ENV 761 (Fay) - Conservation GIS</h2> </div> # Background Riparian buffers have beneficial qualities in maintaining stream water quality. Nutrients and sediments originating from non-point sources within the catchments are intercepted by these buffers and flashy flows are dampened by the resistance of roots and other woody debris. What exactly constitutes a "riparian buffer", however, is a tricky question. Oftentimes, land managers just look at how much forested land cover is found within a certain distance of a stream -- something easily calculated using GIS. However, GIS offers some far more sophisticated and more precise ways of quantifying the buffering potential of different landscape configurations. Baker et al. (2006) examines a sequence of spatial analytical approaches to measuring riparian buffering potential. In this tutorial we examine these approaches with two goals in mind: First is to master these different approaches in the event that you need to evaluation riparian buffers at some point in your career. Second is to appreciate their crafty use of geoprocessing tools that Baker et al. use to solve a specific issue. Using tools productively in ways they weren't exactly intended is a great GIS skill to have... We don't stop with Baker et al.'s approaches to riparian buffer analysis. Our very own Brian McGlynn has developed a few other ways at quantifying "hydrologic connectivity", which isn't exactly the same as riparian buffering, but does indeed relate. We examine a technique inspired by his research to delineate floodplains defined by the area adjacent to streams that are within a set elevation above the stream into which it drains. ## Materials and lab approach These exercise require the data in the **RiparianTutorial.zip** file found on Canvas. This zip file contains a preformatted workspace that includes (in the data folder) the 30m DEM (NED) and the 2001 NLCD for roughly the same area that Baker et al. examined (Maryland, USA). Also provided is a toolbox with models that will help you kick start your analysis. The sections below outline and explain the procedures required to execute the various riparian buffer analyses, from simple to elegant. Follow these steps to better understand the spatial analyses required by each. The assumptions in this exercises include: 1) That any land cover classified as forest (deciduous, evergreen, or mixed), and only forested land has the capacity to buffer non-point source pollutants within a catchment. (Baker et al include wetlands in their analysis, but for simplicity, we'll just go with forest.) 2) That cropland land cover types contribute to nonpoint source pollution (NPS). The more cropland within a catchment, the more potential pollution can enter the streams. ## Data Prep Prior to executing the various buffer analyses, you'll have to derive flow direction, flow accumulation, and a streams layer (thresholded at 1000 cells of accumulation). You will also create watersheds for the 4 gauge sites, which [ ]{.underline} will have to be snapped to a distance of 60m. Note that the DEM has already been filled and contains no sinks. You will also create binary rasters of the buffering land cover (forests, NLCD 41, 42, & 43) and NPS contributing areas (cropland, NLCD 82). By "binary raster", I refer to rasters with two values: 1 (e.g. if it's a forest pixel) and 0 (if it is not). A geoprocessing model is provided to you to speed you along these analyses. ## 1. Catchment wide The crudest method for estimating the buffering potential within a catchment is simply to quantify the relative amount of land that contributes nutrients and sediments to the land that intercepts these pollutants. Although different cover types contribute different levels of pollutants, we will consider only land classified as croplands to be a significant non-point source. Therefore, our first metric is to calculate the percentage of row crops found within the area draining a particular point and compare that to the percent forest found within the same drainage area. You should recognize this analysis as a **zonal** operation on your rasters. Because you are calculating statistics for each watershed in your landscape, the watersheds will be the zones. And the calculation you want to compute can either be the proportion or total area of buffer and of NPS contributing cells. Use zonal mean to compute proportion and zonal sum to compute total area (in cells). Or you can use the Tabulate Area tool to compute areas as well. ## 2. Fixed distance A more spatially explicit predictor of NPS impacts looks at the distances of sources and sinks relative to the stream course. The logic here is that the filtering effects of forests and wetlands are more likely to benefit the stream if they are found between the stream and the non-point sources. A common estimator of this is simply to calculate the percent of forest or wetland occurring within a fixed distance -- we use 100 m here -- of stream courses. This is the same analysis as above, except we want to limit -- or mask - our tabulation of forest pixels to those within 100 m of the stream. The tabulation of the NPS cells will remain the same. ## 3. Unconstrained This approach is quite a bit more complex than the previous two. The fixed distance approach has two major drawbacks. First is that it ignores any buffering that may go beyond the 100m zone we are looking at; a catchment could have solid forest that extends 500 m beyond streams and it would be considered equal to one with just enough buffer to the 100 m extent. The second drawback of the fixed distance approach is that neglects *where* the gaps in buffer occur. A catchment that's 50% forested within the 100 m zone could have nice solid 50 m buffer on either side of the stream, which would be ideal given the proportion of forest. But it could also be solidly buffered to 100 m on one side and have zero buffer on the other, meaning NPS pollution would have a clear, unbuffered path to the stream in half the catchment. The unconstrained approach designed by Baker et al. overcomes these two drawbacks. It calculates, for each stream cell, the minimum amount of buffer between it and the nearest NPS contributing cell. We then compute the average distance of all the stream cells, but also compute the number of streams cells that have [no]{.underline} buffer between them and NPS cells. Applied to our above hypothetical catchments with 50% forest within 100 m, the nicely buffered one would have an average distance of 50 m and no unbuffered cells. The other one, however, is that you could have a great deal of buffering forest within 100 m of streams, but the places without buffer could possibly allow direct, unbuffered pathways between NPS contributing areas and the stream.