SESYNC HEALTHY WATERSHED PROJECT

Goals:

• Improve git and R skills

• Improve group collaboration skills

• Develop reproducible workflows to add/update and evaluate variables to the Chesapeake Healthy Watershed Assessment (CHWA)

Variable: Percent forest cover

Data:

• Maryland Biological Stream Survey data attributed with numeric and categorical stream condition (point shapefile)
• CHWA database (National Hydrologic Dataset (100K) catchments attributed with CHWA variables (CSV or polygon shapefile).
• 2013 Tree Canopy for Maryland (1-meter raster)

Initial workflow:

• Subset CHWA for Maryland
• Summarize (zonal) tree canopy area by NHD catchment and add/update attribute
• Optional-- Accumulate tree canopy area downstream (optional call to Sarah’s Python code)
• Add accumulate tree canopy area as a new attribute to CHWA database
• Relate/join MBSS point ID to NHD+ COMID
• Regress subset of CHWA metrics, including new one, against MBSS condition
• Metrics vs MBSS numeric
• Metrics vs MBSS categorical
• Visualize regression results
• Evaluate regression results and report pass/fail based on user defined significance threshold
• If pass- commit changes to CHWA (aka NHD database), if fail- don’t commit changes to CHWA
• Add QAQC steps

Collaborators

• Renee Thompson, USGS
• Peter Claggett, USGS
• Sarah McDonald, USGS
• Labeeb Ahmed, Attain LLC

---

SESYNC Healthy Watersheds Final Presentation

Background on Chesapeake Healthy Watersheds Assessment

The Chesapeake Bay Program, through its Maintain Healthy Watersheds Goal Implementation Team, has a goal of maintaining the long-term health of watersheds identified as healthy by its partner jurisdictions. Quantitative indicators are important to assess current watershed condition, track future condition, and assess the vulnerability of these state-identified watersheds to future degradation.

Chesapeake Healthy Watershed Assessment report

Reproducible pipeline

Can you aquire data programatically?Yes and No

• some data is available through Maryland iMap online, Chesapeake Open Data
• some data was provided manually through a data agreemnet

Discuss the final steps of your reproducible pipeline. Should you create a data package? Try to get that regression model working or finish that data visualization. Prepare to present on your mini-project!

Workflow for adding/updating and evaluating variable(s) into CHWA Assessment:

1. Spatially join the healthy watersheds metrics to the MBSS point data

    chwa_mbss <- st_join(mbss, chwa)
   

2. Buffer the MBSS points by 100 meters

    mbss_100 <- chwa_mbss %>%  st_buffer(100)
   

   

3. Read in land cover raster containing forest and impervious classes

   LC <- raster("LC_MD_5m.tif")
   

4. Calculate Percent Forest and Percent Impervious cover within the 100m buffers

    aggLC <- extract(LC, mbss_100)
    
    pctForest <- rep(0, length(aggLC))
    
    pctImp <- rep(0, length(aggLC))
    
    for(row in 1:length(aggLC)){tb <- as.data.frame(table(aggLC[row]))
      forest <- tb[tb == 1, 'Freq']
      imp <- tb[tb == 2, 'Freq'] #(tb[tb == 5, 'Freq'] + tb[tb == 6, 'Freq'])
      if (length(forest) != 0){
        pctForest[row] <- forest / length(aggLC[[row]]) #percent forested}
      if (length(imp) != 0){
        pctImp[row] <- imp / length(aggLC[[row]])  # percent impervious } }
   

   

5. Add newly calculated data to the joined data

   chwa_mbss <- as.data.frame(chwa_mbss)
   chwa_mbss['BUFPctFor'] <- pctForest
   chwa_mbss['BUFPctImp'] <- pctImp
   

Leaflet maps Visualizing Visualizing percent forest buffer

6. Clean and fix column names

7. Filter and exclude catchments with missing data

8. Rank multiple data points within each catchment by year to use as weights

9. Group-by and calculate weighted means

10. Perform linear regression and step-wise regression

Benthic IBI:

                     Estimate Std. Error t value Pr(>|t|)    
      (Intercept)   2.415e+00  3.463e-01   6.975 4.28e-12 ***
      PctNatural   -1.605e+00  3.953e-01  -4.060 5.12e-05 ***
      Mngd_TF_HCZ  -8.109e-01  2.116e-01  -3.833 0.000131 ***
      PctForeWshed  2.436e+00  3.140e-01   7.758 1.43e-14 ***
      PctImpWshed  -4.177e+00  6.678e-01  -6.255 4.94e-10 ***
      HabConditi    1.188e-01  3.484e-02   3.409 0.000667 ***
      SPARROWTN    -9.723e+01  2.135e+01  -4.555 5.59e-06 ***
      SPARROWTP     6.522e+01  7.648e+00   8.528  < 2e-16 ***
      CBP_Ag_N      6.705e+01  1.406e+01   4.770 1.99e-06 ***
      BUFPctFor     5.819e-01  8.689e-02   6.697 2.83e-11 ***
    
      Multiple R-squared:  0.2943,    Adjusted R-squared:  0.2846 

Fish IBI:

                     Estimate Std. Error t value Pr(>|t|)    
      (Intercept)     3.92367    0.25772  15.225  < 2e-16 ***
      PctNatural     -3.13719    0.48225  -6.505 1.01e-10 ***
      PctForeWshed    1.46428    0.42876   3.415 0.000652 ***
      PctImpWshed    -3.15918    0.68925  -4.584 4.90e-06 ***
      RoadStream     -0.15386    0.04157  -3.701 0.000222 ***
      CBP_Ag_P      202.00419   29.70632   6.800 1.43e-11 ***
      CBP_WW_S     -140.51963   34.18429  -4.111 4.13e-05 ***
      BUFPctFor       0.50275    0.11679   4.305 1.76e-05 ***

      Multiple R-squared:  0.1129,    Adjusted R-squared:  0.1048

Lessons learned:

• GITHUB, R, and SLACK work great for collaborative coding and managing coding projects
• Project demonstrated the importance of understanding and exploring data initially- something R is very good at
• Important to request code for all outsourced geospatial projects

Future:

• Need to find a way to better manage code from disparate projects
• Need a key for navigating statistical options in R
• Will apply collaborative approach taught in this workshop to future geospatial analysis projects