to this is evaporation at reservoirs which removes water from the system, but not constituent mass.

Parameter Value Estimation Procedures: The equations relating component concentration to flow at San Marcial were developed using limited data available at San Marcial and data available at the Rio Grande below Caballo. Regression equations between concentration and flow developed for the Rio Grande at Caballo were adjusted in a trial and error process for application at San Marcial to account for the effect of evaporation in the reservoirs. These equations are shown in the first row of Table Four.

Location TDS Sodium Chloride Sulfate Rio Grande at San Marcial * 700 - 52 x lnQ 100 - 8 x lnQ 60 - 4 x lnQ 150 - 5 x lnQ Incremental Inflow to Elephant Butte 450 80 60 150 Incremental Inflow to Caballo 700 120 60 150 Gains/Spills in Rincon 700 150 100 200 Garfield Drain 1300 300 250 400 Hatch Drain 1300 300 250 400 Angostura 900 250 200 300 Rincon Drain 1500 400 300 500 Selden Drain 1500 400 300 400 Picacho Drain 1500 400 300 400 Del Rio Drain 1500 400 300 400 La Mesa Drain 1500 400 300 400 East Drain 2200 600 400 600 Montoya Drain 1900 500 350 600 Gains/Spills in Mesilla 900 250 200 300 Gains/Spills in Hueco 1000 250 200 300 * Note: Q is flow at San Marcial in units of cfs

        The concentrations used for TDS at the drains were established based on all available historical data. No meaningful relationship between TDS concentration and flow was determined for any drain. The maximum and mean concentration determined from the data for each specific drain was used to calculate TDS mass for the inflow for that drain. Most of these values were not adjusted during calibration, but slight adjustments were made to some of the drains in the Rincon Valley. For the constituent components, data were not available for the drains. Estimated concentrations were developed for each constituent based on analysis of constituent data available on the mainstem during the non-irrigation season. During the non-irrigation season most of the flow in the river is from the drains, therefore, the water quality being measured reflects the water quality of the drains. These values were adjusted during calibration and the final values used are displayed in the Table Four.

        Concentrations for TDS and the constituent components for incremental inflows or reach gains were estimated based on data available from gauges on the mainstem at Caballo, Leasburg, and Rio Grande at El Paso. A flow and mass balance approach was used to calculate these values. Flows and concentrations for inflows at the head of a reach, outflows at the bottom of the reach, diversions within the reach, and drains within the reach were known, the remaining flow and remaining mass were attributed to the incremental inflows or reach gains. The concentration values calculated from this process were adjusted during calibration and the final values used are displayed in Table Four.

Operations

        Institutional and legal constraints govern releases of water from the dams and diversions of water throughout the system. The principal institutional and legal constraints were incorporated into the model in the form of water right priorities, operating rules and allocation requirements.

Demands: Historical diversions were used as demands for the Percha, Leasburg, Mesilla, American, International, and Riverside Diversion Dams. Historical diversions were used as demands for the Bonita Lateral. Demands were estimated for the River Pumps and California Lateral node on the basis of limited historical data.

Diversions: Each diversion dam was assigned two operating rights to meet demands. The first right was for a direct diversion from the river and the second right was to call for water from Caballo Reservoir. Priorities were assigned to diversion rights for the diversion demands starting at the downstream end and proceeding sequentially upstream. No direct diversion rights were assigned to the Percha Lateral or Arrey Canal due to their close physical proximity to Caballo. Priorities for releases from Caballo were next in order and assigned first to the Bonita Lateral, then to reservoir release rights at Percha Diversion Dam, then for diversions starting at the downstream end and proceeding sequentially upstream to Leasburg Diversion Dam. Employing this priority structure forced the use of flow in the river, including return flows, before calling for water out of Caballo Reservoir, and also appropriately represents that all diversions are of equal priority.

Reservoirs: Storage rights to fill Project Accounts in Elephant Butte and Caballo Reservoirs were given top priority with Elephant Butte first and Caballo second. Releases from Elephant Butte were limited to prevent flows greater than 5,000 cfs below the reservoir. If uncontrolled spills occur in excess of 5,000 cfs, releases would be zero, but flow downstream would be greater than 5,000 cfs.

        Past operations of Caballo Reservoir were varied by the Bureau of Reclamation to accommodate changing needs and circumstances from year to year. In general, storage in Caballo was kept between 80,000 and 150,000 acre-feet during the irrigation season, with storage drawn down to about 50,000 acre-feet by the first of October. Because these varying operations cannot be described simply by a consistent set of operating rules, these operations were simulated by invoking a fill-to-target rule. This pulled water from Elephant Butte to Caballo to meet the end-of-month storage targets as specified. Historical end-of-month storage contents were used as storage targets for calibration purposes. For simulation of alternatives, operating rules representing current operations or modified rules appropriate for use in representing the alternatives will be invoked for those model runs, rather than keying on historic levels at Caballo as was done for calibration.

        The area-capacity curves used for Elephant Butte and Caballo Reservoirs were the most recent available from the Bureau of Reclamation, dated 1981. Flood pools of 50,000 acre-feet in Elephant Butte and 100,000 acre-feet in Caballo were specified in accordance with the international treaty between the U.S. and Mexico. When reservoir storage begins to fill flood pools, this water is released as quickly as possible, as limited by downstream channel capacity limitations, to evacuate flood storage. For simulation of alternatives, operating rules representing current flood control operations will be used.

Calibration Results

        Once developed, the model was calibrated by comparing simulated versus observed flows and constituent concentrations at nodes coinciding with stream gauges and measurement stations.

Quantity: The ability of the model to represent the hydrologic and operational aspects of the system was assessed by comparison of simulated and observed flows at the stream gauges on the Rio Grande and comparison of simulated versus observed end-of-month storage in the reservoirs.

        The following Rio Grande stream gauges were used for comparison: below Elephant Butte Reservoir, below Caballo Dam, below Leasburg Diversion Dam, at the El Paso Gauge, and below American Dam. The observed flows below American Dam were not actually gauged, but were calculated as the difference between flows at the El Paso Gauge and American Canal Diversions to allow a point of comparison near the end of the system. An example of the comparison of simulated versus observed monthly flows at the stream gauge below Leasburg Diversion Dam is shown on Figure Two. In all cases the match between simulated and observed is extremely good. The simulated flows below American Dam tend to be slightly higher than the “observed” values, which may be the result of gains or inflows between the El Paso Gauge and American Dam which were not considered in the calculation of the “observed” data but which are captured in the model results. The differences are very small and considered to be insignificant.

        Simulated and observed end-of-month storage contents were compared at Elephant Butte and Caballo Reservoirs. These comparisons are shown on Figures Three and Four, respectively. By invoking the fill-to-target rule for Caballo, a nearly perfect match between simulated and observed results was obtained at Caballo. Consequently, virtually all simulation error in the system from all sources is concentrated in the results at Elephant Butte. The match at Elephant Butte is very good, but simulated content values tend to be slightly lower than observed. This would indicate that there is additional inflow to the system which is not represented in the model, or there are inaccurate gauge measurements, or more flow is modeled as passing Riverside Dam than there is in reality. For an example of additional inflow, more return flow in a particular year might occur in the Rincon Valley than was estimated with an average efficiency of 80 percent, thereby requiring less draw on the reservoirs. For an example of gauge error, a 10 percent gauge error in one or two of the diversion gauges in the same direction in a month or a season, which is still a good gauge, might result in an error which is accumulated in the reservoir. The lack of gauged data below Riverside Dam precludes a definitive conclusion on the accuracy of the amount of flow modeled as leaving the system. However, review of these values by people knowledgeable of the system found these numbers to be reasonable.

        Note that reservoir content plots have a “memory” in that a deviation in reservoir contents in only one month can separate the simulated and observed lines and this difference can persist for months or years. Therefore, if the patterns of the two lines are consistent, the difference between the lines is of less concern. Further, it is better that the simulated line be below the observed line because this is a conservative representation. In water supply planning , it is advisable to be conservative regarding water supplies.

Quality: The ability of the model to represent the water quality conditions in the system is assessed by comparison of simulated and observed concentrations of TDS, sodium, chloride, and sulfate at measurement stations on the Rio Grande.

        For TDS, data were available at the following stations on the Rio Grande; below Elephant Butte Dam, below Caballo Dam, below Leasburg Diversion Dam, below Mesilla Drain, and at the El Paso Gauge. For sodium, chloride, and sulfate, data were available at the following stations on the Rio Grande; below Caballo Dam, below Leasburg Diversion Dam, and at the El Paso Gauge.

        Example plots of simulated and observed values in the form of scatter plots of concentration versus flow are shown on Figure 5 for TDS, on Figure 6 for sodium, on Figure 7 for chloride, and on Figure 8 for sulfate. The regression equations and values of the squares of the correlation coefficients (r²) shown on the plots apply to the observed data, not the simulated. The observed values shown on these figures are all the data available, whether or not it was from the calibration period. Consequently, some of the extreme observed points may correspond to conditions outside the range seen in the calibration period. Further, the simulated values represent monthly average values while the observed values represent instantaneous values. Monthly average values will show less variability and dampen out the extremes when compared to instantaneous values. For these reasons, the comparison is not exactly between like data, but it is, nevertheless, a valid, valuable comparison illustrating a good correspondence between simulated and observed values. In general, the model provides a good match between simulated and observed values for all water quality components, and captures a significant portion of the variability and dynamics of the system.

        In addition to the plots of concentration versus flow as a means to assess the quality calibration, sufficient data were available at the Rio Grande at the El Paso Gauge to develop a time series plot of simulated and observed concentrations for all of the quality components. Figure 9 shows the TDS values from this simulation test. Again, the simulated values are monthly average values and the observed values are instantaneous measurements. Average values in comparison to instantaneous values will not capture all of the extremes, as is seen on the Figure. During high flow conditions with good water quality, the model provides a good match between simulated and observed values for each parameter. During low flow conditions with poorer water quality, the model provides a generally good match between simulated and observed values, missing only the highest extremes.

Conclusion

        The model appears to appropriately represent system behavior for each quality component both with respect to specific threshold levels and water quality for all flow levels. That is, the model appears to predict with reasonable confidence whether the TDS concentration will be above or below 1,000 mg/L and whether sulfate concentrations will be above or below 300 mg/L. These threshold values are critical for treatability determinations. Further, the model appears to predict with reasonable confidence the concentration of each component from high to low flow conditions. Accurately predicting quality over the entire range is important in the evaluation of drain mitigation strategies.

        Subsequent analyses planned as part of the sensitivity analyses will further probe the behavior of the quality simulations and the confidence level of the predictions, but strong evidence has been presented to consider the water quality component of the model to be calibrated. The calibration results presented for both quantity and quality demonstrate that the linked BESTSM/MODFLOW model of the Rio Grande System from San Marcial to Riverside Diversion Dam as developed represents the behavior of the actual system in an accurate, appropriate manner. The calibrated model should prove to be a valuable tool for evaluating alternatives as conceived for the Las Cruces - El Paso Sustainable Water Project.