PreppingEnvironmentData - PNHP/Regional_SDM GitHub Wiki

Background

This serves as a guide to examine, compile, and pre-proprocess environmental data for use in our aquatic modeling approach. This page also provides information about the source, compilation, and pre-processing of NHD stream reaches (and associated catchments) that are used as modeling units for this project.

Data Downloads and Compilation

National Hydrography Dataset (NHD)

The NHDPlus Version 2 dataset that includes stream reaches and catchments was downloaded for the Ohio, Mid-Atlantic, and Great Lakes Regions. Access to the dataset can be found here and documentation providing more information about the dataset can be found here. Once downloaded, we used ArcGIS software to merge stream reaches and catchments into single polyline and polygon feature classes, respectively. We then selected all catchments and stream reaches within our area of interest to produce our final stream reach and catchment data to be used in further data development, processing, and modeling. NOTE - we used intersect to select catchments and stream reaches within area of interest instead of clip to avoid clipping catchments/stream reaches at an inaccurate edge due to resolution differences found in HUC watersheds.

StreamCat Data

StreamCat datasets were downloaded for the Ohio, Mid-Atlantic, and Great Lakes Regions. More information about the StreamCat data and for links to the data dictionary describing StreamCat variables can be found here. The StreamCat data for the different regions were merged. Although the StreamCat data are exclusively tabular, the values in the StreamCat dataset can be linked to the stream reaches and catchments in the NHDPlusV2 by their COMID. Through this unique identifier, values for StreamCat variables can be joined to the NHD stream reaches and/or catchments for a geospatial representation of the variables.

The following environmental variables found in the StreamCat dataset were used:

Data Processing and Preparation

Land Cover Data

We used the 2011 National Land Cover Dataset (NLCD) to derive catchment level riparian fragmentation metrics to be used in the aquatic models. The 16-class land cover classification scheme used by the 2011 NLCD was reclassified to 4 broad land cover classes that included natural, agriculture, developed, and open water (Table 1). The broader 4-class land cover classification scheme allowed for more appropriate representation of land cover for wide-scale modeling.

Table 1. Reclassification scheme used for preparing the NLCD data.

Deriving Riparian Land Cover

Using vectorized flowlines to derive riparian land cover classifications would have inaccurately represented riparian land cover in areas where streams and/or rivers are large enough to appear on the NLCD layer. For example, in Figure 1, capturing the land use within the buffered flowline would have resulted in the majority of the riparian being classified as 'Open Water', instead of the true riparian that surrounds the open water. Therefore, we took the following steps to more accurately represent the riparian land cover classes (see Figure 2 for example of the riparian raster layer produced by these steps and notice how it excludes the open water):

  1. downloaded and compiled National Wetland Inventory datasets for states within the extent area
  2. selected 'Estuarine and Marine Deepwater', 'Freshwater Pond', 'Lake', and 'Riverine' types from the NWI layer
  3. from those selected in the NWI layer, selected those that intersected with NHD flowlines to exclude non-riverine water features
  4. converted selected NWI layer from feature to raster
  5. expanded the NWI raster layer by 3 cells to capture all potential 'Open Water' cells in the NLCD raster
  6. used Con tool with NLCD layer as the input conditional raster, 'value = 5' as expression, the NWI_expand raster as the input true raster, and NWI_openwater raster as the output.
  7. used Expand tool with NWI_openwater as input and 4 cells as the expand distance to create a 120 meter buffer that represented the riparian
  8. used Cell Statistics tool to combine flowline buffer and NWI buffer - use riparian_combine120m as output
  9. used Con tool to delete open water cells from riparian by using the NLCD as the conditional raster, 'Value <> 5' as expression, and riparian_combine120m as true raster
  10. used Con tool to extract NLCD landcover within riparian

Figure 1. Example of how flowline did not accurately represent the width of the river and how a generic riparian buffer on the flowline would have caused inaccuracies in the classification of riparian land covers.

Figure 2. Example of riparian land cover raster layer that was created to exclude open water and include 120m of riparian land cover cover classes surrounding streams/rivers.

FRAGSTATS Metrics

The 'FragstatsPrep.py' script was used to prepare the riparian land cover raster layer for use with FRAGSTATS. This script clips the riparian land cover raster layer by catchment and produces a new .tif file for each catchment. The script then produces a batch file that can be used as input into FRAGSTATS. FRAGSTATS was then used to calculate the following fragmentation metrics for each catchment:

  1. Percentage of Landscape
  2. Number of Patches
  3. Patch Density
  4. Mean Patch Area
  5. Radius of Gyration
  6. Shape Index
  7. Contiguity Index
  8. Proximity Index
  9. Patch Richness