Groundwater Availability Model (GAM) FAQs
Welcome to the Groundwater Resources frequently asked questions page.
The following GAM related questions were either brought up during a March 9, 2006 meeting held to receive feedback and input from consultants and other specialized interested parties on the technical components of the regional water planning process or have been asked directly to GAM staff.
- Why is average recharge different in various groundwater availability model run reports?
- I need help figuring out rules for well spacing, what can I use?
- If the GAMs are too regional, should I refine the model to just my area of interest?
- It would have been nice if they [GAMs] were done in time for the Regional Water Planning process.
- There was a question on GAMs that extend into two or more Regional Water Planning regions and what should one region use as "boundary conditions" for pumpage in other region.
- Water Availability Models (WAM) and Groundwater Availability Models (GAM) Interaction. This was mentioned and pulled out as a separate discussion topic, however most of the discussion focused on the WAM side. For example, whether the WAMs would take into account change in discharge to rivers in the naturalized flows and doing a "sensitivity analysis" on channel losses/gains.
- GAM Sensitivity and Uncertainty: The one-at-a-time sensitivity analyses included in the GAMS are very helpful, but could be improved by setting the minimum and maximum value of each parameter to represent some level of the uncertainty in that parameter (e.g., an expected minimum and maximum) instead of a uniform +/- percentage. This would, relatively easily, result in a hybrid sensitivity/uncertainty approach that in- corporates a minimum level of uncertainty information in the output. Much better would be to perform an uncertainty analysis. This may range from the fairly straightforward first-order second moment method (e.g., as described in Glasgow, H.S., M.D. Fortney, J. Lee, A.J. Graettinger, and H.W. Reeves, 2003, "MODFLOW 2000 head uncertainty, a first-order second moment method", Ground Water, 41(3):342-350) to more computationally intensive methods such as Monte Carlo Simulation (e.g., as is available in the MODFLOW graphical user interface "Groundwater Vistas"). Such an analysis would provide numerous benefits, including an identification of data gaps, recognition of the uncertainty in model predictions, quantification of safety factors, and improved prioritization of groundwater water management strategies. The results of uncertainty analyses are particularly valuable in situations, like the regional water planning process, where budget constraints require strict prioritization of short-term tasks and long-term strategies. [A] For these reasons, TPWD recommends that TWDB select and promote a standardized uncertainty analysis framework for the GAMs. [B] Furthermore, TPWD recommends that the sensitivity and uncertainty analyses be used to identify the most important data gaps and that TWDB attempt to collect such data. Hopefully, ongoing data collection efforts, such as the TPWD Edwards Plateau Ecoregion spring research and sampling program (of which TWDB is a contributor), will fill some data gaps.
- Finalization of GAMs: Given the relative lack of data and other model constraints, Texas Parks & Wildlife Department recommends that TWDB staff be given the primary responsibility to identify future improvements and changes to the GAMs as well as judge them complete. While the GAMs should not be considered finished if the Regional Water Planning Groups (RWPGs) are dissatisfied with the models, this is a technical issue and, like the population projections, should have some level of state oversight.
- Groundwater-Surface Water Interactions as Represented in the GAMs and Water Availability Models (WAMs): The GAMS and WAMs, and the data they are built on, provide boundary conditions to adjoining models. For example, recharge to the GAM is dependent on streamflows. It is unclear how future GAM predictions take into account predicted changes in streamflows from the WAMs. Similarly, the WAM naturalized flows implicitly include historical gains and losses to/from aquifers. Many aquifers are projected to experience additional drawdowns in the next 50 years. It is unclear how these projected drawdowns are recognized in the WAMs. In addition, future GAM predictions for one aquifer affect overlying and underlying aquifers, through the cross-formational flow term. This is true even for local groundwater sources that are not designated aquifers. It is unclear if projected head changes in adjoining aquifers and groundwater sources are used to project changes in cross-formational flows. TPWD recommends that updates to the GAMS and WAMs include the best predictions available of future changes in adjoining surface water bodies and groundwater sources. TPWD recognizes that this process will not always be ideal, but the best available predictions are more appropriate than historical data for defining boundary conditions to future projection models.
- Groundwater-Surface Water Interaction Determinations: The GAM reports, Water Availability Models (WAM) reports, and recent publications by The University of Texas Bureau of Economic Geology (BEG; e.g., Scanlon et al., 2005, "Groundwater-Surface Water Interactions in Texas") describe numerous methods for quantifying existing groundwater-surface water interactions. Such data may be used to constrain the WAM and GAM predictions; however there are numerous methodologies, each with advantages and disadvantages in certain circumstances. Recognizing that a uniform approach would not be reasonable, TPWD recommends that TWDB develop a short guidance document on appropriate methods for estimating groundwater/surface water interactions. Additionally, TPWD recommends that future GAM and WAM reports briefly mention each method and present the rationale for the method(s) chosen.
- In counties with overlapping GAMs, the TWDB might consider telling the Regional Water Planning Groups which GAM is considered the most accurate and, in essence, make the decision for the groups as to which GAM to use. This was a comment that many participants agreed would be useful.
- Relating to the above question, there was a general question as to whether the differences between overlapping GAMs had been resolved, for example the central and northern Gulf Coast GAMs.
- Many participants agreed that the use of GAMs in the Regional Water Planning process went quite well. However, one Groundwater Conservation District suggested that the GAMs for Regions A and O (northern and southern Ogallala) needed refinement, with "hydrograph traces that were 30 to 50 feet off in some areas."
- Suggest looking at using "alluvial deposits" as supplies. There was a comment that this may be a surface water-groundwater interaction issue.
Answers to Frequently Asked Questions
1. Why is average recharge different in various groundwater availability model run reports?
Some districts are comparing the average recharge reported in groundwater availability model runs for management plans to groundwater management area predictive model runs. The management plan runs are designed to address Texas State Water Code, Section 36.1071, Subsection (h), which states that in developing its groundwater management plan, a groundwater conservation district shall use groundwater availability modeling information provided by the Executive Administrator of the Texas Water Development Board. Both types of model runs are produced by the Texas Water Development Board staff but for different purposes. In many cases the average recharge reported is different. There are several reasons why this may occur:
- Most of the groundwater management area predictive model runs use a 30-year average recharge amount since the variability in the magnitude and frequency of droughts and wet periods on recharge is hard to predict. The years used to calculate the average recharge for these model runs are typically noted in the “Parameters and Assumptions” section of the model run report. For the groundwater conservation district management plans we extract the historical water budget information to address the various requirements of Texas State Water Code, Section 36.1071, Subsection (h). The recharge values provided for groundwater conservation district management plans is averaged from the historical calibrated transient period, which typically covers 1980 to 1999. Because we are reporting other flow conditions for this 20-year period it would be inconsistent to report a 30-year average recharge value so we report the average recharge for the same 20 years. In addition Texas State Water Code, Section 36.1071, Subsection (h) specifies that we report only the amount of recharge derived from precipitation so in some cases we have to back out other recharge factors, for example irrigation return flow, from the recharge values extracted from the model. Comparing recharge averaged for a 20-year period to a 30-year period will probably result in different values, especially if we had to “back out” other non-precipitation related recharge factors.
- The groundwater availability models are living tools and some have already been updated to improve recharge estimates. It is important to note the model version used for any particular model run. If the model version is not noted in the “Parameters and Assumptions” section of the model run report then the model used was probably based on the initial version of the official groundwater availability model. We recommend that you recognize whether you are comparing older, dated information to new and improved information.
- Each model is composed of hundreds to thousands of grid cells that do not align up to political boundaries, such as county or groundwater conservation district boundaries. Extracting information out of the model in a consistent manner has been challenging: How do we avoid double accounting? Are we using the most accurate data set? Are we reporting information based on the official aquifer boundaries, which typically contains fresher water, or are we including parts of the formation that typically contains poorer quality of water and falls outside the official aquifer boundaries? For the most part we use a geographical data extraction approach we call the “centroid method”. We overlay the political boundaries on top of the model and assign each grid cell in the model to a particular political boundary based on where the center of the top of the grid cell (the centroid) falls. For example, if a cell contains two counties, the cell is assigned to the county where the centroid of the cell is located. We make every effort to avoid double accounting and we endeavor to use the most current and accurate datasets. Therefore, changes or updates to political or aquifer boundaries may have a bearing on the recharge value reported for any given model run.
- Dry cells. A model cell may go “dry” during the model run. Dry cells occur when the water level in a cell falls below the bottom of the cell. If high pumping is the primary factor for a cell going dry, the model is saying that the pumping may be too great for the aquifer in that particular area. Although some of the models use the MODFLOW rewetting package to allow cells to “rewet”, many do not. It is important to identify why a cell stays dry and address the causes. In reality, the aquifer will probably not go dry because pumping will become uneconomical before the aquifer actually is fully dewatered in any particular area. In many of the models, once a cell goes dry the cell is deactivated for the rest of the simulation. Any recharge that would have been applied to the cell that is dry is no longer considered. Therefore if two simulations are performed on the same model using average recharge, one run with low or no pumping and the other run with high pumping, the recharge extracted from the model runs might differ if ,for example, more dry cells occurred under the high pumping scenario
2. I need help figuring out rules for well spacing, what can I use?
The GAMs are regional models and use grids that are generally too large to help with analyzing individual wells on a local scale. A better tool would be an analytical model. If site specific information is not available on aquifer properties, then information from the appropriate GAM model may be used to estimate various key input parameters. We have developed a web-based tool that estimates potential drawdown of water levels in a well due to various pumping scenarios.
3. If the GAMs are too regional, should I refine the model to just my area of interest?
If additional information/data supports this and the question that you need answered requires refinement of the model grid, then it is important to review and use the regional scale GAMs to understand and adjust boundary conditions.
4. It would have been nice if they [GAMs] were done in time for the Regional Water Planning process.
The GAMs for the major aquifers were completed during the last round of regional planning. The models for the minor aquifers are in progress. Therefore, many existing GAMs were available for the major aquifers for RWP consultants to test strategy options. In order for GAM staff to develop the appropriate pumpage dataset(s), all regional supply data needs to be completed in the appropriate online database. The timing of data needed for model runs will continue to be a challenge in future planning cycles.
5. There was a question on GAMs that extend into two or more Regional Water Planning regions and what should one region use as "boundary conditions" for pumpage in other region.
Pumpage in other regions will be based on previous state water plan supplies and strategies unless, on a case-by-case basis, we are aware of large projects along boundaries.
6. Water Availability Models (WAM) and Groundwater Availability Models (GAM) Interaction. This was mentioned and pulled out as a separate discussion topic, however most of the discussion focused on the WAM side. For example, whether the WAMs would take into account change in discharge to rivers in the naturalized flows and doing a "sensitivity analysis" on channel losses/gains.
TWDB contracted with HDR to investigate the feasibility of linking WAMs and GAMs. The study determined that linking WAMs and GAMs is not feasible.
7. GAM Sensitivity and Uncertainty: The one-at-a-time sensitivity analyses included in the GAMS are very helpful, but could be improved by setting the minimum and maximum value of each parameter to represent some level of the uncertainty in that parameter (e.g., an expected minimum and maximum) instead of a uniform +/- percentage. This would, relatively easily, result in a hybrid sensitivity/uncertainty approach that in- corporates a minimum level of uncertainty information in the output. Much better would be to perform an uncertainty analysis. This may range from the fairly straightforward first-order second moment method (e.g., as described in Glasgow, H.S., M.D. Fortney, J. Lee, A.J. Graettinger, and H.W. Reeves, 2003, "MODFLOW 2000 head uncertainty, a first-order second moment method", Ground Water, 41(3):342-350) to more computationally intensive methods such as Monte Carlo Simulation (e.g., as is available in the MODFLOW graphical user interface "Groundwater Vistas"). Such an analysis would provide numerous benefits, including an identification of data gaps, recognition of the uncertainty in model predictions, quantification of safety factors, and improved prioritization of groundwater water management strategies. The results of uncertainty analyses are particularly valuable in situations, like the regional water planning process, where budget constraints require strict prioritization of short-term tasks and long-term strategies. [A] For these reasons, TPWD recommends that TWDB select and promote a standardized uncertainty analysis framework for the GAMs. [B] Furthermore, TPWD recommends that the sensitivity and uncertainty analyses be used to identify the most important data gaps and that TWDB attempt to collect such data. Hopefully, ongoing data collection efforts, such as the TPWD Edwards Plateau Ecoregion spring research and sampling program (of which TWDB is a contributor), will fill some data gaps.
[A] The GAM program currently follows a standard uncertainty procedure as described in Applied Groundwater Modeling: Simulation of Flow and Advection Transport (1992, M. P. Anderson and W.W. Woessner, pg 246). [B] TWDB staff have discussed using PEST (Parameter ESTimation: a general-purpose, model-independent, parameter estimation and model predictive error analysis package developed by Dr. John Dohert), with the GAM technical advisory group. Budget constraints and other priorities restricted us from pursuing a research project using PEST for fiscal years 2006 and 2007. We hope to pursue this option in the future, as well as updating the GAMs every five years with additional data, as applicable and needed.
8. Finalization of GAMs: Given the relative lack of data and other model constraints, Texas Parks & Wildlife Department recommends that TWDB staff be given the primary responsibility to identify future improvements and changes to the GAMs as well as judge them complete. While the GAMs should not be considered finished if the Regional Water Planning Groups (RWPGs) are dissatisfied with the models, this is a technical issue and, like the population projections, should have some level of state oversight.
We agree.
9. Groundwater-Surface Water Interactions as Represented in the GAMs and Water Availability Models (WAMs): The GAMS and WAMs, and the data they are built on, provide boundary conditions to adjoining models. For example, recharge to the GAM is dependent on streamflows. It is unclear how future GAM predictions take into account predicted changes in streamflows from the WAMs. Similarly, the WAM naturalized flows implicitly include historical gains and losses to/from aquifers. Many aquifers are projected to experience additional drawdowns in the next 50 years. It is unclear how these projected drawdowns are recognized in the WAMs. In addition, future GAM predictions for one aquifer affect overlying and underlying aquifers, through the cross-formational flow term. This is true even for local groundwater sources that are not designated aquifers. It is unclear if projected head changes in adjoining aquifers and groundwater sources are used to project changes in cross-formational flows. TPWD recommends that updates to the GAMS and WAMs include the best predictions available of future changes in adjoining surface water bodies and groundwater sources. TPWD recognizes that this process will not always be ideal, but the best available predictions are more appropriate than historical data for defining boundary conditions to future projection models.
TWDB currently has a contract with HDR to investigate the feasibility of linking WAMs and GAMs. Possible improvements to both or either programs may result from this study.
10. Groundwater-Surface Water Interaction Determinations: The GAM reports, Water Availability Models (WAM) reports, and recent publications by The University of Texas Bureau of Economic Geology (BEG; e.g., Scanlon et al., 2005, "Groundwater-Surface Water Interactions in Texas") describe numerous methods for quantifying existing groundwater-surface water interactions. Such data may be used to constrain the WAM and GAM predictions; however there are numerous methodologies, each with advantages and disadvantages in certain circumstances. Recognizing that a uniform approach would not be reasonable, TPWD recommends that TWDB develop a short guidance document on appropriate methods for estimating groundwater/surface water interactions. Additionally, TPWD recommends that future GAM and WAM reports briefly mention each method and present the rationale for the method(s) chosen.
GAM deliverable reports require that Section 4.5 of the report identify and quantify reaches of streams or rivers with net gains or losses and incorporate the TWDB funded research on surface water/groundwater interactions (Slade and others, 2002) into the analysis.
11. In counties with overlapping GAMs, the TWDB might consider telling the Regional Water Planning Groups which GAM is considered the most accurate and, in essence, make the decision for the groups as to which GAM to use. This was a comment that many participants agreed would be useful.
In the overlap area of GAMs the appropriate model to choose depends on how close to the edge of each of the models that major pumping occurs. We did note that in the case of the northern and central part of the Gulf Coast aquifer models, using the central part of the Gulf Coast aquifer GAM was encouraged because of concerns with pumpage used to calibrate the northern part of the Gulf Coast aquifer GAM.See northern part of the Gulf Coast aquifer GAM; central part of the Gulf Coast aquifer GAM; southern part of the Gulf Coast aquifer GAM; northern part of the Carrizo-Wilcox, Queen City, Sparta aquifers GAM; central part of the Carrizo-Wilcox, Queen City, Sparta aquifers GAM; and the southern part of the Carrizo-Wilcox, Queen City, Sparta aquifers GAM.
12. Relating to the above question, there was a general question as to whether the differences between overlapping GAMs had been resolved, for example the central and northern Gulf Coast GAMs.
Not yet. As soon as resources are available, GAM staff hope to re-evaluate the structure used in all of the Gulf Coast aquifer models and reassess the calibration of all three models. See northern part of the Gulf Coast aquifer GAM and central part of the Gulf Coast aquifer GAM.
13. Many participants agreed that the use of GAMs in the Regional Water Planning process went quite well. However, one Groundwater Conservation District suggested that the GAMs for Regions A and O (northern and southern Ogallala) needed refinement, with "hydrograph traces that were 30 to 50 feet off in some areas."
The southern part of the Ogallala aquifer is currently being revisited to add the Edwards-Trinity High Plains aquifer to the existing model. For a regional model, it is not unreasonable to have 'hydrograph traces' that are 30 to 50 feet off in some places. This is mainly due to the scale of the models. Areas with high topographic relief and/or steep groundwater gradients estimates the average values within the square mile area of the grid used in the model. See northern part of the Ogallala aquifer GAM or southern part of the Ogallala aquifer GAM.
14. Suggest looking at using "alluvial deposits" as supplies. There was a comment that this may be a surface water-groundwater interaction issue.
Several regions include 'alluvial deposits' as supplies. Some of the GAMs included various surficial alluvial deposits as part of the uppermost aquifer layer. Also alluvial river deposits are roughly factored into the model through the streambed conductance factor. See southern part of the Gulf Coast aquifer, central part of the Gulf Coast aquifer, Hueco Bolson aquifer, West Texas Igneous and Bolsons, and Presidio portion of the West Texas Bolsons, Edwards-Trinity Plateau(includes Cenozoic Pecos Alluvium).