Prioritizing riparian corridors for ecosystem restoration in urbanizing watersheds

Riparian corridors

As one of the most diverse of habitat types, riparian corridors influence water quality, flood prevention, wildlife habitat, economics, and various other ecological, physical, biological, and chemical processes (Wagner, 2004). Riparian corridors reduce erosion (Castelle & Johnson, 2000), filter sediments and pollutants out of overland runoff (Staddon, Locke & Zablotowicz, 2001; Dabney, Moore & Locke, 2006; Bongard et al., 2010), block solar radiation to moderate water temperature (Sweeney, 1992), provide habitat (Jones III et al., 1999), and store water thereby moderate flooding (Cooper, Hiscock & Lovett, 2019). However, on a global basis more than 32,000 large dams (defined by the International Commission on Large Dams as a dam with a height of 15 m or greater, or a dam between 5 m and 15 m impounding more than 3 million cubic meters (ICOLD, 2011)), and upwards of 1,000,000 total dams have fragmented systems of streams and rivers (Jackson et al., 2001), leading Dudgeon et al. (2006) to state that freshwater ecosystems “…may well be the most endangered ecosystems in the world”. The US Geological Survey’s John Wesley Powell Center for Analysis and Synthesis further suggests that “fresh water is arguably the most valuable resource on the planet, but human activities threaten freshwater ecosystems” (USGS, 2019). Effective assessment and management techniques are necessary to protect the diversity of the ecosystem services found within these ecosystems, and to mitigate current and future conditions of environmental stressors amplified by rapid urban development.

Scientists have long studied riparian corridors as ecosystems of forested and/or vegetative transition zones that link aquatic and terrestrial environments (see, for example Karr & Schlosser (1978). Studies show that healthy riparian stream corridors perform a multitude of valuable services for their adjacent waterways, including: their overall influence on water quality (e.g., Connolly et al., 2015); biological diversity (e.g., Mlambo et al., 2015); ecosystem maintenance (e.g.,  Montgomery, 2001); and protection of intermittent streams and the residual pools that they provide as refuges for multiple species during dry periods (e.g., Wigington Jr et al., 2006). Nutrient cycling, contaminant filtration, water purification, bank stabilization, stream temperature maintenance, flow stabilization, flood attenuation, and habitat preservation are some of the numerous functions carried out by riparian zones (National Research Council, 2002). In a recent review of riparian restoration literature (Feld et al., 2018), it was stated that riparian restoration provides “a no-regrets management option to improve and sustain lotic ecosystem functioning and biodiversity”.

Riparian corridors exhibit a wide variety of geomorphic characteristics across a continuum of scales. Studies that need to define the lateral extent of those corridors for research purposes have used a range of approaches, from defining each riparian corridor on an individualized case-by-case basis, to defining all corridors simultaneously in a study area based upon more generalized definitional basis. An example that is nearer the individualized end of the spectrum, a recent study that developed a model to measure departures of current riparian conditions from historic conditions (defined as pre-European settlement in the north-western United States) used a stream segment’s valley bottom width as the lateral riparian extent because that represents the maximum possible extent of riparian vegetation (Macfarlane et al., 2017). Using Thiessen polygons with centroids located at the midpoint of each stream segment, they defined the individual lateral extent of each steam segment’s riparian corridor across an irregular planform of geometries and valley bottoms. Nearer the other end of the spectrum, Gilbert, Macfarlane & Wheaton (2016) developed a tool that set the maximum valley bottom (defined as the stream or river channel and the associated low-lying floodplain) as the maximum possible extent of a riparian area. That tool set the lateral extent of riparian areas for a given size of a drainage area (classified as either “confined”, “transition” or “large” based on km²) based on a buffer around the stream channel set to the maximum valley width in the drainage area.

Watershed managers concerned with water quality find that suggesting restrictions on development in high quality riparian corridors, or restoring degraded riparian corridors, is particularly challenging in Texas. Water law in Texas has evolved from conflicts between private landowners and the needs of Texas citizens. Privately owned lands dominate the landscape of Texas; less than 2% are federal lands (Vincent, Hanson & Argueta, 2017). Additionally, two legal doctrines of surface water law are recognized in Texas today: the riparian doctrine, and the prior appropriation doctrine (Sansom, 2008). The fundamental concept of prior appropriation is based on pre-statehood Spanish law that evolved to control water use in irrigation intensive settlements in Texas and the southwest. Those Spanish law concepts have today evolved such that Texas owns most surface water, holding it in trust for the public, and allocates use of that resource through a permitting process. Permits are issued on a “first in time, first in right” for a given allocation of water. The riparian doctrine is based on an older English common law approach to water use. The basic concept is that private water rights are tied to the ownership of the land bordering a natural river or stream. Thus, surface water rights in Texas are tied to both state permitted appropriations and by land ownership (Wurbs, 2004), effectively restricting watershed managers’ options, and for all practical purposes limits the lateral extent of a riparian corridor that a watershed manager can consider for water quality purposes.

Here, we offer a strategy for prioritizing riparian corridors for potential ecosystem restoration focused on water quality. Fortunately, efforts focused on a specific ecosystem service can also enhance other ecosystem service. For example, Fremier et al. (2015) focus on riparian restoration as a means for improving connectivity between protected habitats as a means to increase the ecological resilience of those protected habitats. Their work was motivated in part much the same as ours; increasing a system’s ability to respond to natural and human induced perturbations. Although their focus was for enhancing biodiversity while ours was for improving water quality, the net result of either focus would likely yield benefits to the other. Other beneficial synergistic restoration outcomes could include aesthetics, in-stream habitat, fish passage and bank stabilization (see, e.g., Bernhardt et al., 2005).

Effects of urbanization

Urbanization and its subsequent activities, without proper planning, often leads to the degradation of streams and their riparian corridors. This degradation may affect the natural cycles of biological and physical activities normally carried out within riparian ecosystems (Tanaka et al., 2016), and also cause social and economic problems at both local and regional levels. For example, Withers & Jarvie (2008) state that the problems can influence human health (e.g., algal toxins), species abundance and diversity, amenity value and costs of water treatment for drinking. Streams and rivers on the periphery of urbanizing areas are particularly vulnerable due to population density, sensitivity to land use change and ubiquitous exploitation (Walsh et al., 2005). Furthermore, many of those aquatic systems at the urban periphery have already been exposed to agricultural practices, such as grazing and the direct access of cattle to streams. This has resulted in increased erosion of stream banks due to trampling, as well as direct deposition and indirect flow of animal waste into waterways, a principal component of nonpoint source pollution (Hamilton & Miller, 2002). Ultimately, the dynamic equilibrium of stream ecosystems can be altered by the cumulative effects of channelization, clear-cutting, illegal dumping, and increased chemical usage, all consequences of urbanization of surrounding riparian corridors.

Many places throughout the world are experiencing the same pattern of land use change that can be seen in north central Texas, the application site for this model. Former rural areas are becoming a part of an ever increasing urban landscape. As residential developments, commercial properties, and industrial facilities increase, they cover the natural landscape with roads, buildings, parking lots, and other impervious surfaces (Wang et al., 2019). Stream health can be directly linked to urbanization, the effects of which simultaneously decrease bank stability and increase pollutant presence and transfer. Healthy riparian buffer zones have been shown to filter out up to 97% of soil sediment prior to stream entrance (Lee, Isenhart & Schultz, 2003). However, removing trees in riparian zones is often one of the first activities associated with urban and suburban development, leading to increased soil erosion. Increased erodibility results in a decrease in the depth of fertile topsoil and an increase of sediment within streams. These sediments often contain metals such as lead, chromium or zinc, pesticides such as DDT, other organics such as polychlorinated biphenyls (PCBs) or polycyclic aromatic hydrocarbons (PAHs), or various other synthetic chemicals that may be toxic to aquatic and terrestrial species, and may also be linked to human health via the food chain (see, for example: Pavlović et al., 2016; Walters et al., 2018; Bing et al., 2019).

The 2000 National Water Quality Inventory (US Environmental Protection Agency, 2000) indicated that 39% of the river and stream miles in the United States were listed as impaired or polluted, but the updated 2017 report indicated that the number increased to 46% (US Environmental Protection Agency, 2017). River and stream conditions were assessed with measures of biological quality, chemical stress, physical habitat stress, or human health indicators. Aquatic conditions have led to extinction rates of freshwater fauna that are five times that for terrestrial biota (Ricciardi & Rasmussen, 1999; Sand-Jensen, 2001). Fortunately, aquatic ecosystem restoration can lead to improved biological, chemical and physical conditions, resulting in both improved wildlife habitat and water quality.

Objectives of this study

In an earlier study focused on prioritizing and protecting the highest quality riparian corridors, a spatially explicit modeling and mapping technique was developed; the Water Quality Corridor Management (WQCM) model (Atkinson et al., 2007; Atkinson, Hunter & English, 2010), originally referred to as the “wick ‘em” model because of the acronym. We now call that model the WQCM-P model, because it is focused on “protection” strategies. The model involves a geospatial database that utilizes GIS and remote sensing techniques to assess and prioritize stream reaches according to their overall health and sustainability. That model was developed and then assessed for its ability to document stream corridor quality, and ultimately, establish a ranking system for developing management strategies for protecting the highest quality stream corridors for drinking water quality purposes.

The model’s application area for this research consists of an important watershed in north central Texas, USA. The overall watershed is 240,000 hectares in size (593,000 acres) and drains to a relatively large drinking water reservoir (12,000 hectares; 29,600 acres), Lewisville Lake, which serves the Dallas /Ft. Worth area (current population approximately 6 million). The watershed was divided into 90 sub-watersheds based on US Geological Survey Hydrologic Unit Codes, ranging in size from 230 to 12,000 hectares (570 to 29,000 acres). Atkinson, Hunter & English (2010) provides a map that shows the results of the original WQCM-P modeling applied to that watershed. That map illustrates to watershed managers which sub-watersheds contain the highest priority stream corridors that should receive the initial focus for an overall watershed riparian protection program.

Because high quality riparian corridor protection is only one strategy available for water quality purposes, it became apparent after the release of the WQCM-P model that a tool was also needed to rank riparian corridors in terms of potential aquatic ecosystem restoration. A re-working of the original model was hypothesized to potentially provide that new focus. The new work described in this paper turns the riparian protection question around, and asks the model to prioritize riparian corridors for potential restoration activities in order to improve water quality in the reservoir. The objective of the research reported here was to modify the original model, using the same spatial databases, to one focused on potential “restoration” opportunities, or the WQCM-R model. Using the same databases to reframe the question asked in this research allowed an analysis of the effects of the reframing of the question, without a confounding effect of also changing databases. Therefore, the new model described in this paper used the same sub-watershed delineations, stream network, land use, soils, and topography data as was used the original model.

A subsequent benefit of this new model is that not only will it indicate which sub-watersheds contain riparian corridors that have the greatest restoration needs, it also allows targeting of sub-watersheds that would receive the largest increases in riparian corridor quality based on specified restoration strategies (e.g., revegetation with native plants, soil stabilization, etc.). As Nilsson et al. (2015) state, natural river ecosystems are both self-sustaining and dynamic, one of their stated goals for successful river restoration. Our goal is that WQCM-R will assist watershed managers in choosing not only the riparian areas most in need of restoration, but also help them identify the most appropriate restoration strategy. Using both the WQCM-R and the WQCM-P models, watershed managers should have access to a broad range of riparian corridor options for assisting in water quality considerations in rapidly urbanizing watersheds.

GIS materials

The following spatial data layers were utilized in the generation of the WQCM-R model:

Sub-watershed and its Corridor Component: US Geological Survey (USGS) Hydrologic Unit Codes were used to define 90 sub-watersheds in the overall Lewisville Lake drainage. These sub-watersheds ranged in size from 230 to 12,000 hectares (570 to 29,000 acres). ESRI ArcInfo’s buffer tool was used to generate a 20 meter (66-foot) wide buffer zone around a stream shapefile (40 meter total width) obtained from the USGS National Hydrography Dataset (NHD¹), which is provided at a scale of 1:24,000. The stream buffer was then exported as a separate shapefile to be used as the extent from which the land use, soil erodibility, and surface slope were clipped. Because sub-watershed varied substantially in area, all subsequent data were transformed with a scaling factor: the ratio of total riparian area within a sub-watershed to total sub-watershed area.

Land Use Parameter: Using the same land use classifications from our previous modeling effort (Atkinson et al., 2007; Atkinson, Hunter & English, 2010), surface area of eight classes of land use were extracted for each riparian corridor in the study area: barren, cropland/pasture, forested, residential, shrub/brush rangeland, urban, water, and unclassified. The land use classes were derived from 30-m resolution LANDSAT 8 ETM satellite imagery using Definiens eCognition object-based image classification software. The native scale of 30 m raster cells were utilized for all raster data.

Soil Erodibility Parameter: Using the same erodibility data from our previous model, the Soil Survey Geographic (SSURGO) data from the National Resource Conservation Service was used to create a soil erosion potential shapefile. There were eight erodibility ‘kffact’ categories in the study area, where higher numbers indicate higher potential erosion: Kw = 0, Kw = 0.17, Kw = 0.20, Kw = 0.24, Kw = 0.28, Kw = 0.32, Kw = 0.37, and Kw = 0.43. Soil erodibility classes were extracted for each riparian corridor in the study area.

Surface Slope Parameter: The same Digital Elevation Model (DEM) from our previous model was used for developing surface slope in the study area. Thirty meter raster data were obtained from the USGS National Elevation Dataset (NED) to create a percent slope raster file. Those data were categorized into five potential surface slope categories: <1%, 1% to <2%, 2% to <3%, 3% to <4%, and 4% to 5% (no riparian corridors had areas with slopes greater than 5%).

Methods used in original water quality corridor model for protection –WQCM-P

The original WQCM-P model utilized the same four geospatial variables listed above, but also included a “floodplain parameter” which accounted for the percent of a riparian zone that fell within a Federal Emergency Management Agency (FEMA) 100-year floodplain. This was included in the WQCM-P model because it helped account for some “protective” measures already in place (e.g., restrictions on certain types of developments within the floodplain).

Each of the spatial variables consists of an importance weight and a scaling function (details provided in Atkinson et al. (2007) and (Atkinson, Hunter & English, 2010)). Importance weights and scaling functions assigned to each variable range from 1 to 5, with 5 indicating a greater need for protection. For example, land use class was considered the most important variable in the model, receiving 5 importance points, and the scaling function for land use indicates that forested areas within the riparian buffer receive 5 points while residential areas within the riparian buffer receive 2 points. Each variable’s scaling function was based on the same concept: what conditions are more relevant to protect via preservation? Two of the variables are specific to preservation goals. First, the FEMA 100 year Floodplain variable was considered because areas designated to be inside these floodplains already receive some amount of preservation protection (e.g., certain types of activities require flood insurance which often discourages that activity). Second, the Corridor Ratio variable is the ratio of stream’s corridor area to its sub-watershed area. The Corridor Ratio and was considered because larger corridor ratios suggest less room for development in the sub-watershed and therefore more pressure to develop inside a stream’s corridor area. In the WQCM-P model, higher scores represent higher quality and therefore should be targeted for riparian corridor protection.

Changes in methods for the water quality corridor model for restoration—WQCM-R

The newly developed WQCM-R, established for restoration potential modeling, eliminated the floodplain variable and the corridor ratio variables used in WQCM-P because they were specifically focused on preservation strategies. Because floodplain development restrictions typically do not affect “restoration” of riparian zones, these parameters were eliminated from the WQCM-R model. The remaining four variables were retained for WQCM-R but importance and scaling functions were altered to reflect a focus on restoration potential. Final scores were calculated for each corridor in each sub-watershed’s riparian area based on characteristics of three variables: land use (L), erodibility (E) and slope (S). Values were calculated for the forth variable, the area of the 20 m wide corridor on each side of the stream segments in the study area, generating an overall WQCM-R score for the riparian corridors of each sub-watershed.

The underpinning model is presented in Eq. (1): (1)${\displaystyle \text{WQCM-R Score}={\mathrm{L}}_{\mathrm{i}}{\mathrm{L}}_{\mathrm{f}}+{\mathrm{E}}_{\mathrm{i}}{\mathrm{E}}_{\mathrm{f}}+{\mathrm{S}}_{\mathrm{i}}{S}_{\mathrm{f}}}$ Where the subscripts “i” and “f” represent “importance” and “functional scale” (Table 1).

WQCM-R scores were generated for each sub-watershed’s riparian corridors based on Eq. (1). Each riparian corridor’s score was associated and mapped onto the appropriate sub-watershed, and all sub-watersheds were then placed into one of four priority quartiles: low, moderate, high, and highest restoration priority. Scores can theoretically range from a low of 10 (a sub-watershed whose riparian zones are comprised of 100% barren land, 100% highly erodible soil, and 100% steep surface slope), to a high of 50 (riparian zones comprised of 100% forest or water, 100% low erodibility soil, and 100% negligible surface slope). This model intends watershed managers to focus riparian restoration efforts on lower scores, those that represent the best opportunities for watershed level water quality management.

Statistical analyses

Principal component analyses. Data for 20 variables characterized each sub-watershed’s riparian corridor: area of each of 7 land use classes; area of each of 8 soil erodibility classes, and; area of each of 5 surface slope classes. Principal component analyses were used to examine the sources of variation in the 20 variables and determine how many principal components were needed to explain at least 50% of the total variation, as well as determine if the variables that formed the largest eigenvectors of the first principal component corresponded with the variables that formed the best multivariate linear regression model.

Least squares regression. Least squares multivariate linear regression analysis was also applied to the WQCM-R results, seeking the best regression model using a maximum r-squared approach. Regression modeling set a sub-watershed’s riparian area WQCM-R score as the dependent variable, and each of the 20 riparian characteristics for the sub-watershed’s riparian area as the independent variables. The criteria for “best” included:

1. each variable must be logically consistent in terms of the sign of its coefficient

2. each variable must be statistically significant

3. the addition of each variable must increase the adjusted r² by a minimum of 0.05

4. the overall model must be statistically significant.