Non-Point Source (NPS) • tbeploads

library(tbeploads)

Non-point source (NPS) estimates are obtained for gaged and ungaged locations in the watershed, then combined for the final result. Skip to the final sub-section to understand how to obtain all estimates together. Separate approaches for gaged and ungaged estimates are described first for documentation purposes.

Gaged locations

Gaged estimates are obtained using the anlz_nps_gaged() function that retrieves flow and water quality data, then combines them to calculate TN, TP, TSS, BOD, and hydrologic loads.

Required external flow data are Lake Manatee, Tampa Bypass, and Alafia River Bell Shoals. These are not available from the USGS API and must be obtained from the contacts listed in util_nps_getextflow(). USGS flow data are obtained from an API for stations 02299950, 02300042, 02300500, 02300700, 02301000, 02301300, 02301500, 02301750, 02303000, 02303330, 02304500, 02306647, 02307000, 02307359, and 02307498. A preprocessed USGS flow data frame can be provided to the usgsflow argument. The usgsflow data object is provided with the package to avoid re-downloading the data. Similarly, all flow data can be provided to the allflo argument. The allflo data object included with the package has both external and USGS flow data.

Water Quality data are obtained from the FDEP WIN database API, tbeptools, or local files as described in util_nps_getwq(). Chosen stations are ER2 and UM2 for Manatee County and station 06-06 for Pinellas County. Environmental Protection Commission (EPC) of Hillsborough County stations retained are 105, 113, 114, 132, 141, 138, 142, and 147. Manatee or Pinellas County data can be imported from local files using the mancopth and pincopth arguments, respectively. If these are not provided, the function will attempt to retrieve data from the FDEP WIN database using read_importwqwin() from tbeptools. The EPC data are retrieved using read_importepc() from tbeptools. The allwq data object included with the package has external data and USGS flow data and can used with the allwq argument.

The function assumes that the water quality data are in mg/L and flow data are in cfs. Missing water quality data are filled with previous five year averages for the end months, then linearly interpolated using util_nps_fillmiswq().

Water quality and flow inputs to anlz_nps_gaged() can be provided numerous ways:

Manatee and Pinellas County water quality data provided as local files and EPC data automatically retrieved from the API
All water quality data provided locally using a single object with the allwq argument (see the allwq dataset for the format, created using util_nps_getwq())
Flow data provided using local files for Lake Manatee, Bell Shoals, and Tampa Bypass, USGS data retrieved automatically from the API
Flow data provided using local files for Lake Manatee, Bell Shoals, and Tampa Bypass, USGS data provided locally using the usgsflow argument (see the usgsflow dataset for the format, created using util_nps_getusgsflow())
All flow data provided locally using the allflo argument (see the allflo dataset for the format, created using util_nps_getflow()).

In all cases, input data can retrieved from APIs with the exception of flow data for Lake Manatee, Bell Shoals, and Tampa Bypass (see the help file for util_nps_getextflow() for how to get these data). The following example uses combined water quality and flow data included with the package for convenience.

# external files included with the package
data(allwq)
data(allflo)

# get gaged NPS loads
nps_gaged <- anlz_nps_gaged(yrrng = c('2021-01-01', '2023-12-31'), allwq = allwq, allflo = allflo)
#> Estimating gaged NPS loads...

head(nps_gaged)
#> # A tibble: 6 × 13
#>   basin      yr    mo tn_mgl tp_mgl tss_mgl bod_mgl   flow h2oload tnload tpload
#>   <chr>   <dbl> <dbl>  <dbl>  <dbl>   <dbl>   <dbl>  <dbl>   <dbl>  <dbl>  <dbl>
#> 1 023005…  2021     1  0.862  0.222      NA      NA 4.71e9  4.71e6  4063.  1046.
#> 2 023005…  2021     2  1.13   0.358      NA      NA 9.10e9  9.10e6 10280.  3257.
#> 3 023005…  2021     3  0.941  0.327      NA      NA 4.62e9  4.62e6  4351.  1512.
#> 4 023005…  2021     4  0.964  0.44       NA      NA 8.04e9  8.04e6  7755.  3540.
#> 5 023005…  2021     5  0.334  0.286      NA      NA 2.07e9  2.07e6   692.   593.
#> 6 023005…  2021     6  0.871  0.341      NA      NA 4.12e9  4.12e6  3585.  1403.
#> # ℹ 2 more variables: tssload <dbl>, bodload <dbl>

In all use cases for the function, a data frame is returned with columns for basin, year, month, TN in mg/L, TP in mg/L, TSS in mg/L, BOD in mg/L, flow in liters/month, hydrologic load in m3/month, TN load in kg/month, TP load in kg/month, TSS load in kg/month, and BOD load in kg/month.

Ungaged locations

Ungaged (unmonitored basins) estimates are obtained using anlz_nps_ungaged(). The approach combines spatial land use data, rainfall patterns, hydrologic modeling, and empirical relationships to estimate monthly nutrient and sediment loads. The function requires combined spatial data for bay segment, basin, entity jurisdiction, land use data, and soils (obtained with util_nps_tbbase()), rainfall data (obtained with util_getrain()), and flow data (obtained with util_nps_getflow()).

The first step is updating the combined spatial data if any of the input datasets have changed. The required inputs are land use (tblu2023), soil data (tbsoil), jurisdiction (tbjuris), and sub-basin data (tbsubshed). The function util_nps_tbbase() combines these datasets into a single spatial object that is used for ungaged load estimation using util_nps_union(). The function requires GDAL to be installed and accessible in the system PATH, or the path to GDAL binaries can be provided using the gdal_path argument.

Land use and soil data can be updated using the util_nps_getswfwmd() function. These data are also stored internally with the package for easy retrieval, as are the remaining datasets.

tblu2023 <- util_nps_getswfwmd('lulc2023')
tbsoil <- util_nps_getswfwmd('soil')

Then, the combined layer, tbbase, can be created (takes an hour or two).

data(tbsubshed)
data(tbjuris)
data(tblu2023)
data(tbsoil)
tbbase <- util_nps_tbbase(tbsubshed, tbjuris, tblu2023, tbsoil, gdal_path = "C:/OSGeo4W/bin", chunk_size = 1000)

The tbbase data object is also included with the package for convenience.

head(tbbase)
#> # A tibble: 6 × 9
#>   bay_seg basin    drnfeat entity     FLUCCSCODE CLUCSID IMPROVED hydgrp area_ha
#>     <dbl> <chr>    <chr>   <chr>           <dbl>   <dbl>    <int> <chr>    <dbl>
#> 1       1 02304500 LAKE    HILLSBORO…       1100       1        1 A      0.00253
#> 2       1 02304500 LAKE    HILLSBORO…       1100       1        1 A/D    0.00653
#> 3       1 02304500 LAKE    HILLSBORO…       1200       2        1 A      0.0321 
#> 4       1 02304500 LAKE    HILLSBORO…       1200       2        1 A/D    0.0126 
#> 5       1 02304500 LAKE    HILLSBORO…       1400       4        1 A      0.00969
#> 6       1 02304500 LAKE    HILLSBORO…       1400       4        1 A/D    0.00779

Next, rainfall data must be obtained for the watershed. These data can be obtained using the util_getrain() function. The function retrieves daily rainfall data from NWIS NCDC stations in the watershed and returns a data frame with daily rainfall totals for each station. These data are always provided to the anlz_nps_ungaged() using the rain argument and not downloaded automatically to reduce execution time. The rain data object included with the package is provided for convenience.

Flow data are also required to estimate ungaged loads. This is the same dataset used to estimate gaged loads. These data can be obtained using the util_nps_getflow() function. The function retrieves flow data from USGS stations in the watershed and also uses external flow data for Lake Manatee, Tampa Bypass, and Bell Shoals (see the help file for util_nps_getextflow() for how to retrieve these data). A preprocessed USGS flow data frame can be provided to the usgsflow argument. The usgsflow data object is provided with the package to avoid re-downloading the data. Similarly, all flow data can be provided to the allflo argument. The allflo data object included with the package has both external and USGS flow data.

Once the required data are prepared, ungaged loads can be estimated using anlz_nps_ungaged(). The following describes the general methods for how the ungaged loads are estimated in four steps.

1. Data Preparation

Using the inputs described above, the first step processes land use data for logistic regression modeling and calculates inverse distance-weighted rainfall data for each sub-basin. This ensures that each basin receives rainfall estimates that account for spatial variability across the watershed based on the proximity and influence of nearby rain gauges.

2. Flow Estimation

A logistic regression model predicts monthly streamflow in ungaged basins using several key variables:

Rainfall variables: Current month rainfall plus 2-month lagged rainfall to capture antecedent moisture conditions
Land use percentages: Proportions of urban, agriculture, wetlands, and forest cover within each basin
Seasonal patterns: Separate treatment for wet season (July-October) and dry season (November-June)
Urban development intensity: Basins are classified into Group A (<19% urban) or Group B (≥19% urban)
Hydrologic soil characteristics: Soil group properties that influence infiltration and runoff

3. Runoff Coefficient Application

Land use and soil-specific runoff coefficients are applied to distribute the predicted basin flows across different landscape types within each basin.

4. Load Calculation

Pollutant loads are estimated using Event Mean Concentrations (EMCs) for different land use categories, calculating:

Total Nitrogen (TN) loads
Total Phosphorus (TP) loads
Total Suspended Solids (TSS) loads
Biochemical Oxygen Demand (BOD) loads
Stormwater-specific loads (with different EMCs for certain categories)

The fundamental equation for pollutant load estimation is:

Load = Flow × EMC × Unit Conversions

Where EMCs (Event Mean Concentrations) represent the average pollutant concentrations in stormwater runoff for different land use types. EMCs vary by land use category (CLUCSID) based on empirical studies of stormwater quality. Special handling is applied for water bodies and certain wetland types (CLUCSIDs 18, 20), which are assigned zero stormwater loads since these areas do not generate surface runoff in the same manner as terrestrial land uses.

All together, the above can be implemented as follows. Flow inputs (as cubic feet per second) to anlz_nps_ungaged() can be provided numerous ways:

Flow data provided using local files for Lake Manatee, Bell Shoals, and Tampa Bypass, USGS data retrieved automatically from the API
Flow data provided using local files for Lake Manatee, Bell Shoals, and Tampa Bypass, USGS data provided locally using the usgsflow argument (see the usgsflow dataset for the format, created using util_nps_getusgsflow())
All flow data provided locally using the allflo argument (see the allflo dataset for the format, created using util_nps_getflow()).

In all cases, input data can retrieved from APIs with the exception of flow data for Lake Manatee, Bell Shoals, and Tampa Bypass (see the help file for util_nps_getextflow() for how to get these data). The following example uses combined flow data included with the package for convenience. See above for how to create the tbbase and rain data objects if updated data are needed.

# required inputs
data(tbbase)
data(rain)
data(allflo)

nps_ungaged <- anlz_nps_ungaged(yrrng = c('2021-01-01', '2023-12-31'), tbbase = tbbase, rain = rain,
                                allflo = allflo)
#> Prepping rain data...
#> Estimating ungaged NPS loads...

head(nps_ungaged)
#> # A tibble: 6 × 12
#>   bay_seg basin     yr    mo clucsid h2oload tnload tpload tssload bodload  area
#>     <dbl> <chr>  <dbl> <dbl>   <dbl>   <dbl>  <dbl>  <dbl>   <dbl>   <dbl> <dbl>
#> 1       1 02306…  2021     1       1   7661.   14.6   2.40    137.    33.7  210.
#> 2       1 02306…  2021     1       2  59344.  133.   20.2    2156.   439.   971.
#> 3       1 02306…  2021     1       3 108271.  225.   39.9    6912.  1191.  1267.
#> 4       1 02306…  2021     1       4  81852.  159.   22.9    6766.  1408.   481.
#> 5       1 02306…  2021     1       5  58995.   96.6  15.8    5542.   566.   371.
#> 6       1 02306…  2021     1       7  27446.   32.4   4.12    549.   225.   281.
#> # ℹ 1 more variable: bas_area <dbl>

A data frame is returned with columns for bay segment, basin, year, month, clucsid (land use code), hydrologic load in m3/month, TN load in kg/month, TP load in kg/month, TSS load in kg/month, BOD load in kg/month, area (hectares, basin/clucsid combination), and basin area (hectares).

Combined gaged and ungaged loads

Refer to the prior sections for details on how the separate loads for gaged and ungaged portions of the watershed are calculated. The anlz_nps() function described below can be used to combine all steps. The function estimates non-point source (NPS) loads for Tampa Bay by combining gaged and ungaged NPS loads. Gaged loads are estimated using flow and water quality data. Ungaged loads are estimated using rainfall, flow, event mean concentration, land use, and soils data. The function also incorporates atmospheric concentration data from the Verna Wellfield site.

Pre-processed inputs for tbbase, rain, allwq, and allflo are used. See the help file for util_prepverna() for how to obtain the file for atmospheric concentration data. See above for how to recreate these files if updated data are needed.

data(tbbase)
data(rain)
data(allwq)
data(allflo)
vernafl <- system.file('extdata/verna-raw.csv', package = 'tbeploads')

nps <- anlz_nps(yrrng = c('2021-01-01', '2023-12-31'), tbbase = tbbase, rain = rain,
                vernafl = vernafl, allwq = allwq, allflo = allflo)
#> Estimating ungaged NPS loads...
#> Estimating gaged NPS loads...
#> Combining atmospheric data with ungaged NPS loads...
#> Combining ungaged and gaged NPS loads, estimating final...

head(nps)
#> # A tibble: 6 × 10
#>    Year Month source segment     basin tn_load tp_load tss_load bod_load hy_load
#>   <dbl> <dbl> <chr>  <chr>       <chr>   <dbl>   <dbl>    <dbl>    <dbl>   <dbl>
#> 1  2021     1 NPS    Boca Ciega… 207-5    2.42   0.404     80.8    14.5    1.22 
#> 2  2021     2 NPS    Boca Ciega… 207-5    1.65   0.276     55.3     9.94   0.834
#> 3  2021     3 NPS    Boca Ciega… 207-5    1.37   0.228     45.6     8.21   0.689
#> 4  2021     4 NPS    Boca Ciega… 207-5    1.58   0.263     52.6     9.46   0.794
#> 5  2021     5 NPS    Boca Ciega… 207-5    1.21   0.200     40.1     7.20   0.604
#> 6  2021     6 NPS    Boca Ciega… 207-5    2.64   0.441     88.4    15.9    1.33

Unlike the individual gaged and ungaged functions, anlz_nps() returns loading results in tons per month or year depending on the summary arguments. Hydrologic load is returned as million cubic meters per month or year.

The following functions are used internally and are provided here for reference on the components used in the calculations. Not all are used depending on the inputs provided.

anlz_nps_ungaged(): Estimates ungaged NPS loads.
anlz_nps_gaged(): Estimates gaged NPS loads.
util_nps_fillmiswq(): Fills missing water quality data with linear interpolation.
util_nps_getflow(): Gets flow estimates for NPS gaged and ungaged calculations.
util_nps_getusgsflow(): Gets USGS flow data for NPS calculations, used in util_nps_getflow().
util_nps_getextflow(): Gets external flow data (Lake Manatee, Tampa Bypass, and Bell Shoals), used in util_nps_getflow().
util_nps_getwq(): Gets water quality data for NPS gaged calculations.
util_nps_preprain(): Prepares and formats rainfall data.
util_nps_preplog(): Prepares land use data for logistic regression modeling.
util_nps_segment(): Assigns basins to bay segments.
util_prepverna(): Prepares and fills missing data with five-year means for the Verna Wellfield site data.

Results can be summarized by basin, segment, baywide, monthly, or annually using the summ and summtime arguments. By default, loads are returned monthly for each basin. Note that Boca Ciega Bay and Boca Ciega Bay South results are returned separately. Only Boca Ciega Bay South is used when estimating total bay loads.

Loads by land use type (using CLUCSID) can also be returned if aslu = TRUE. These results only apply to ungaged loading estimates.

npslu <- anlz_nps(yrrng = c('2021-01-01', '2023-12-31'), tbbase = tbbase, rain = rain,
                allwq = allwq, allflo = allflo, vernafl = vernafl, aslu = TRUE)
#> Estimating ungaged NPS loads...
#> Combining atmospheric data with ungaged NPS loads...
#> Summarizing ungaged NPS loads by land use...

head(npslu)
#> # A tibble: 6 × 11
#>    Year Month source segment       basin lu    tn_load tp_load tss_load bod_load
#>   <dbl> <dbl> <chr>  <chr>         <chr> <chr>   <dbl>   <dbl>    <dbl>    <dbl>
#> 1  2021     1 NPS    Boca Ciega B… 207-5 Barr… 7.11e-5 5.74e-7 0.000631  8.32e-5
#> 2  2021     2 NPS    Boca Ciega B… 207-5 Barr… 4.86e-5 3.92e-7 0.000431  5.69e-5
#> 3  2021     3 NPS    Boca Ciega B… 207-5 Barr… 4.02e-5 3.24e-7 0.000356  4.70e-5
#> 4  2021     4 NPS    Boca Ciega B… 207-5 Barr… 4.63e-5 3.73e-7 0.000411  5.41e-5
#> 5  2021     5 NPS    Boca Ciega B… 207-5 Barr… 3.52e-5 2.84e-7 0.000313  4.12e-5
#> 6  2021     6 NPS    Boca Ciega B… 207-5 Barr… 7.78e-5 6.27e-7 0.000690  9.10e-5
#> # ℹ 1 more variable: hy_load <dbl>

Note that results from anlz_nps() are returned in units of tons/month for TN, TP, TSS, and BOD and million cubic meters/month for hydrologic load if summtime = 'month' or tons/year for TN, TP, TSS, and BOD and million cubic meters/year for hydrologic load if summtime = 'year'. Results from the gaged and ungaged functions are returned in kg for TN, TP, TSS, and BOD and m3 for hydrologic load.

Removing point source loads from gaged NPS estimates

Because stream gauges measure total flow at a location, gaged NPS load estimates from anlz_nps() include any point source discharges upstream of the gauge. The anlz_nps_psremove() function subtracts the IPS and DPS loads occurring in gaged basins from the NPS totals so that point source contributions are not double-counted. Only point source loads in basins classified as “Gaged” in the dbasing lookup table are subtracted. Loads in ungaged basins are unaffected.

The function requires pre-computed basin-level monthly loads from anlz_nps(), anlz_ips(), and anlz_dps(), all called with summ = 'basin' and summtime = 'month'. Nested basin identifiers (e.g., 02301000, 02301300) are reassigned to their parent basins (02301500, 02304500) before summing, consistent with the handling already applied inside anlz_nps().

An optional AD/AP TN adjustment (ad_ap = TRUE by default) applies fixed monthly reductions from the 2007 RA allocation analysis to the segment-level NPS totals: Old Tampa Bay (−2.41 tons/month), Hillsborough Bay (−4.31 tons/month), Middle Tampa Bay (−2.29 tons/month), Lower Tampa Bay (−0.36 tons/month), and Manatee River (−2.74 tons/month, representing the combined reduction for segments 55, 6, and 7). Only TN is adjusted. TP, TSS, BOD, and hydrologic loads are unchanged.

# pre-compute basin-level monthly loads for each source
ipsfls <- list.files(system.file('extdata/', package = 'tbeploads'),
  pattern = 'ps_ind_', full.names = TRUE)
dpsfls <- list.files(system.file('extdata/', package = 'tbeploads'),
  pattern = 'ps_dom', full.names = TRUE)

nps_basin <- anlz_nps(yrrng = c('2021-01-01', '2023-12-31'), tbbase = tbbase,
  rain = rain, vernafl = vernafl, allwq = allwq, allflo = allflo,
  summ = 'basin', summtime = 'month')
#> Estimating ungaged NPS loads...
#> Estimating gaged NPS loads...
#> Combining atmospheric data with ungaged NPS loads...
#> Combining ungaged and gaged NPS loads, estimating final...
ips_basin <- anlz_ips(ipsfls, summ = 'basin', summtime = 'month')
dps_basin <- anlz_dps(dpsfls, summ = 'basin', summtime = 'month')

# subtract gaged PS loads and apply AD/AP TN adjustment
nps_psremoved <- anlz_nps_psremove(nps_basin, ips_basin, dps_basin)

head(nps_psremoved)
#> # A tibble: 6 × 9
#>    Year Month source segment        tn_load tp_load tss_load bod_load hy_load
#>   <dbl> <dbl> <chr>  <chr>            <dbl>   <dbl>    <dbl>    <dbl>   <dbl>
#> 1  2021     1 NPS    Boca Ciega Bay    2.42   0.404     80.8    14.5    1.22 
#> 2  2021     2 NPS    Boca Ciega Bay    1.65   0.276     55.3     9.94   0.834
#> 3  2021     3 NPS    Boca Ciega Bay    1.37   0.228     45.6     8.21   0.689
#> 4  2021     4 NPS    Boca Ciega Bay    1.58   0.263     52.6     9.46   0.794
#> 5  2021     5 NPS    Boca Ciega Bay    1.21   0.200     40.1     7.20   0.604
#> 6  2021     6 NPS    Boca Ciega Bay    2.64   0.441     88.4    15.9    1.33

Results are returned at the segment/month level with the same column structure as anlz_ips() and anlz_dps(). Annual totals can be obtained by setting summtime = 'year'. Load units are tons for TN, TP, TSS, and BOD and million cubic meters for hydrologic load.