7 Section 7: Land to Water

Size: px

Start display at page:

Download "7 Section 7: Land to Water"

Ralf Dorsey
6 years ago
Views:

1 7 Section 7: Land to Water 7.1 Introduction As discussed in Section 1, the multiple modeling approach permits P6 to represent processes on a finer scale than previous versions of the Watershed Model. Table 7-1 provides an overview of the transport processes for nutrients and sediment represented in P6. This section discusses the processes that control transport from field to edge of small order streams. For sediment, these processes represent hillslope transport that connect the edge-of-field losses, represented by the RUSLE2 land simulation targets, with the edge-of-small-streams (EOSS). These delivery factors are called interconnectivity factors in P6. As discussed in Section 1, nutrient land simulation targets do not represent edge-of-field nutrient export, but rather the average EOSS nutrient export, without regard to variation in nutrient delivery. In P6, the variation in delivery is represented by the Delivery Variation Factors (DVFs). These are calculated based on USGS SPARROW simulations of the Chesapeake Bay. Because of the key role SPARROW plays in determining the DVFs and the delivery factors for nutrient in small streams, discussed in Section 9, this section will open with an extended discussion of SPARROW, before turning the DVFs themselves. The planned implementation of the interconnectivity factors, which were not used in the Beta 1.0 version of P6, will then be briefly discussed, followed by a brief discussion of the possible roles that wetlands may play in subsequent version of P6. Table 7-1: Transport Processes Represented in the Phase 6 Watershed Model Process P6 Nutrients P6 Sediment (non-urban) P6 Sediment (urban ) Edge-of-Field Land Simulation Target + RUSLE2 targets RUSLE2 targets Hillslope DVFs Interconnectivity Factors Interconnectivity Factors Small Stream S2R factors USGS Regression Models of Streambank Erosion and Floodplain Deposition + LIDAR Stream Source Ratio Large River River simulation River simulation River simulation 7.2 Sparrow SPARROW is a non-linear regression model which predicts constituent fluxes on the basis of reach and catchment attributes. SPARROW can best be explained by example. In this case, it is convenient to choose the latest version of the SPARROW models of total nitrogen and total phosphorus loads in the Chesapeake Bay watershed, CBTN_v4 and CBTP_v4, respectively, because results from these specific models were incorporated in the Prototype. As the model names suggest, these are the fourth versions of SPARROW models of the Chesapeake Bay watershed. Ator et al. (2011) document the development of the models and analyze their results in detail. 7-1

2 The catchments and reaches used in CBTN_v4 and CBTP_v4 are taken from the National Hydrography Dataset Plus (NHDPlus), version 1.1. NHDPlus catchments and river reaches are delineated at a much finer scale than P6. Over 80,000 reaches and catchments are represented in NHDPLus in the Chesapeake Bay watershed. The average catchment size is about 500 acres. Figure 7-1 illustrates the difference in scale between NHD and P6. It shows the NHDPlus reaches and catchments for Montgomery County, MD. Montgomery County is a single land segment in the P6 model. The darker blue lines show the river P6 reaches represented in the county. Figure 7-1:Comparison of NHDPlus Catchments and Watershed Model Land and River Segments Reach and catchment attributes are the independent variables used in the non-linear regression model of nitrogen and phosphorus fluxes in the Chesapeake Bay watershed. These attributes can be divided into three groups: (1) sources of nutrients; (2) attributes which control the transport of sources from land to water; and (3) reach characteristics which determine nutrient losses (aquatic decay) in the reach network. The nutrient load in reach i is determined by the following equation (Preston and Brakebill, 1999): Equation 7-1: Sparrow 7-2

3 where N L i = β n s n,j e ( αz j) e ( δt i,j) n=1 jεj(i) L i = load in reach i; n,n = source index where N is the total number of considered sources; J(i) = the set of all reaches upstream and including reach i, except those containing or upstream of monitoring stations upstream of reach i; β n =estimated source parameter; s n,j = contaminant mass from source n in drainage to reach j; α = estimated vector of land-to-water delivery parameters; Z j = land-surface characteristics associate with drainage to reach j; δ = estimated vector of instream-loss parameters; and T I,j = channel transport characteristics. The reach and catchment attributes used in CBTN_v4 and CBTP_v4 are shown in Table 7-1. These attributes represent conditions in The α s, β s, and δ s are estimated through non-linear regression. These parameters are adjusted to minimize the sum of the square differences between modeled fluxes and empirically calculated mean annual fluxes. The empirically calculated fluxes are determined using the USGS software FLUXMASTER. FLUXMASTER estimates concentrations based on the following linear regression model: Equation 7-2: FLUXMASTER C t = γ 0 + γ 1*q t + γ 2*q 2 t + γ 3*T t + γ 4*T 2 t + γ s*sin(2πt t) + γ c*cos(2πt t) + e t where C t = natural log of concentration at time t; q t = natural log of daily average flow at time t; T t = time in years as decimal; e t = error term; and γ s = estimated coefficients FLUXMASTER uses Tobit regression to treat censored concentration values and corrects for retransformation bias (in converting back from natural log units) using a method that approximates that used in the USGS software LOADEST (Runkle et al., 2004). FLUXMASTER was used to calculate mean annual detrended nitrogen and phosphorus loads, using water quality data from 1994 through Loads were adjusted to reflect mean hydrologic conditions over a 30-year flow period. Thus although SPARROW attributes represent watershed conditions in 2002, model loads represent long-term conditions centered around that year. There were 181 FLUXMASTER load estimates used to calibrate the parameters in the nitrogen model and 184 load estimates used in the phosphorus model. Table 3 shows the estimated parameters. It should be noted that the land-to-water catchment variables are input into the regression model centered around their average values. This changes the interpretation of their effects. The overall effect 7-3

4 of the land-to-water variables has been called the delivery variation factor (DVF) (Hoos and McMahon, 2009): Equation 7-3: Delivery Variance Factor DVF i = exp ( α *Z i) where α = vector of estimated land-to-water coefficients: and Z i = vector of land-to-water catchment variables for catchment i, centered around average value for watershed. When the catchment variables equal the average value of the variables, DVF is equal to one; correspondingly, when the variables on the whole are greater than the average value, DVF is greater than 1, and when they are less than the average values, DVF is less than 1. Therefore, DVF is not a true delivery factor, but measure the deviation from average effects of transport from land to reaches. The overall effects of aquatic decay, however, can never be greater than one. For phosphorus, decay only takes place in impoundments. For nitrogen, in addition to impoundments, decay takes place in reaches representing streams and rivers. The decay rate is a function of travel time, and varies with the size of the stream (as measured by mean annual flow) and average maximum temperature, 1971 through

5 Table 7-2: Estimated Coefficients from SPARROW Nitrogen and Phosphorus Models of the Chesapeake Bay Watershed, Version 4 Type Nitrogen Phosphorus Variable Coefficient Variable Coefficient Source Point sources (kg/yr) Point sources (kg/yr) Crop fertilizer (kg/yr) Crop fertilizer (kg/yr) Manure (kg/yr) Manure (kg/yr) Atmospheric deposition (kg/yr) Siliciclastic rocks (km 2 ) 8.52 Urban (km 2 ) Crystalline rocks (km 2 ) 6.75 Urban (km 2 ) 49.0 Land-to-Water ln(mean enhance vegetative index) Soil erodibility (K factor) 6.25 Aquatic Decay ln(mean soil available water capacity ln(mean groundwater discharge) (mm) ln(percent well-drained soils) Percent Coastal Plain 1.02 ln(percent Piedmont carbonate) Ln(precipitation) (mm) 2.06 Impoundments: Inverse hydraulic load (yr/m) Small streams 1 : travel time (d) large streams, cool temperature 2 : travel time (d) large streams, warm temperature 3 : travel time (d) 5.93 Impoundments: Inverse hydraulic load (yr/m) mean average flow 3.45 m 3 /s 2 mean average flow > 3.45 m 3 /s; mean annual maximum temperature > 18.5 C 3 mean average flow > 3.45 m 3 /s; mean annual maximum temperature 15.0 C SPARROW Simulation with Sector Categories as Sources The USGS performed new SPARROW simulations of nitrogen and phosphorus in the Chesapeake Bay watershed explicitly to help inform P6 global sector and targets and P6 DVFs and S2R delivery factors. These simulations used the acreage of the broad sector categories cropland, pasture, developed land, and natural land as source categories, in place of the original source categories in the CB_V4 SPARROW models. The only source category retained from CB_v4 was point sources, though estimates of point source loads were updated using information from P6. Combined sewer overflows were also added as a source which, like point sources, is directly applied to river reaches. Like the SPARROW CB_v4 models, the new SPARROW simulations were set up to simulate inputs under 2002 conditions. The 2002 P6 land use, which is tabulated at land-river segment scale, was disaggregated to the NHDPlus scale appropriate for inputs into SPARROW. The land-use disaggregation was based on the 2011 land use rasters prepared by Peter Claggett of the USGS for calculating the P6 land use, as described in Chapter 5. Assignment of P6 land use to catchments was based on the ratio of 2011 land use in the catchment to the land use acreage in the catchment. This method of disaggregating 7-5

6 land use does not necessarily preserve catchment area, but was deemed appropriate, since the land use acreages are only being used as sources of nutrients, and not catchment areas, which are derived directly from NHDPlus. Table 7-3 gives the coefficients estimated by SPARROW in the simulations using land uses as sourced. The coefficients SPARROW calculates for these land-use sources provide an estimate of the average export rate of nutrients (in kg/km 2 /yr) across the Chesapeake Bay watershed. These nutrient export rates were used to estimate the ratio of nutrient export among the sector categories as described in Section 2 Table 7-3: Estimated Coefficients from SPARROW Nitrogen and Phosphorus Models of the Chesapeake Bay Watershed, P Land Use Acreage as Sources Nitrogen Phosphorus Type Variable Coefficient Variable Coefficient Source Point sources (kg/yr) Point sources (kg/yr) Cropland (kg/ha/yr) Cropland (kg/ha/yr) Pasture (kg/ha/yr) Pasture (kg/ha/yr) Developed land (kg/ha/yr) Developed land (kg/ha/yr) Natural (kg/ha/yr) Natural (kg/ha/yr) CSO (kg/yr) Land-to-Water ln(mean enhance vegetative Soil erodibility (K factor) index) ln(mean soil available water ln(percent well-drained capacity soils) ln(mean groundwater Percent Coastal Plain discharge) (mm) ln(percent Piedmont carbonate) Ln(precipitation) (mm) Aquatic Decay Impoundments: Inverse hydraulic load (yr/m) Impoundments: Inverse hydraulic load (yr/m) Small streams 1 : travel time (d) large streams, cool temperature 2 : travel time (d) large streams, warm temperature 3 : travel time (d) 1 mean average flow 3.45 m 3 /s 2 mean average flow > 3.45 m 3 /s; mean annual maximum temperature > 18.5 C 3 mean average flow > 3.45 m 3 /s; mean annual maximum temperature 15.0 C Land-to-Water Delivery Variation Factors Land-to-water (L2W) delivery variation factors (DVFs) represent the effect of transport processes from nutrient application or generation to delivery to small-order streams. The DVFs are derived from 7-6

7 SPARROW. SPARROW estimates coefficients for catchment properties which can impact transport. Coefficients from the SPARROW simulations using the sector categories were used to calculate the DVFs in the Beta 1.0 version of P6. These properties and their coefficients are shown in Table 7-3. In the Prototype, DVFs are applied at the LRS scale. A DVF at the LRS scale is the average DVF at the NHDPlus catchment scale, weighted by the area of the catchment in the LRS, according to the formula where DVF LR = DVF at LRS scale DVF i = DVF in NHDPlus catchment i A i = Area of catchment i in LRS A LR = Area of LRS N DVF LR = DVF i A i /A LR i=1 N = number of catchments wholly or partially in the LRS This is equivalent to assuming the loads from a LRS are uniformly distributed over the segment, i.e. there is no correlation between variation in loading rates introduced by variation in land use and the variation in L2W factors over the segment. The Modeling Team is currently investigating whether DVFs can be calculated on a land use basis within a LRS, so that variations in L2W factors with land use can be captured in P6. As discussed above, the individual L2W factors are centered around their average values, so the DVF measures the effects of transport as they deviate from average conditions. For this reason, the DVFs do not behave like sediment delivery factors which estimate the delivery from edge-of-field (EOF) to EOS. Similarly, the land simulation targets tacitly assume average transport conditions and therefore do not represent EOF loads. Only the product of the target and the DVF has physical meaning as the EOSS load. For phosphorus, land simulation targets incorporate sensitivity to runoff and erosion into the export targets for specific land segments. The sensitivity to runoff and erosion captures the impact that the precipitation and erosivity land-to-water factors has on phosphorus transport. For that reason, these two factors were dropped from the calculation of phosphorus DVFs on most land uses. In other words, the phosphorus DVFs were calculated based on the percent land in the Coastal Plain and percent welldrained soils. All four land-to-water factors were used to calculate DVfs for animal feeding operations, where the sensitivity to runoff or erosion was not used to set the land simulation targets. Figures 7-2 and 7-3 show the DVFs on the LRS scale for nitrogen and phosphorus, respectively. 7-7

8 Figure 7-2: Nitrogen Delivery Variation Factors, P6 Prototype 7-8

9 Figure 7-3: Phosphorus Delivery Variation Factors, P6 Prototype 7-9

10 7.4 Interconnectivity Factors and Sediment Delivery Ratios The Beta 1 version of P6 uses sediment delivery factors (SDFs) to calculate the losses in transport between edge-of-field and the river reaches represented in P6. The SDFs are calculated the same way as the Phase 5 Model. The Phase 5 Model uses a formula from the NRCS (SCS, 1983) to calculate SDFs where SDF = * A A = watershed area (square miles) SDFs are calculated by land-river segment and land use. For each land use, the mean distance between land use parcels in a segment and the river reach is calculated, and the watershed area is set equal to the area of a circle with radius equal to the mean distance. As discussed in Section 9, future versions of P6 will use explicitly calculate net sediment transport in small streams. These versions will use interconnectivity factors, which are sediment delivery factors that represent the effect of hillslope transport or the transport between edge-of-field and edge-of-smallstreams. The current version of interconnectivity factors is based on the NRCS s equation for sediment delivery factors used in the Phase 5 Watershed Model. For interconnectivity factors, the area is determined from distance between the land use pixel and the NHDPlus reach. Efforts are underway, however, to incorporate slope and the location of the slope in the landscape into future calculation of the interconnectivity factors. 7.5 Wetlands Wetlands are represented in P6 as land uses but an effort is also underway to represent their impact on nutrient loads originating on other land uses. The cornerstone of this effort is the work by Jordan et al. (2008), who developed a simple formula for calculating the removal efficiency of restored or created wetlands: where Removal = 1 e k(area) Removal = fraction of nitrogen or phosphorus input load removed by the wetland area = the proportion of the watershed occupied by wetlands k = fitted parameter Based on an analysis of reported nutrient reductions in the scientific literature, Jordan et al. estimated that k equals 7.90 for nitrogen and 16.4 for phosphorus. Area, as fraction of the watershed, occupied by wetlands, is a surrogate for detention time. Increasing detention time increases removal rates by increasing the time nutrient loss mechanisms like denitrification, deposition, and biological uptake effect nutrient inputs. 7-10

11 The USGS is currently planning on calculating a removal efficiency on the NHD catchment scale for every wetland in the Chesapeake Bay watershed. The efficiency would be based on applying Jordan s et al. s formula to the watershed upstream of each wetland as far as the next upstream wetland. In other words, if there is a wetland upstream, the watershed associated with the downstream wetland would terminate at the upstream watershed. In this manner a removal efficiency can be applied to each NHD catchment upstream of a wetland. The explanatory value of the proposed wetland removal efficiencies will be tested by using them as land-to-water inputs to a SPARROW simulation. The SPARROW simulation will also help determine to what extent wetland effects are already accounted for by the other land-to-water factors contributing to the overall DVF. The impact of wetlands may also be confounded with the effect of forested buffers. SPARROW simulations used to test the effect of wetlands will also use percent unbuffered cropland as potential explanatory factor to attempt to separate the impacts of wetlands and forest buffers. 7-11

Nutrient Loading: NAWQA Regional SPARROW model

Nutrient Loading: NAWQA Regional SPARROW model Pee Dee River Basin and Winyah Bay Estuary Waccamaw Water Quality Data Conference Wednesday, September 19, 2013 Celeste Journey, Water-Quality Specialist