Landcover, Intertwine Regional Conservation Strategy

Executive SummaryPrior to November 2010, when The Intertwine Alliance launched the Regional Conservation Strategy (RCS) and Biodiversity Guide (RBG) efforts for the Portland-Vancouver metropolitan region, conservation priorities in the metropolitan region were identified at a broad regional scale that generally excluded urban areas (e.g., state conservation strategies and Willamette Synthesis); were regional but based solely on expert opinion (e.g., Natural Features); and consisted of localized priorities that abruptly ended at jurisdiction boundaries. The goal of the RCS was to fill in the gaps between broad and local scales of information related to conservation priorities. RCS members envisioned a data-driven approach that could add a regional perspective to local efforts and facilitate cross-scale cooperation toward protecting remaining valuable habitat in the Portland-Vancouver metropolitan region. Also, RCS members expected that the product would complement rather than replace local knowledge, by validating what is known and expanding to areas that are less well known.The RCS Technical Working Group hired the Institute for Natural Resources to develop a land cover layer at a spatial resolution appropriate for highly fragmented areas (5m versus the typical 30m), then to develop a conservation priority model that addressed both aquatic and terrestrial conservation needs. Several key products resulted from the project: the High-value Habitat Model describing high-value terrestrial habitat within the metropolitan region, the Riparian Habitat Model describing high-value habitat adjacent to streams and rivers, and the high spatial resolution land cover data set describing land cover at a 5 m spatial resolution. In June 2011, INR completed an initial proof-of-concept product describing high value conservation areas in the Portland-Vancouver region. The product demonstrated a methodology that enabled stakeholder involvement while also being data-driven. In May 2012, a second version of this product was completed. While the product is considered final at this time, it is expected and hoped that the models and data will be updated and improved upon into the future as more funds and better information becomes available so that the product functions as a “living work” rather than a one-time snapshot in time.Among the data used, the habitat prioritization modeling makes use of multiple data sets including high, 5 m spatial resolution imagery, improving on past efforts that were mapped at 30 m spatial resolution and nationally available data. The 5 m spatial resolution allows users to distinguish individual features on the landscape, such as large tree canopies. Because urban landscapes are widely diverse in terms of the vegetation types and types of surfaces (e.g., sidewalks, rooftops, plants, etc.), and many materials may be located in small areas, high resolution spatial data is essential to understanding and cataloging urban areas. The nationally available data allows the products to use spatially consistent data across the whole metropolitan region. Local data sets were used to supplement region-wide data sets. Data SummaryThe Intertwine High Resolution Land Cover data set (IHRLC) was developed in support of the Intertwine’s Regional Conservation Strategy effort to catalogue natural resources in the Portland-Vancouver metropolitan region. The land cover data set used for this project is the result of several stages of image classification and post-processing procedures. The initial stage land cover data set was a combination of 4 m and 30 m spatial resolution derived classification data; the 30 m data set filled in locations where 4 m data was unavailable. The high spatial resolution (4 m) portions of the classification map were developed using data from six LiDAR flights acquired from 2002-2009 (aggregated to a spatial resolution of 4 m) and National Agriculture Imaging Program (NAIP) imagery (0.5 m). The NAIP data was used to derive a 4 m spatial resolution Normalized Difference Vegetation Index (NDVI). The 30 m spatial resolution data was classified using a random forest classification technique. Normalized Differenced Vegetation Index (NDVI), Normalized Differenced Moisture Index (NDMI), and Tasseled Cap Wetness (TCW) spectral indices were used in combination with a digital elevation model (DEM), slope image, and atmospherically corrected and converted top of atmosphere (ToA) Landsat 5 Thematic Mapper bands 1-5 and 7 to complete the 30 m classification. Overall accuracy of the 30 m data set was 86%. The high and moderate resolution classified data were combined to create the first generation 4 m spatial resolution data set. Land covers within the urban growth boundary (UGB) classified as agriculture were reclassified as residential land cover in a post-processing step. This data set was then aggregated to a 5 m spatial resolution. The second generation 5 m spatial resolution data set was created by applying rule-based post-processing techniques to the first generation data set. Rules were used to distinguish between land cover such as agriculture and low-stature vegetation that were not well separated in the classification process. Rules were based on location relative the urban growth boundary (UGB) and elevation (600 feet above sea level; Table 1). The resulting classification contained 33 classes.The second generation data classes were also aggregated to yield two coarser levels, “level 1” and “level 0” classification schemes. The level 1 classification was not used for analyses, but is useful for display purposes. This classification resulted in 15 classes. Level 0 was created and used for regional statistics as well as cartographic purposes; it contains 6 generalized classes.Land cover (level 2): Developed originally by INR using LiDAR vegetation heights, National Agriculture Imagery Programimagery (NAIP; http://www.fsa.usda.gov/FSA/apfoapp?area=home&subject=prog&topic=nai), and Landsat ETM imagery.Augmented by Metro to more fully distinguish between land covers/land uses such as agriculture and low-stature vegetation.Primary data set used for analysisConsists of 33 classes (Table 1).Land cover (level 1):Level 1 land cover data set categories/classes were grouped to from the level 2 classification.Consists of 15 classes (Table 1). Land cover (level 0):Level 1 land cover data set categories/classes were grouped to form the level 2 classification.Consists of 6 classes (Table 1). Accuracy Assessment of Land Cover DataA heads-up accuracy assessment was completed. To assess accuracy, an analyst was provided with a data sheet and a set of points that were created through geographically stratified, random methods. The analyst used NAIP imagery to assess whether the conditions on the grown matched the land cover class. Overall accuracy was 94.3%. LimitationsSome of the earlier LiDAR flights were flown leaf off; resulting in inaccurate heights for deciduous trees in those areas. In areas where the 30 m classification was included, different land cover classes were developed due to the nature of the data sources. Classes in the 30 m data set were more generalized than the high spatial resolution classified data. Table 1. A nested classification scheme was developed during the course of this project. Level 0 was the most coarse classification level and level 2 was the most detailed.Level 0Class Name, level 0Level 1Class Name, level 1Level 2Class Name, level 2Class Description, level 21Water1Open water1WaterOpen water2Developed2Paved2Paved, built smallMost paved areas2Developed2Paved3Buildings (burned in), built mediumBuildings burned in From Metro's building layer and Clark County's building layer. Taller buildings (> 30 ft.) and other structures (e.g., bridges); includes some edge portions of the canopies of tall shrubs and short trees (sometimes very dark shadows from steep embankments/cliffs)2Developed4Buildings4Buildings (detected), built tallShorter buildings and other structures (e.g., bridges), semi trucks and rail cars; includes some edge portions of the canopies of tall shrubs and short trees (sometimes very dark shadows from steep embankments/cliffs)3Low vegetation5Herbaceous I - Low sparse veg (0 - 2 ft.)5Herbaceous, low, inside UGBSparse and/or very short vegetation (0 - 2 ft.; e.g., lawn); includes some water with emergent or submersed vegetation, vegetation canopy overhanging water surfaces, or shadows cast on water surfaces; may also include ball fields, mowed areas, golf courses, etc.)3Low vegetation5Herbaceous I - Low sparse veg (0 - 2 ft.)6Herbaceous, medium, inside UGBFairly sparse and/or short vegetation (2 - 5 ft.; e.g., crops, pastures, lawn, Phalaris); may include ball fields, mowed areas, golf courses, etc.3Low vegetation5Herbaceous I - Low sparse veg (0 - 2 ft.)22Reclassified to herbaceous, low, from developedBare ground/pervious surface with sparse vegetation; manual corrections made via heads-up digitizing; these pixels were originally classified as developed.3Low vegetation5Herbaceous I - Low sparse veg (0 - 2 ft.)27Herbaceous, low, outside UGBOUTSIDE UGB, > 600 ft. elevation - Sparse and/or very short vegetation (0 - 2 ft.; e.g., lawn); includes some water with emergent or submersed vegetation, or with overhanging vegetation canopy or shadow being cast on water surface3Low vegetation5Herbaceous I - Low sparse veg (0 - 2 ft.)28Herbaceous, medium, outside UGBOUTSIDE UGB, > 600 ft. elevation - Fairly sparse and/or short vegetation (2 - 5 ft.; e.g., crops, pastures, lawn, Phalaris)3Low vegetation7Herbaceous II - Low vegetation (2 - 7 ft.)7Herbaceous, high, inside UGBHerbaceous (5 - 13 ft.; e.g., low shrubs, tall crops, medium-sized shrubs, medium-sized tree regeneration); may include ball fields, mowed areas, and golf courses3Low vegetation7Herbaceous II - Low vegetation (2 - 7 ft.)29Herbaceous, high, outside UGBOUTSIDE UGB, > 600 ft. elevation - Fairly sparse and/or short vegetation (5 - 13 ft.; e.g., crops, pastures, lawn, Phalaris)4Tree cover13Large shrub/small trees (7 - 30 ft.)8Conifers, smallConifer woody crops, tall shrubs, small trees, largely tree regeneration (13 - 30 ft.)4Tree cover13Large shrub/small trees (7 - 30 ft.)13Hardwood, smallWoody crops, tall shrubs, small trees (e.g., willow, ash), large tree regeneration (13 - 30 ft.)4Tree cover9Conifers (30-120 ft.)9Conifers, mediumConifers 30 - 70 ft. tall; includes some broadleaved trees with shaded canopies, adjacent to water, or with bright, sparsely vegetated backgrounds (e.g., in urban environments)4Tree cover9Conifers (30-120 ft.)10Conifers, medium - tallConifers 70 - 120 ft. tall4Tree cover14Broadleaf (over 30 ft.)14Hardwood, mediumBroadleaved trees 30 - 70 ft. tall (e.g., ash); includes some conifers with brightly illuminated canopies4Tree cover14Broadleaf (over 30 ft.)15Hardwood, medium-tallBroadleaved trees 70 - 120 ft. tall (e.g., red alder)4Tree cover14Broadleaf (over 30 ft.)16Hardwood, tallBroadleaved trees > 120 ft. tall (e.g., big leaf maple, cottonwood)4Tree cover11Conifers (over 120 ft.)11Conifers, tallConifers 120 -200 ft. tall4Tree cover11Conifers (over 120 ft.)12Conifers, very tallConifers > 200 ft. tall, old growth4Tree cover55Mixed forest from low resolution (non-LiDAR) areas55Mixed forestMixed forest from low resolution (non-LiDAR) areas4Tree cover56Conifers from low resolution (non-LiDAR) areas56ConiferConifers from low resolution (non-LiDAR) areas4Tree cover57Hardwoods forest from low resolution (non-LiDAR) areas57HardwoodHardwoods from low resolution (non-LiDAR) areas4Tree cover17Clear cuts17Clear cuts, oldestSome cuts detected from 2000 or even earlier, most likely is representative of herbaceous or even shrub by now. 4Tree cover17Clear cuts18Clear cuts, 2006-2008Clear cut between 2006 and 2008, most likely is representative of herbaceous or bare ground.4Tree cover17Clear cuts19Partial cuts, 2006-2008Less than 50% volume removal, most representative of mature conifer forest >= 70 ft.4Tree cover17Clear cuts20Clear cuts, 2008-2010Clear cut between 2008 and 2010, representative of bare ground.4Tree cover17Clear cuts21Partial cuts, 2008-2010Less than 50% volume removal, most representative of mature conifer forest >=70 ft.4Tree cover17Clear cuts41Digitized clear cuts OUTSIDE UGB, > 600 ft elevation, patches > 4 acres; manually identified areas of herbaceous classes larger than 4 acres that resembled clear cuts5Agriculture26Agriculture26Agriculture, reclassified (inside UGB)Manually digitized agriculture inside the UGB, < 600 feet elevation, patches > 4 acres5Agriculture26Agriculture36Agriculture, digitized (outside UGB)OUTSIDE UGB, < 600 ft elevation, patches < 2 acres; manually identified5Agriculture26Agriculture40Agriculture, digitized (outside UGB)OUTSIDE UGB, > 600 ft elevation, patches > 4 acres; manually identified areas of herbaceous classes larger than 4 acres that resembled agriculture6Sand bars61Sand bars61Undeveloped areas; sandbarsFormerly paved pixels (class ID = 2) near rivers; manually reclassified; class composed mostly of sand bars