3 Methods

3.1 Field Assessments

3.1.1 Site Reviews

Field site reviews were conducted in September and early October 2024 to assess current conditions at historic prescription sites, past restoration locations, and newly proposed sites. Sites were selected based on:

  • Locations with documented historic prescriptions from MacKay et al. (1998)
  • Sites proposed for restoration by the Wet’suwet’en First Nation (Gaboury and Smith 2016)
  • Sites with completed restoration work (Smith and Gaboury 2016)
  • Past Healthy Watersheds Initiative restoration sites
  • Locations of documented traditional use fishing sites (Gottesfeld and Rabnett 2007; Wilson and Rabnett 2007)
  • Newly proposed sites identified through landowner engagement and local knowledge
  • Erosion protection sites in the Fraser watershed for comparison


At each site, field crews documented current riparian conditions, evidence of cattle access and bank erosion, existing infrastructure, and restoration potential. Standardized digital forms were developed for the project and deployed within the collaborative GIS environment (Mergin Maps), with photos georeferenced and linked to site records. UAV flights were conducted at select sites to provide high-resolution baseline imagery for monitoring. Raw field data is stored within the shared QGIS project in the Project Specific/Field Data/2024 group.

3.1.2 Benthic Invertebrate Sampling

Benthic invertebrate sampling was conducted following the Canadian Aquatic Biomonitoring Network (CABIN) wadeable streams protocol (Environment Canada 2012). Samples were collected using the traveling kick-net method with a 400 μm mesh net. At each site, a 3-minute timed sample was collected by traversing a zigzag pattern in an upstream direction within the erosional zone (riffle, rapid, or fast-flowing run habitat).


Concurrent with benthic sampling, standardized habitat and water quality data were collected. Substrate composition was characterized using a 100-pebble count, measuring the intermediate axis of randomly selected particles along transects. Embeddedness—the degree to which particles are buried in fine sediments—was assessed for every 10th particle. Channel slope was measured using a lazer level.


Water quality parameters recorded at each site included temperature, pH, and conductivity. Instream habitat features documented included macrophyte coverage (percent cover in categories: 0%, 1–25%, 26–50%, 51–75%, 76–100%), periphyton coverage (5-category scale based on slipperiness and thickness), and canopy cover. Channel dimensions including bankfull width, wetted width, and depths were also recorded.


Three sites were sampled on the Neexdzii Kwah mainstem within known high-value Chinook spawning and rearing habitat. Site locations were selected to assess aquatic health upstream and downstream of potential point-source impacts, allowing comparison of species composition and community health metrics across the watershed. Triplicate kick samples were collected at each site to quantify within-site variability and enable statistical distinction between true site differences and natural spatial heterogeneity.


Samples were preserved in 10% formalin and submitted to Cordillera Consulting Inc. (Summerland, BC) for sorting and taxonomic identification to genus/species level. Results will be uploaded to the federal CABIN database.

3.2 Remote Sensing & Imagery

3.2.1 Historic Aerial Photographs

Provincial aerial photograph coverage of the Neexdzii Kwah watershed spanning 1963 to 2019 was georeferenced and served as a STAC collection using the fly R package and the stac_airphoto_bc pipeline. The fly package extracts photo centre coordinates and camera geometry from provincial metadata, computes approximate ground footprints, applies rotation correction, and fetches thumbnail images from the BC Government photo warehouse. Thumbnails were converted to Cloud-Optimized GeoTIFFs (COGs) with their computed spatial footprints, cataloged according to the STAC specification, and served via the same API endpoint used for UAV imagery at https://images.a11s.one. The resulting collection contains 9,741 georeferenced aerial photographs queryable by spatial extent and time range using rstac or QGIS 3.42+ (Simoes et al. 2021; Radiantearth 2024).


These are free provincial thumbnails — sufficient for establishing baseline conditions, identifying change trajectories, and selecting specific photos of interest for high-resolution acquisition. The georeferencing and rotation correction allow direct overlay on modern imagery for visual comparison of conditions across 60 years. This collection forms part of an integrated STAC-based analysis stack: flooded delineates floodplain extents that become areas of interest for photo queries, drift quantifies recent land cover change (2017–2023) to identify where change is occurring, the airphoto collection documents what was there before, and stac_uav_bc provides high-resolution current condition from UAV flights. Together, the pipeline grounds restoration prescriptions in observed history rather than modelled targets alone.

3.2.2 UAV Imagery

Scripted processing and serving of UAV imagery collected during the project is available at https://github.com/NewGraphEnvironment/stac_uav_bc/ (Irvine [2025] 2025). OpenDroneMap was utilized to produce orthomosaics, digital surface models (DSMs), and digital terrain models (DTMs) (OpenDroneMap Authors [2014] 2025). To support efficient web-based access - imagery products were converted to cloud-optimized GeoTIFFs (COGs) using rio-cogeo, then collated according to the SpatioTemporal Asset Catalog (STAC) specification with pystac and uploaded to S3 storage Amazon Web Services (2025). A titiler tile server was set up to facilitate interactive viewing of the orthoimagery and an Application Program Interface (API) leveraging stac-fastapi-pgstac is served at https://images.a11s.one to enable linking of collection images through QGIS as well as remote spatial and temporal querying using open source software such as rstac (Development Seed [2019] 2025; stac-utils 2025; Simoes et al. 2021).

3.2.3 Land Cover Classification

To focus the land cover analysis on areas ecologically relevant to fish habitat — rather than measuring change across the full watershed including uplands — modelled floodplain extents were delineated using the Valley Confinement Algorithm (VCA) implemented in the flooded R package. The VCA estimates bankfull flood depth from upstream contributing area and mean annual precipitation, then identifies valley-bottom cells below the modelled flood surface that also meet slope, cost-distance, and proximity criteria. Stream networks were extracted for coho rearing/spawning habitat on streams of 3rd order and greater using the fresh R package, with lakes and wetlands included to fill gaps where water surfaces read differently than surrounding terrain. The flood_factor parameter (set to 6 for the baseline scenario) scales bankfull depth to approximate the functional floodplain — the area regularly influenced by high flows that sustains riparian processes critical to salmon habitat. Sub-basins were defined interactively using a watershed picker tool that snaps user-selected points to the BC Freshwater Atlas stream network and delineates contributing areas via fwapg. Within each sub-basin’s modelled floodplain, IO Land Use Land Cover imagery (10 m, Sentinel-2) was fetched for 2017, 2020, and 2023 from Microsoft Planetary Computer via STAC queries and classified using the drift R package. At this resolution, Crops, Rangeland, and Bare Ground were grouped into an “Agriculture” superclass because a single field can be classified differently depending on the satellite overpass date relative to the growing season. Sub-basin boundaries are adjustable — reanalysis requires only re-running the pipeline with updated break points. Full results are presented in Appendix - LULC.

3.2.4 Climate Anomaly Analysis

Regional climate trends for the Neexdzii Kwah watershed were analyzed using ERA5-Land hourly reanalysis data from the European Centre for Medium-Range Weather Forecasts, accessed via the Copernicus Climate Change Service. ERA5-Land is a gridded reanalysis dataset combining observational and model data at approximately 9 km spatial and hourly temporal resolution. Climate anomalies were computed relative to 1981–2010 climatological normals (30-year means). For temperature, vapor pressure deficit, and relative humidity, anomalies were calculated as absolute deviations from the normal (\(x - \bar{x}\)). For precipitation and soil moisture, anomalies were expressed as percentage of normal (\(\frac{x - \bar{x}}{\bar{x}} \times 100\)). Anomaly data were regridded to approximately 30 km spatial resolution and cropped to the study area boundary.

Trend magnitude was estimated using the Theil-Sen estimator and statistical significance assessed using the Mann-Kendall test for two periods: 1950–present and 1980–present. Climate parameters analyzed included mean, maximum, and minimum temperature, total precipitation, vapor pressure deficit, relative humidity, and volumetric soil moisture (0–1 m depth). Seasonal (winter, spring, summer, fall) and annual summaries were extracted. Spatial anomaly maps were generated using the BC Climate Anomaly app developed by the BC Ministry of Forests Future Forest Ecosystems Centre. Tabular summaries were extracted programmatically from the same NetCDF source data. Full spatial anomaly maps are presented in Appendix - Climate Anomaly Data.

3.3 Background Research & Analysis

3.3.1 Historic Information

Historic information regarding impacts and restoration initiatives, prescriptions and activities is an important component of the restoration planning process. A wealth of information is present for the Neexdzii Kwah however it is located in many different locations and not always spatially represented. To address this, a comprehensive review of past impact indentification and restoration recommendations is underway. This review included a review of the literature, interviews with local experts, and a review of the restoration prescriptions and activities that have been conducted in the Neexdzii Kwah. For information regarding relatively recent restoration efforts such as physical works conducted through Healthy Watersheds Initiative, representatives from the Morice Watershed Monitoring Trust amalgamated all information located on physical drives and either provided emails with the data attached or provided links to the data for the study team for download and review. This information was then subsequently downloaded and re-uploaded to a OneDrive folder for the study team to access.

3.3.2 Future Restoration Site Selection

3.3.2.1 Evaluation of Historic and Current Imagery

Evaluation of historic and current imagery to understand and quantify watershed characteristics and morphological changes in the Neexdzii Kwah and its major tributaries over time and guide future restoration efforts. This will require that the study team acquire, georeference, archive and analyze historic imagery for the Neexdzii Kwah watershed and compare with an analysis of recent data to quantify historic changes in stream morphology resulting in loss of quantity and quality of water and fish habitat. Through this process we hope to highlight areas of historic dredging, realignment, and floodplain disconnections due to infrastructure that have degraded watershed health and fish habitat.

3.3.2.2 Fish Passage

At the time of writing, extensive work related to fish passage restoration planning was underway in the Neexdzii Kwah watershed as well as other areas of the greater Skeena watershed. Methodology for this work is presented in Irvine and Schick (2025), Irvine and Schick (2024), Irvine and Wintersheidt (2023), Irvine et al. (2023), Irvine (2021) and Irvine (2018). Although these references detail work primarily related to linear connectivity (ie. upstream/downstream) there is also ongoing work related to lateral connectivity (ie. floodplain connectivity) with methodology related to disconnections due to the linear infrastructure documented in bcfishpass (Norris [2020] 2024) here. The results of this analysis and future analysis incorporating major roadways as well (currently under development) will be added to the shared QGIS project and used to inform future restoration activity prioritization.

3.3.2.3 Sub-Basin Prioritization via Land Cover Change

To identify reaches with the greatest riparian disturbance, sub-basins were ranked by tree cover loss and agriculture gain within the modelled floodplain between 2017 and 2023. This ranking provides the spatial framework from which specific restoration sites are selected. Sub-basin boundaries are not fixed — they can be adjusted through the interactive watershed picker and the analysis re-run at minimal cost, making the approach adaptable as project understanding evolves. Details are presented in Appendix - LULC.

3.3.2.4 Local Knowledge for Site Selection

To facilitate project activities in the short term - the project team conducted site visits to areas of known riparian removal and bank erosion where landowners are amenable to project activities including erosion protection, installation of cattle exclusion fencing and riparian planting.

3.3.2.5 Delineation of High Fisheries Value Areas

Past work to spatially delineate areas of high value habitat known to be utilized historically for chinook and sockeye salmon spawning (among other data) undertaken by DFO, Arocha Canada and others has been stored within a secure location and linked within the shared GIS project. Wilson and Rabnett (2007) includes descriptions of traditional Wet’suwet’en Fisheries sites within the Neexdzii Kwah which will be spatialized for the project.

3.3.2.6 Parameter Ranking

Analysis of background and current data can be used to inform priority ranking criteria used to select sites for restoration activities in the long term. Although the process is ongoing - GIS and user input parameters were selected based on ongoing project team meetings and review of available spatial information to rank future restoration sites. Using custom functions and scripted workflows based on layers kept within the shared GIS project - we prioritized proposed sites dynamically based on a range of metrics and user defined ranking. Scripts to conduct the analysis can be found here with custom functions here.

3.4 Collaborative Data Management

3.4.1 GIS Environment

A collaborative GIS environment (restoration_wedzin_kwa) was established using QGIS and Mergin Maps to enable project team members to view, edit, and analyze shared spatial data on desktop computers and mobile devices in the field. The environment is used to develop and share maps, conduct spatial analyses, communicate restoration plans to stakeholders, and provide standardized digital forms for field assessments. The platform also tracks restoration activity progress and monitors landscape changes over time.


The QGIS project was created using scripts in dff-2022 that:

  • download and clip layers from the BC Data Catalogue and custom AWS buckets for the watershed area of interest
  • create a project directory with layer symbology, naming conventions, and metadata
  • store layer metadata in the rfp_tracking table within background_layers.gpkg, along with user-supplied stream width/gradient inputs to bcfishpass for modeling high-value fish habitat

3.5 Open Source Reporting Framework

This report is produced using open-source tools (R, bookdown) with full version control via git, enabling iterative updates as restoration planning progresses. All code, data, and revision history are publicly accessible in the project repository, with ongoing tasks tracked via GitHub issues and version history maintained in the project changelog.