Scientific Investigations Report 2007–5007
U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2007–5007
Back to Table of Contents
Potential sources of error in the estimates of recharge are described by following the flux of water through the land-surface system as calculated by the models. This description starts with the weather data and ends with the water leaving the root or soil zone.
Both errors in the daily weather data and the interpolation of weather data to HRUs produce errors in recharge calculations (Bauer and Vaccaro, 1990). In the uplands, the total annual runoff for a watershed (subbasin) was simulated to be lower in some years and higher in others compared to the observed/estimated value (Mastin and Vaccaro, 2002a); the differences were mainly attributed to the weather data and the interpolation of the data to HRUs. Thus recharge was either under- or over-estimated for such years. However, the simulated and estimated mean annual runoff volume for the basin were similar (Mastin and Vaccaro, 2002a) and the mean annual water budget under predevelopment conditions was reasonable (mean annual recharge was similar to mean annual discharge). Mastin and Vaccaro (2002a) used regression analysis to estimate that about 165 ft3/s of the mean annual streamflow under predevelopment conditions was produced by the 14 areas modeled with DPM, which is consistent with the mean annual recharge estimate of 187 ft3/s, assuming that most of the recharge ultimately becomes streamflow for these areas. At shorter time scales (daily to annual), both the spatial and temporal distribution of recharge will contain some unknown amount of error due to the weather data and its interpolation.
Estimates of irrigation application rates are another source of error because their quantities are similar to or greater than the precipitation quantities in some areas. The I/P ratio ranged from 0.2 to 5.9 and averaged 1.96 for the DPM-modeled areas, and only four areas had I/P ratios of less than 1.0. Therefore, errors in the estimates of irrigation application rates, which averaged about 82 percent of irrigation diversions, can have a large effect on the estimated recharge. Detailed information on irrigation operations (diversion, operational spill, canal/lateral loss, and deliveries) was available for several years for one irrigation district, and the information indicated that for the later years of simulation the effective application rate might be 10 percent too large. If application rates were similarly overestimated in other areas, the estimated recharge rate for the DPM-modeled areas with surface-water irrigation may be more than 1 in. too large. The application rates were assumed to be constant over the period of simulation and, although this captures the long-term mean annual recharge, there will be errors for years when the rate was either lower or higher than average. For example, in the severe drought of 2001, much less water than usual was applied in several of the irrigation districts with junior water rights, and thus the recharge estimates would be too large. Conversely, the recharge may be too small during the early years of simulation (typically prior to 1977) because there was a plentiful supply of water for irrigation; differences between annual calculated unregulated and regulated streamflow increase as the unregulated flow increases, indicating that more water is diverted and lost in wet years (Vaccaro, 1986). Assuming that the ratio of R/T for an area is applicable if more or less water is applied, and adding or subtracting 4 in. (the range in the potential crop water use for 1950–2003) to the application rate, suggests that using a constant application rate can result in errors of about 10 percent in some years if the application rate was 4 in. too low or too high. In more extreme drought years such as 1992, 1994, and 2001, the recharge estimate may be as much 20 percent too high.
The mean monthly precipitation values for the HRUs were based on a national database of grids that are 4- by 4-km cells (PRISM, Daly and others, 1998), leading to large differences across the boundaries of those grids in areas with large gradients in monthly precipitation. In turn, there are similar changes in the interpolated precipitation values at the HRUs near the grid boundary. As a result, there are differences in calculated recharge values across these boundaries, especially for predevelopment estimates of recharge in the semiarid to arid parts of the study area (see for example, fig. 11). The total estimated mean annual recharge for these areas is reasonable but the spatial distribution will be in error—some areas have recharge values that are too low and other areas have values that are too high. This type of error is a discretization problem due to differences in scale.
The next potential source of error in recharge estimates is the assignment of the LULC to a HRU. For example, under natural conditions, in a typical plant community throughout much of the non-forested parts of the study area, sagebrush and grasses are generally intermixed, with grasses starting to predominate in the forest-transition zone. For predevelopment conditions, the DPM HRUs with human influences (agriculture, urban areas, etc.) were assigned a LULC of sagebrush. Sagebrush, being more deeply rooted than grasses, uses more water, and thus, recharge is less. If a HRU was predominantly grassland instead of sagebrush, the predevelopment recharge estimate would be too small. For current conditions, the LULC was based on the composite LULC database, and several factors can lead to errors in the estimated recharge. Assignment of a low water-use crop type in contrast to a high water-use crop (and the opposite) can lead to both over-estimated and under-estimated recharge and the amount would be dependent on the application rate. Although some error is introduced due to assigning a HRU the wrong LULC, more error is due to assuming that the composite database was valid for all years. The composite distribution does not capture the spatial and temporal changes in crop types and amount of irrigated lands. The potential error would be greater in calculations for the earlier years and less for the surface-water irrigated areas established prior to 1950 and greater for ground-water irrigated areas because of the increase in acres irrigated with ground water since 1950.
After calculating the amount of precipitation and irrigation water intercepted by plants, soil column calculations are performed on any remaining water (including snowmelt). Therefore, the next important source of error is due to the soil information for each HRU. The largest potential source of error is due to the differences between the generalized STATSGO data and the detailed SSURGO data. The SSURGO data provides more detail both spatially and with depth. STATSGO retains the effective information for the mapped soils in the survey areas, but the actual spatial distribution is coarse. Thus, the total recharge for the models that used the STATSGO data should approximate the actual recharge but there would be errors in the spatial distribution; this is another discretization problem. Differences in hydrologic properties for the same soil-mapping unit occur across some of the boundaries of the different SSURGO survey areas. For example, the same soil-mapping unit may have a high TAWC in one survey area but a low TAWC in the adjacent area, or the soil depths may vary; in either case, calculated recharge would be different across those boundaries. It was beyond the scope of this study to rectify these differences, and thus there are differences in calculated recharge across these discontinuities. Lastly, to simplify the data management and input for estimating recharge for such a large study area, an ‘effective’ depth-weighted TAWC was calculated for each soil type. This averaging of the TAWC may introduce some error into the estimated recharge, especially for predevelopment conditions in the semiarid to arid areas. For example, if a clay soil layer overlies a sandy soil layer, more water would be retained in the upper clay soil and thus more water would be used by evapotranspiration than calculated by using an averaged value. However, analysis of the layering for selected soil types suggests that any error from such averaging is not large.
The amount of water leaving the root or soil zone is a function of the vertical infiltration rate below the soils for a HRU—the basal (subsoil) rate or infiltration capacity. The rate was estimated to be high for soils overlying coarse-grained deposits and low for soils overlying bedrock. The rate was determined from the depth-to-bedrock information in the soil databases. Shallow soils were assigned a low infiltration rate and deep soils (typically no bedrock) were assumed to overlie deposits with a high infiltration rate. Errors in the identification of this HRU parameter can lead to errors in recharge. In some areas, deep soils may overlie clayey deposits and consequently, the recharge estimate would be too high; in other areas soils may overlie bedrock units with a high infiltration capacity and the recharge estimate would be too low. The potential error in the recharge estimate from misidentification of the infiltration rate is not known, but for much of the area included in the DPM models, the type of underlying deposits is reasonably defined.
Considering all sources of error and the known sensitivity of recharge to different factors (Bauer and Vaccaro, 1990), the error in the recharge estimates for the modeled areas is uncertain. However, mean annual recharge for predevelopment conditions is consistent with the information presented in Mastin and Vaccaro (2002a), and the error for the basin-wide mean annual predevelopment conditions is considered to be on the order of 10 percent. The estimated difference between the mean annual streamflow leaving the basin for predevelopment conditions compared to current conditions is about 2,000 ft3/s, and the estimated increase in mean annual AET from predevelopment to current conditions is about 1,720 ft3/s, suggesting about a 15 percent error in the estimated total mean annual recharge for current conditions. Area-averaged annual recharge values for predevelopment and current conditions likely have an error on the order of 25 percent.
The relative error in recharge values increases with an increase of spatial and temporal resolution. Although the overall spatial distribution of the estimated annual recharge in any modeled area is reasonable, the error in the estimated recharge for a particular HRU can be quite large. Similarly, estimates of daily to monthly recharge likely have larger relative errors than estimates of annual recharge.
Back to Table of Contents