Model specifications for place-specific water supply models developed for the San Pedro and Rocky Mountains case studies

Water supply is a complex ecosystem service to spatially model, given that groundwater and surface water are closely connected but move based on different controlling factors, that these influences on hydrology and hydrologic models differ greatly based on the spatial and temporal scale and the region of the world considered, and that available spatial data can rarely support modeling at both high spatial and temporal resolutions. Given these limitations, our initial water supply models operate at an annual scale (which is matched by available spatial data for important variables including precipitation, infiltration, snowmelt, and evapotranspiration). We currently consider only flows of surface water, though we do model the infiltration of surface water into groundwater and groundwater extraction from wells. We also use a set of generalized models to represent sources of surface water (precipitation, snowmelt, springs, baseflow to rivers, and incoming inter-basin water transfers), sinks of surface water (evapotranspiration, infiltration), beneficiaries or users of surface water, and the flow of surface water across the landscape (routed using SRTM elevation data). The long-term intent of our modeling process is to incorporate existing hydrologic models wherever appropriate that will more realistically account for hydrologic processes. Over time this includes modeling groundwater and its movement and use. As spatial data continue to improve, it may also be possible to model water supply at finer temporal scales. Please see the ARIES modeling guide for full documentation and references for these models.


Surface water source: Areas where surface water enters the landscape and becomes available for human use, via precipitation, snowmelt, springs, baseflow to rivers, and incoming inter-basin water transfers.
Surface water sink: Areas that where surface water transitions to groundwater (via infiltration) or atmospheric water (via evapotranspiration).
Surface water beneficiaries: Human users of surface water, obtained via water intakes or surface diversions.
Surface water flow: Surface water flows are routed across the landscape through topography and stream networks.
Groundwater source: Areas where groundwater enters the water table and becomes available for human use, via infiltration or artificial groundwater augmentation.
Groundwater sink: Areas where groundwater transitions to surface water (via springs or baseflow).
Groundwater beneficiaries: Human users of groundwater, obtained via groundwater wells.
Groundwater flow: Groundwater flows, which depend on subsurface geology (e.g., porosity of rock layers) and groundwater elevations.

Model structure and assumptions

Case studies 1. San Pedro River Watershed, Arizona and Sonora 2. Rocky Mountains

Water supply source models. Past studies have used a variety of spatial data to map water supply and regulation services on the landscape. These have typically included overlays of supply and demand (Boyd and Wainger 2003, Wundscher et al. 2008), estimates of water stored in soils and aquifers using infiltration data (Egoh et al. 2008), precipitation and evapotranspiration data (Chan et al. 2006), or the SCS curve number (SCS 1972, Gately 2008) or Budyko Curve method to account for precipitation and evapotranspiration across the landscape (Tallis et al. 2011). Given the difficulty in developing a generalized model of hydrologic processes that is applicable at multiple spatial scales and in different ecological contexts, the initial ARIES water supply models include direct data or Bayesian models (for surface water sinks) that are applicable to our case study regions but that incorporate many of the influences on hydrologic processes that were used by the above authors. In cases where vegetation-hydrology relationships are poorly understood, such as in tropical forests (Bruijnzeel 2004), ARIES’ data-driven modeling approach may be more appropriate than using process-based approaches. In many other cases, future generation ARIES models will link existing hydrologic models, improving model quality and credibility. Spatial data or calibrated hydrologic model outputs can generally be used as the source value for surface and groundwater supply, with no Bayesian model needed. There are at least five potential sources of surface water, which can be summed to obtain the total annual surface water source value: precipitation, snowmelt, springs, baseflow to rivers, and incoming inter-basin water transfers where water is discharged into surface water bodies. If we run the model using annual average values, snowmelt only becomes important in locations with glaciers (i.e., annual snowmelt in all other locations is included in annual precipitation totals). Sources of groundwater include areas of infiltration and deep percolation that lead to aquifer recharge, along with artificial groundwater recharge. For the San Pedro, we use annual precipitation as the source value for surface water. We show initial results for a representative dry (2002) and wet year (2007), since the 30-year average data from PRISM is less meaningful in arid environments where annual precipitation is highly variable. If desired, a user could also input precipitation data from other years to use as the surface water source value. For groundwater, we can compare spatial data for soil infiltration, infiltration results from the surface water sink Bayesian network model, and the results of hydrologic models (once incorporated) as possible source values. In the future, we could also incorporate data on the location of groundwater recharge facilities in the Sierra Vista area, assuming the data were available. We do not include snowmelt in the source model, as there is no persistent snowpack in the mountains within the San Pedro River Watershed. Until detailed surface and groundwater models are incorporated, we lack data on baseflow. Although incoming interbasin water transfers are proposed (Bureau of Reclamation 2007), there are currently no incoming water transfers from outside the basin. Finally, while we have data on the location of springs in the San Pedro, we do not use these data in the source model as we do not know their discharge volume, and most spring discharge quickly infiltrates back into the soil via ephemeral stream channels. Glacial extents in the Rocky Mountains total approximately 73 km2 in Wyoming and 4.8 km2 in Colorado. Additionally, numerous interbasin water transfers have been constructed, particularly to supply water for Colorado’s Front Range communities from the western slope of the Continental Divide. ARIES’ surface water source models for the Rocky Mountains currently account for annual precipitation and interbasin water transfers but not glacial melt. For the San Pedro, there is no persistent mountain snowpack. We also lack detailed hydrologic data or models on springs, baseflow, incoming interbasin water transfers, and groundwater recharge, so do not include these as sources of surface or groundwater.

Water supply sink models. Surface water sinks include areas of evapotranspiration and infiltration. Conversely, groundwater sinks include springs and baseflow to rivers. Lacking an external groundwater model, we currently do not include springs or baseflow as groundwater sinks for the San Pedro. For all case studies, we set the total surface water sink as the sum of evapotranspiration and deep soil infiltration. Runoff data will play a role in training of the Bayesian network models to help account for the difference between precipitation and sinks. For the U.S.-based case studies, nationwide data are available for deep soil infiltration and global data for actual evapotranspiration. While these data sources can be used as training data for Bayesian network models, both datasets are problematic. The evapotranspiration dataset has low spatial resolution (0.5 x 0.5 degree) and does not capture local variation in vegetation type, tree canopy cover, and temperature, all of which are key influences on evapotranspiration. The infiltration data, having been developed at the national level, are unlikely to account for the limited area over which infiltration actually occurs in the semiarid Basin and Range region of the southwestern United States. For the San Pedro and Rocky Mountain case studies, we therefore use a Bayesian network that considers vegetation type, percent tree canopy cover, and annual maximum temperature as influences on evapotranspiration. For the San Pedro, we set the locations of stream channels and limestone bedrock and the intersection of valley fill alluvium and the mountain fronts to account for the two key locations of deep percolation and groundwater recharge: the mountain fronts, stream banks, and ephemeral stream channels (Pool and Dickinson 2007). In the future, we could also incorporate data on the location of groundwater recharge facilities in the Sierra Vista area, assuming the data are available. For the Rocky Mountain case study, we set deep infiltration as a function of hydrologic soils group, slope, and percent impervious surface cover. We set priors for all nodes based on a review of the corresponding spatial data. We set the highest values for the evapotranspiration in the San Pedro and Rocky Mountain conditional probability table under greater percent tree canopy cover and higher temperatures, all else being equal. We set the highest evapotranspiration rates for vegetation type to riparian, followed by forests, then mesquite woodland, oak woodland, agriculture, urban, and grassland, with the lowest values set for desert scrub. In the San Pedro case study, we set the conditional probability for infiltration as highest at the mountain fronts and as slightly lower in stream channels, and set it as extremely low elsewhere. We assume infiltration to be greatest for the Rocky Mountain case study on shallow slopes, low levels of impervious surface cover, and hydrologic soils groups A and B.

Water supply use models. Users access groundwater through wells and surface water through surface diversions, direct pumping from water bodies, and outgoing inter-basin water transfers. Users can be split by use type (e.g., agriculture, domestic, industrial use) if deemed relevant to the case study of interest. For the San Pedro, we mapped the location and volume of the two surface water diversions on the river at St. David and Pomerene. We use well data and capacity to identify groundwater use. Although the state wells database identifies users, they are not explicitly grouped by use, so at this time we do not separate out agricultural, mining, military, or domestic water uses. Also, since we do not currently have an integrated groundwater flow model, we do not explicitly connect sources, sinks, and users for groundwater. Water use data for the Rocky Mountains included agricultural use, domestic use, and interstate compacts. Per capita residential water use data was combined with Census population data to estimate residential water demand. Agricultural parcel data was combined with crop water use data to generate an agricultural water demand layer. Finally, for Colorado, whose rivers supply water to many downstream states, we estimated the quantity of water obligated to other states as part of legally binding interstate compacts. Well data and well capacity could be used to identify groundwater use. In either case, legally binding water rights would also be informative for further identifying beneficiaries and use. We mapped four distinct beneficiary classes, including agriculture, aquaculture, industrial, and residential water use.

Water supply flow models. The source models determine the annual quantity (in mm3/yr) of precipitation and other surface water sources (in the surface water models) or infiltration to groundwater (in the groundwater source models). The sink models estimate the annual quantity of water transitioning between surface and groundwater, and vice versa, and the use models estimate the quantity of water used by beneficiaries in each location. Surface and groundwater flows must be modeled separately, as they move at different rates, with flows governed by different factors. Currently, we map surface water flow using a simple water routing model. This model relies on the SRTM elevation data to identify flow directions for water. Water is moved across the landscape using this derived flow direction layer until it encounters a stream (represented using a hydrography layer), at which point it moves downstream through the stream network. Users or sinks encountered in transit reduce the quantity of water carried across the landscape. Subsurface water flows are considerably more complex, and are governed by factors including geology (i.e., porosity of rock layers) and groundwater elevations. Subsurface flows are commonly modeled using the MODFLOW model (Harbaugh et al. 2000). Future releases of ARIES will investigate the feasibility of linking groundwater models to source, sink, and use models to fully and more accurately represent water flows using accepted hydrologic models. Key outputs from the flow models include: 1. Possible surface water flow: The movement of surface water via topography and stream networks, and groundwater via appropriate groundwater flow paths while disregarding sinks. 2. Possible surface water supply: Atmospheric, ground, or surface water transitions providing an initial source quantity of surface or groundwater, that are capable of providing water to human beneficiaries when accounting for surface or groundwater flow paths but not sinks. 3. Possible surface water use: Water actually reaching a user, but not accounting for the activity of sinks. 4. Actual surface water flow: The movement of surface and groundwater, accounting for flow topology and sinks. 5. Used surface water supply: Atmospheric, ground, or surface water transitions that result in an initial source quantity surface or groundwater, that are capable of providing water to human beneficiaries when accounting for surface or groundwater flow paths and sinks (i.e., locations actually providing water to human beneficiaries). 6. Actual surface water sink: Locations where surface water transitions into groundwater (via infiltration) or atmospheric water (via evapotranspiration), or where groundwater transitions into surface water (via springs or baseflow). 7. Satisfied surface water demand; Satisfied groundwater demand: The portion of demand for water satisfied by extraction of surface or groundwater. 8. Sunk surface water flow: Water flow that fails to reach a user because it encountered a sink and transitioned from surface or groundwater into the atmosphere, surface, or groundwater. 9. Sunk surface water supply: Source locations of surface or groundwater that fail to reach a user due to their encountering a sink and that transitions water to the atmospheric, surface, or groundwater. 10. Blocked surface water demand: Demand for surface or groundwater that goes unsatisfied due to the action of sinks that transition water between surface, atmospheric, and groundwater.

Spatial data

Models Data theme Source Spatial extent Spatial resolution Year
Source – All models Annual precipitation WorldClim Global 30 arc seconds 1950-2000
Source – San Pedro, Rocky Mountains Annual precipitation PRISM/Oregon State Univ. United States 800 m 1971-2000
Source – Rocky Mountains Interbasin water transfers Colorado Division of Water Resources Colorado 1 km n/a
Source & sink – San Pedro, Rocky Mountains Soil infiltration USGS Continental United States 1 km Derived from 1951-1980 runoff data
Springs Arizona Geographic Information Council Arizona Rasterized point data n/a
Source – All models Snowmelt Univ. Delaware Global Water Balance Archive Global 0.5 degree 1950-1999
Sink – All models Average annual actual evapotranspiration SAGE/Univ. Wisconsin Global 0.5 degree 1950-1999
Average annual runoff SAGE/Univ. Wisconsin Global 0.5 degree 1955-1990
Tree canopy cover GLCF/Univ. Maryland Global 1 km2 2000
Sink – San Pedro, Rocky Mountains Tree canopy cover NLCD 2001 United States 30 m 2001
Annual maximum temperature WorldClim Global 30 arc seconds 1950-2000
Annual maximum temperature PRISM/Oregon State Univ. United States 800 m 1971-2000
Mountainfront recharge zones Derived from Arizona Geographic Information Council Arizona Vector polygon data n/a
Sink – San Pedro, Rocky Mountains Vegetation type Southwest Regional Gap Analysis (SWReGAP) LULC AZ, CO, NM, NV, UT 30 m 2000
Sink – Rocky Mountains Vegetation type Northwest GAP Analysis (NWGAP) CA, ID, MT, OR, WA, WY 30 m 1999-2001
Sink & flow – San Pedro, Rocky Mountains Hydrography National Hydrography Dataset Arizona Vector line data n/a
Hydrography EPA San Pedro Data Browser Upper San Pedro in Sonora, Mexico Vector line data n/a
Sink – Rocky Mountains Slope Derived from global SRTM data Global 90 m n/a
Sink – Rocky Mountains Land cover NLCD 2006 United States 30 m 2006
Sink – Rocky Mountains Hydrologic soils group SSURGO soils data United States 30 m n/a
Use – Rocky Mountains Agricultural water use Irrigated parcels data combined with crop water use data Colorado 800 m n/a
Use – Rocky Mountains Interstate water compacts Colorado Division of Water Resources Colorado 1 km n/a
Use – Rocky Mountains Residential water use USGS per capita water consumption data plus Census population density estimates Colorado 800 m 2005
Use – San Pedro Surface diversions Digitized locations of St. David & Pomerene Diversions San Pedro Rasterized point data 2010
Well user type, capacity, depth Arizona Dept. of Water Resources Wells 55 Database Arizona Rasterized point data 2010
Flow – All models Elevation SRTM Global 90 m n/a



Boyd, J and L. Wainger. 2003. Measuring ecosystem service benefits: The use of landscape analysis to evaluate environmental trades and compensation. Discussion Paper 02-63, Resources for the Future: Washington, DC.

Bruijnzeel, L.A. 2004. Hydrological functions of tropical forests: Not seeing the soil for the trees? Agriculture, Ecosystems, and Environment 104: 185-228.

Bureau of Reclamation. 2007. Appraisal report: Augmentation alternatives for the Sierra Vista Sub-watershed, Arizona: Lower Colorado Region. U.S. Department of Interior Bureau of Reclamation: Denver, CO.

Chan, K.M.A., et al. 2006. Conservation planning for ecosystem services. PLOS Biology 4 (11): 2138-2152.

Egoh, B, et al. 2008. Mapping ecosystem services for planning and management. Agriculture, Ecosystems and Environment 127: 135-140.

Gately, M. 2008. Dynamic modeling to inform environmental management: Applications in energy resources and ecosystem services. MS Thesis, University of Vermont, Burlington, VT.

Harbaugh, A.W., et al. 2000. MODFLOW-2000, The U.S. Geological Survey modular ground-water model – User guide to modularization concepts and the ground-water flow process: USGS Open-File Report 00-92. USGS: Reston, VA.

Pool, D.R. and J.E. Dickinson. 2007. Ground-water flow model of the Sierra Vista Subwatershed and Sonoran portions of the Upper San Pedro Basin, southeast Arizona, United States, and northern Sonora, Mexico. USGS Scientific Investigations Report 2006-5228. USGS: Reston, VA.

Soil Conservation Service (SCS). 1972. National Engineering Handbook, Section 4, Hydrology. SCS: Washington, DC.

Tallis, H.T., et al. 2011. InVEST 2.0 beta User’s Guide. The Natural Capital Project: Stanford.

Wundscher, T., et al. 2008. Spatial targeting of payments for environmental services: A tool for boosting conservation benefits. Ecological Economics 65: 822-833.

Acknowledgements and additional contributors

An expert review panel including individuals from the U.S. Geological Survey, University of Arizona, Bureau of Land Management, and other organizations provided data and model review for the San Pedro case study. Initial ARIES data and models for the Rocky Mountains were developed by students participating in a graduate level ecosystem services modeling course taught in the University of Denver’s Department of Geography in the fall of 2011; James Reed, Darius Semmens, and Todd Hawbaker assisted with further data and model refinement.

Share This