Using the RGDSS Groundwater Model to Compute Response Functions

Transcription

1 Using the RGDSS Model to Compute Response Functions The computation of response functions for Subdistrict No. 1 involves making a set of paired model simulations using the Rio Grande Decision Support Model ( Model ). In each pair of simulations, one simulation represents a reference case that is referred to as the historical run. The second simulation is referred to as the impact run. The term impact refers to the fact that the simulation is intended to quantify the impact of some change to historical conditions. The Model is therefore applied in a change mode. The difference in predicted stream gains and losses calculated by subtracting the impact run from the historical run is the quantification of the impact of differences in the Model inputs between the runs. The impact of well pumping in Subdistrict No. 1 on streams is evaluated by generating a set of model inputs that represent a different level of well pumping inside Subdistrict No. 1. Specifically, the amount of well pumping is adjusted such that the amount of consumption matches the amount of consumption of imported water quantified with the methodologies of the recharge decrees ( Imported Water Offset ). In the impact run the amount of pumping for land served by ditches without recharge decrees is zero. The impact run for Subdistrict No. 1 is a simulation where the inputs to the model are changed inside Subdistrict No. 1 to reflect that the amount of groundwater consumptive use is equal to the Imported Water Offset. A comparison of the stream aquifer interaction between the impact run and the historical run reflects changes in the stream flow that result from changes in the inputs. Response Functions In order to facilitate prediction of future stream depletions as a function of activities inside Subdistrict No. 1, a response function will be used. The critical Model input is the difference between consumption due to well pumping and the Imported Water Offset. The Model output of interest is the impact to streams. The goal is therefore to generate a relationship between stream impacts in location, time and the amount from groundwater consumption in excess of the Offset. This relationship is called a response function. The response function is a fraction of the applied stress that translates to a change in stream flow at a particular location at a particular time. It is assumed that when the applied stress is changed, that the change in stream flow changes proportionally. This requires the response to be linear. The Model is, of course, nonlinear, but for moderate changes in the applied stresses the response varies close to linear. Once the response functions are generated, by applying the model and post processing the results, stream depletions can be readily calculated. Specifically, the response functions generated for

2 Subdistrict No. 1 translate the annual difference between consumption of groundwater due to well pumping and the Offsets to monthly stream depletions by stream. Therefore in order to calculate the impact of well pumping in a particular year on a particular stream for a particular month, what is required is: 1. determine the difference between consumption due to well pumping and the Offset (in acre feet annually), 2. multiply it by the appropriate response function value, and 3. the result is the impact to the particular stream for a particular month. Model Input Data use calculations are done using the StateCU program. This program uses historical irrigated area and crop information to calculate how surface and ground water would be applied and consumed. These calculations are done on a ditch service area basis. Two sets of consumptive use calculations are available. The historical use for the period is contained in rg27.dwb and represents the best estimates of historical consumptive use. An alternative scenario which considers the same period but estimates what consumptive use would have occurred had there been no wells is contained in rg27_noq.dwb. This is the so called No Pumping scenario. The amount of historical consumptive use under each ditch is estimated by StateCU. Specifically the consumptive use from surface water application (Surface Water Farm Diversion to CU) and from groundwater application ( Diversion to CU) is reported. Davis Engineering Service, Inc. provided a summary of the Offsets. A program called mkrc was used to summarize the amount of historical consumption from surface water and groundwater and compare it to the Offsets for those ditches with recharge decrees. Table 1 shows these results for the four canals with recharge decrees. For each year (column 1), the Offset, as reported by Davis Engineering, is listed in column 2. The amount of consumptive use from direct surface irrigation as reported by StateCU is listed in column 3. The amount of groundwater consumptive use of the Offset is the difference of columns 2 and 3 and is shown in column 4. The amount of historical groundwater consumptive use estimated by StateCU is shown in column 5. Column 6 is the difference between column 5 and 4, and shows the amount of historical groundwater consumptive use in excess of the Offsets. Column 7 shows the ratio of the Offset and the historical groundwater consumptive use (column 4 divided by column 5). The ratio is the key result from this calculation. consumptive use is basically the difference between total groundwater pumping ( Diversion ) and return flow (Non Consumed ). In order to adjust the groundwater consumptive use to match the Offset, multiplying all three of these quantities by the ratio will match the groundwater consumption to the Offset while maintaining the correct relationship between total

3 pumping and return flow. The program mkrc creates a file ratio.dat which reports the ratio needed to match groundwater consumptive use to the imported water offsets for each of the ditches with recharge decrees. The program mkrcdwb then reads this table as well as the historical consumptive use budget file rg27.dwb and the no pumping consumptive use budget file rg27_noq.dwb and produces a new consumptive use budget file called rg27_rc.dwb. In rg27_rc.dwb, the historical ditch water budgets for the ditches with recharge decrees are adjusted by multiplying the groundwater components by the appropriate ratio. This scales the annual and monthly pumping and recharge numbers such that they match the Offsets. For those ditches that do not have recharge decrees, the rg27_rc.dwb file contains the no pumping budgets from rg27_noq.dwb. The rg27_rc.dwb budget file contains ditch budgets where groundwater pumping under every ditch is reduced to the amount of the Offset, which is zero in the absence of a recharge decree. StatePP is then used to generate model input files using the historical ditch budget file rg27.dwb which is called H5P5 and using rg27_rc.dwb which is called H5P5 87. The output from StatePP is a set of files consisting of an M&I well pumping file (.mi), agricultural well pumping file (.wel), precipitation recharge file (.ppt), irrigation return flow file (.irr), canal leakage file (.can) and rim recharge (.rim), as well as native (.ets) and subirrigation (.sub1,.sub2 and.sub3) evapotranspiration files. These files cover the period using monthly stresses. The H5P5 87 files represent a situation where the impact of all groundwater pumping in the model domain would be evaluated. In order to isolate just the pumping in Subdistrict No. 1, the program mksub is used to generate a new set of output files called H5P5a. How the mksub program is run is shown in the file Mksub. The mksub program switches between areas inside Subdistrict No. 1 and outside Subdistrict #1 on a model cell by cell basis so that the H5P5a data contains the H5P5 87 (impact) results inside Subdistrict No. 1 and H5P5 (historical) results outside Subdistrict No. 1. The impact run for Subdistrict No. 1 is therefore performed using the H5P5a data set, while the historical run is performed using the H5P5 data set. The reason for this multi step procedure is that Subdistrict No. 1 does not line up with the ditch service areas. By generating an impact data set for the entire model domain, the mksub program can then select the appropriate data for just Subdistrict No. 1. The mksub program also has a mode where the subdistrict analysis can be done in a response function mode. The u flag instructs the mksub program to use the impact data (in this instance H5P5 87) only during the first year. For all years after the first year, the stresses match the historical data. This allows the impact analysis to isolate the effects associated with a particular year. The y flag on the mksub program allows the user to select the order of the years. So, for example, using the command line parameters y , will instruct mksub to produce an output file where the historical years are output in the order followed by

4 Input data sets for a response function analysis starting with each year from 1988 to 25 and cycling through the historical sequence of years 1988 to 25 was generated using the instructions in Mkurf. The input data for the historical simulation for each year was named H5P5_YY where YY represents the last two digits of the year. The input data for the impact simulation for each year was named H5P5aYY. Numerical experiments with these simulations showed that response functions so derived introduced some dependency on the exact sequence of years that followed the year being analyzed. This is inappropriate because when using the response function to determine the impacts on streams from a particular year, it is unlikely that the following years will occur in exactly the same historical sequence. In order to avoid this problem, the response functions were generated using average monthly conditions. The Average Monthly simulation is a simulation where stresses for each calendar month is averaged. These stresses represent what each month typically looks like, and captures the seasonal variations throughout the year. The mkurf program was therefore used to generate response function input data sets that consist of the historical and impact inputs for each of the years , followed by average monthly inputs for each year following the first year. Specifically the H5P5=YY data sets represent the historical data for each year from 1988 to 25 as the first year in the simulation, followed by 99 years of average monthly data. Similarly, the H5P5AYY data sets represent the impact data set for these years, followed by 99 years of average monthly data. Model Simulations Thirty six simulations were generated using the Model, eighteen for historical and eighteen for impact conditions. The historical simulations were named H5AP12=YYX where YY represents the last two digits for the year. For each simulation the starting heads for that calendar year was extracted from the transient calibration simulation. The simulation was then run for 1 years, where the first year represented the specific historical year and the next 99 years represented monthly average conditions. The impact simulations were named H5AP12AYYX. The first year represented the specific year but using the pumping levels appropriate for the recharge decrees. The next 99 years represented monthly average conditions. The Model outputs consist of four main files. The text output produced by MODFLOW is the.out file and contains information about the simulation such as input files read, calculations, volumetric budgets and the like. A detailed budget is saved using the budget package in the.sbb file. This is a binary file in HYDMOD format which contains a detailed budget for every time step in the simulation for subsequent analysis using the mkbgt program. The stream flows for every stream segment in the model and water level observations at selected well locations are stored in the.sfi file using the HYDMOD package. This file is subsequently analyzed using the mksum program to generate detailed stream gain and loss analysis. Heads throughout the domain at selected times are stored in the.head files. The simulations require 6 8 hours of computer time each.

5 Response Function Calculation The water budget calculated by the Model is summarized using two programs. The mkbgt program is used to analyze all components of the water budget using output saved by MODFLOW in the.sbb file. It produces summary tables of budget terms such as well pumping, recharge, evapotranspiration, stream flows, etc. on an annual basis for the domain as a whole, as well as for geographic regions. Alternatively, when presented with two model simulations, the mkbgt program will report the differences in the budget terms between the model simulations. Since for this application the interest is in the differences in the inputs and output of the model between the historical and impact simulations, the mkbgt program was primarily employed in this mode. The mksum program operates similarly to the mkbgt program, but analyzes stream flows. The program reads detailed stream flow information from the.sfi files and produces a summary of the inflows, outflows, diversions and gains and losses to the streams. Like mkbgt, the mksum program will report the difference in these terms when presented with two simulations. The mkbgt and mksum programs for the simulations listed above are run using the program MkurfbgtX. The output is summarized in data sets named P12AyyX bgt and P12AyyX str. The.htm files are tables in HTML format that can be viewed using a web browser or spreadsheet program. In addition, detailed monthly values are also saved in DBF format which can be accessed using a spreadsheet or database program. In the application to Subdistrict #1, the mksum was specifically run with the x flag. This flag causes the program to summarize stream depletions by specific reaches of interest for purposes of administration. Specifically, on the Rio Grande, the stream depletions are reported for the reach from Del Norte to the Excelsior Ditch headgate, from the Excelsior Ditch to the Chicago Ditch headgate, and from the Chicago Ditch headgate to the State Line. For smaller streams in the Closed Basin, stream depletions are summarized from the edge of the model to the last diversion. This subdivision of the stream is rather important. Since the Model is a physical flow model, and not a water rights model, the Model cannot increase historical diversions to simulate how stream gains would be put to beneficial use in accordance with the priority system. By limiting the the impact calculation to only those gains and losses that occur upstream of the last water right, the output more accurately estimates depletions that may affect water rights. Discussion of Response Functions In order to evaluate the appropriateness of the response functions as well as ascertain the magnitude of the stream depletions, a model simulation was created similar to those used to determine the response functions described above. The paired runs consisted of the calibration simulation from 197 to 25 and an impact simulation which applied the Subdistrict #1 impact stresses for all years from 197 to 25. By comparing the predicted stream flows in these simulations the cumulative impact of

6 groundwater pumping from the area covered by Subdistrict #1 can be assessed. Table 2 shows the computed stream depletions as a result of the groundwater pumping in Subdistrict #1 as an annual average for the last ten years of this model simulation. Using 5 acre feet of average annual depletion as a lower limit for reporting, it was determined that response functions for Subdistrict #1 impacts are appropriate for each of three streams: the Rio Grande, the Conejos River and La Jara Creek. The Rio Grande was divided into three administrative reaches based on discussions with the State Engineer's Office. The first reach is from Del Norte to the headgate of the Excelsior Ditch. The second reach is from the Excelsior Ditch to the Chicago Ditch. The third reach is from the Chicago Ditch to the State Line and includes the Norton Drain. A response function was generated for each of these three reaches. For the Conejos River, a single response function was generated for the Conejos River proper, the Rio San Antonio and McIntyre Spring. The response function represents the combined impact to these three sources. The La Jara Creek response function represents only La Jara Creek. Table 3 shows the computed Net from the simulation as well as the stream depletions computed by the model for the five stream reaches for which response functions were calculated. The Net was calculated as the change in groundwater pumping minus the change in recharge in the model simulations. Here the change refers to the difference between the impact and historical simulations. The stream depletions shown in Table 3 are the cumulative stream depletion for the period Therefore, these values can be used to test the predictive ability of the response functions.

7 Table 1A: Rio Grande Canal (2812) Year Offsets Surface Water of Offsets (2)-(3) Excess Use (5)-(4) Use Ratio (4)/(5) (1) (2) (3) (4) (5) (6) (7) Average

8 Table 1B: Farmers Union Canal (2631) Year Offsets Surface Water of Offsets (2)-(3) Excess Use (5)-(4) Use Ratio (4)/(5) (1) (2) (3) (4) (5) (6) (7) Average

9 Table 1C: Prairie Ditch (2798) Year Offsets Surface Water of Offsets (2)-(3) Excess Use (5)-(4) Use Ratio (4)/(5) (1) (2) (3) (4) (5) (6) (7) Average

10 Table 1D: San Luis Valley Canal (2829) Year Offsets Surface Water of Offsets (2)-(3) Excess Use (5)-(4) Use Ratio (4)/(5) (1) (2) (3) (4) (5) (6) (7) Average

11 Table 1E: All Four Canals with Recharge Decrees Year Offsets Surface Water of Offsets (2)-(3) Excess Use (5)-(4) Use Ratio (4)/(5) (1) (2) (3) (4) (5) (6) (7) Average

12 Table 2: Subdistrict #1 Computed Stream Depletions Average Depletions Stream (af) Rio Grande 5599 LaJara Creek 136 McIntyre Spring 18 Werner Arroyo 48 Alamosa River 44 Conejos River 41 Saguache Creek 36 Norton Drain 24 San Luis Creek 21 Rio San Antonio 11 Sand Creek 1 Trinchera Creek 7 Deadman Creek 7 Big Spring Creek 5 Rito Alto 2 Little Spring Creek 1 Crestone Creek 1 Willow Creek SanIsabel Creek Cottonwood Creek SangreDeCristo Creek Culebra Creek Kerber Creek Spanish Creek Cotton Creek Zapata Creek Costilla Creek WildCherry Creek Major Creek Ute Creek Medano Creek Carnero Creek Garner Creek LaGarita Creek

13 Table 3: Modeled Net and Stream Depletions Year Modeled Net Rio Grande Del Norte to Excelsior Modeled Stream Depletions Rio Grande Chicago to State Line Rio Grande Excelsior to Chicago Conejos River Rio San Antonio and La Jara Creek and Norton Drain McIntyre Spring