Study Sites
The present study involves two salmonid-bearing waterways in the metropolitan area of Portland, Oregon: (1) Johnson Creek begins in an agricultural region east of Portland, receiving multiple tributaries and draining 138 km2 while flowing westward for 42 km through agricultural, natural, suburban, urban, and industrial habitats before emptying into the Willamette River near Milwaukie, Oregon (Lee and Snyder 2009). (2) The Columbia Slough is a slow-moving, tidally-influenced, and heavily leveed water body running parallel to the Columbia River and draining 132 km2 while flowing 33 km westward from Fairview Lake in Troutdale, OR to the confluence of the Willamette and Columbia Rivers (Columbia Slough Watershed Council 2023) through mostly industrial and natural areas. The riparian areas of both waterways support Oregon ash and both waterbodies have been identified by the State of Oregon as impaired due to high water temperatures (DEQ 2022). Each of these streams can be approximately bisected at their midpoints into a western reach that is dominated by urban landscape features and an eastern reach dominated by exurban and rural landscape features (Fig. 1). As such, we separately analyzed these west and east reaches.
To randomize geographic sampling, we used ArcGIS 10.7.1 software (ESRI 2019) to subdivide our study’s streams into 30 m lengths and assign a unique ID to each segment. These IDs were assigned random numbers using the ‘random()’ function in Microsoft Excel, and sequentially sorted by random number. The first 5% of rows (quadrats) for each stream reach were selected as sample sites. We wanted to evenly split our sample sites between north bank and south banks and so used the same randomization procedure to evenly assign sample sites to these banks. Finally, these sampling sites were exported back to ArcGIS, and geoprocessing functions allowed us to create 10 m x 30 m bands adjacent to these stream reaches/segments (quadrat design is further explained in the next section and illustrated in Fig. 2).
Modeling Scenarios
We modeled three scenarios as part of this effort to compare conditions on the landscape and concomitant effects on solar loading:
-
Current Conditions: an assessment of riparian shade as represented in 2019 LiDAR data, before the functional extinction of Oregon ash due to EAB.
-
System Potential: a hypothetical future condition representing the maximum potential riparian shade. All existing canopy vegetation from 2019 is assumed to have reached full-grown heights and all areas with currently no canopy vegetation are assumed to have full-grown trees. This scenario represents the maximum possible effective shade in each stream.
-
Ash Loss: modeled riparian shade based on 2019 conditions, but assuming complete removal of Oregon ash from the landscape due to EAB.
To simulate the impact of the complete loss of Oregon ash, we modified the “Current Conditions” scenario described above by removing portions of the existing riparian canopy based on the ash abundance estimated by the field surveys. In ArcGIS we subdivided the three 10 m x 30 m bands every 10 m to create areas that are approximately 10 m x 10 m. The areas were designed to approximate the canopy extent of a fully grown Oregon ash tree. We then randomly selected a subset of these 10 m x 10 m areas. The number of selected areas were modeled as having no trees present to simulate the loss of ash trees. We generated and modeled 100 scenarios to evaluate the loss of riparian canopy by EAB.
Field Surveys to Estimate Canopy Contribution of Oregon ash:
Field surveys served two objectives: (1) to quantify the proportional contribution of Oregon ash to the riparian canopy to forecast the impacts of canopy loss, and (2) provide precise counts of Oregon ash boles while characterizing the distribution of all co-dominant and dominant riparian shrubs and trees. Our survey focused on 30 m-wide riparian zones directly above ordinary high-water of both river-right and river-left along the full length of the streams—a total potential survey area of 2.59 km2 and 2.05 km2 for Johnson Creek and the Columbia Slough, respectively. We sampled the total riparian area by surveying the randomly assigned 30 m2 quadrats described in the previous section (Fig. 2). We further subdivided each quadrat into three 10 m x 30 m bands running parallel to the stream to provide greater resolution for two reasons: (1) vegetation growing near the water’s edge typically provides more shade than plants growing further from the stream, and (2) fine scale habitat characteristics such as soil nutrient and moisture content can vary close to streams and may influence species distributions (Giller and Malmqvist 1998). Thus, these “bands” are our sampling units. We surveyed 429 bands from 143 quadrats along Johnson Creek, and 348 bands from 116 quadrats along the Columbia Slough, representing 5.2% and 5.3% of the total defined riparian areas, respectively.
Field surveys were conducted when leaves were present, between July and October 2020 (and prior to detection of EAB in the Pacific Northwest). Procedures varied among four scenarios:
Scenario 1: Desktop analysis We first employed recent aerial satellite photographs (Oregon Metro 2020) to visually identify sampling quadrats where Oregon ash would be excluded from field surveys due to structural or other land-use factors such as parking lots and roads (34 quadrats identified for this category). These sites nonetheless provided canopy cover data for the model, included as 0% canopy.
Scenarios 2–4: Field surveys
In the field, vegetation was functionally categorized in each band into overstory (tallest tree cohort within band), understory (trees > 2.0 meters but distinctly below overstory species), and shrub (≤ 2.0 meters). When the same species occupied two or more functional layers it was separately recorded within each category.
Scenario 2: No Oregon ash
When no Oregon ash was present, we conducted a rapid vegetation assessment to inform future mitigation and restoration efforts by recording up to five of the most abundant species within overstory, understory, and shrub functional layers.
Scenario 3: Overstory exclusively Oregon ash
For field surveys where Oregon ash was the only overstory species within a given band, we recorded numbers of Oregon ash boles, and measured their height ranges within each band using a laser rangefinder. As in Scenario 2, we documented up to five co-dominant species within the understory and shrub layers.
Scenario 4: Co-dominant overstory
When Oregon ash was co-dominant with other species in the overstory, we recorded numbers of Oregon ash boles, measured their height ranges, and estimated percent canopy contribution as the average of two independent visual estimates. Up to five co-dominant species within the understory and shrub layers were recorded.
Canopy Composition Analysis
To quantify the overall canopy cover within randomly-selected stream-lengths, we used first-return LiDAR data collected in 2019 to generate a riparian canopy model, from which we quantified canopy heights, distribution, and the total canopy cover. The quadrat-specific zonal canopy cover results were combined with the associated field surveys to estimate the total area of each band composed of Oregon ash. From this we were able to calculate the relative proportion of the riparian canopy that was composed of Oregon ash.
Effective shade, thermal loading models
To assess the potential impacts of riparian canopy loss along Johnson Creek and the Columbia Slough, we modeled solar loading and effective shade using remote sensing data. Effective shade is the proportion of solar radiation that is attenuated or scattered before reaching the stream by the cumulative biotic and abiotic landscape features (Fig. 3). To calculate effective shade, we used the shade module of Heat Source version 26 (Michie et al. 2021), a computer model used and maintained by Oregon DEQ (Boyd and Kasper 2003). We modeled effective shade for both waterways from July 1 to August 31 at a 15-minute timestep. We summarized these results as the July–August mean at each modeling node: the summertime period when stream temperatures in the Portland area tend to be greatest.
The model relies on GIS inputs to characterize the surrounding topography and land cover features (including vegetation canopy cover) that affect the amount of solar radiation reaching the stream surface. For modeling effective shade of Johnson Creek and the Columbia Slough, these GIS inputs were compiled using TTools version 9.0 (Michie 2021), a GIS tool that assembles stream channel and land cover geospatial data for input into the Heat Source model.
For both waterbodies we modeled the effective shade every 25 m along the center of the channel. For each modeling node, we sampled the surrounding topography and land cover within 75 m of the center of the stream, in 3 m increments (Fig. 2).
Table 1
Inputs used to model solar loading, including the datasets used to characterize land cover surrounding the modeled streams and the Heat Source modeling parameters. See the References section for links to complete GIS metadata
Feature | Purpose | Input/Value |
Topography | Streambank and nearby topography used to calculate topographic shade. | 2019 bare earth LiDAR (City of Portland, 2021) |
Riparian Vegetation | Tree canopy height and tree type (coniferous or deciduous) used to characterize the riparian vegetation. Tree heights and canopy density attenuate solar loading. | 2019 first return LiDAR (City of Portland, 2021) ; Canopy Classification (Oregon Metro, 2016) |
Impervious Features | Identified buildings, roads, and parking lots used to construct land cover codes to characterize adjacent land use. Buildings assumed to provide shade. | Impervious Areas (City of Portland, 2017) |
Model Time Period | Range of dates for which incoming solar loading is calculated | July 1 to August 31 |
Model Time Step | Modeling time increment for each modeled day | 15 minutes |
Model Node Spacing | Distance between points along the stream line where solar loading is calculated | 25 meters |
Landcover Sampling Transects | Number of transects per steam node where adjacent landcover characteristics are sampled (e.g., tree heights). | 8 transects |
Landcover Transect Sample Spacing | Distance between landcover sampling along each transect at each stream node. | 3 meters |
Model Calibration
Riparian vegetation attenuates less than 100% of light; to account for this the Heat Source model includes a canopy density parameter that represents the proportion of incoming solar radiation that is blocked by vegetation. The canopy density values used in the initial model runs were based on the vegetation characteristics described in the Lower Willamette temperature TMDL (DEQ, 2006) and included a 75% density value for deciduous vegetation and an 80% density value for coniferous vegetation.
We refined the canopy density parameter using riparian canopy cover measurements recorded as part of the City of Portland’s watershed monitoring program (Portland Area Watershed Monitoring and Assessment Program; City of Portland 2023a) and the Bureau of Environmental Service’s Restoration Monitoring Program (City of Portland 2023b). Densiometer readings were collected and converted to canopy cover using the methods of Lemmon (1957). We found that the field measured canopy cover values were typically substantially higher than the TMDL canopy density values; particularly in riparian areas dominated by conifers. For the purposes of this assessment, we used the 25th percentile canopy cover values to better reflect measured field conditions. As a result, both density values were increased, with the density value for deciduous canopy of 78% and a density for coniferous canopy of 90%.