Historical fire regimes Current fire severity Fire-regime condition class Fire-behavior fuel models Crown fire behavior Crown bulk density Height-to-crown Stand height Wildlife habitats Weed susceptibility and threat
General Limitations
These data were derived using field plots from many different sources (e.g., FSVEG, ECODATA, FIA, DNRC) as well as remotely sensed data (e.g., satellite imagery, DEMs). The sampling designs for collecting these data were not intended to sample across environmental gradients. The spatial distribution of field plots was extremely variable. In general, expected accuracy is believed to be much lower in areas where plot data was sparse and relatively higher in areas with concentrated plot locations.
These data were designed to characterize broad scale patterns for regional and subregional assessments. Any decisions based on these data should be supported with field verification, especially at scales finer than 1:100,000. Although the resolution of the PNV theme is at a 90 meter cell, the expected accuracy does not warrant their use for analyses of areas smaller than about 10,000 acres (for example, assessments that typically require 1:24,000 data).
The data provide a coarse-filter approach to ecosystem assessments. Consequently, not every occurrence of every PNV is mapped; instead, only larger, more generalized distributions of certain types were mapped.
Input data sources Digital Elevational Model (DEM), DAYMET climate datasets, Landcover Vegetation datasets (SILC1 and SILC3), Interior Columbia Basin Subwatershed (HUC), and PNV point data were used as is. Accuracy of the attributes as well as the positional accuracy can be found with the respective metadata. US Forest Service Timber Stand Management Record System and stand coverages were used as is. Accuracy of this dataset is not known.
Along the edges of this dataset, either Potential Natural Vegetation (PNV) information, SILC1/SILC3 information, or both are missing in some areas. These areas are either right along the edge of the analysis area, or in the buffer zone between the actual USFS Northern Region boundary and the edge of the analysis area.
Although this dataset is listed as 1:100000 and its parent dataset at 1:60000, the raster data is better described by spatial resolution; in this case 30 meters.
Although this dataset is listed as 1:60000, the raster data is better described by spatial resolution; in this case 30 meters.
In combining east- and west-side cover, canopy, size and lifeform grids, there is an overlap zone between SILC1 and SILC3 datasets. This particular PNV grid utilized HUC boundaries to cut SILC3 data on the western edge following the Beaverhead-Deerlodge NF boundary with the Bitterroot NF northward, then across the Clark Fork and Blackfoot rivers to the boundary between the Lolo NF and Helena NF moving northeast to the Continental Divide, then northward along the Divide to the northern edge of the SILC3 dataset. This boundary line followed 4th and some 5th code HUC boundaries, except where it followed 6th code HUCs in the Clark Fork river area. The version is often referred to as 'HUC' because HUC boundaries were used to split the SILC datasets.
Lifeform potential was created by combining the existing lifeform (LF EXIST huc) and regeneration harvest (HARV) datasets into one grid. If the existing lifeform was 'shrubland', 'grassland', or 'rock/barren', and there had been a regeneration harvest, the potential lifeform became 'forest'.
In combining cover grids, a new list of possible covertypes for the entire analysis area was developed. There appeared to be some "duplicate" covertypes, i.e. different covertypes that are essentially the same. After discussions with Chip Fischer, Wildlife Spatial Analysis Lab, the following covertypes were re-coded to similar, if not identical, covertypes: 3150 to 6200; 3170 to 6200; 3610 and not riparian to 3200; 3610 and riparian to 6300; 4150 and not riparian to 4140; 4150 and riparian to 6120; from 4203 through 4244 and riparian to 6110; 3625 to 3350; 4142 to 4260; 4242 to 4270; 4244 to 4290. To simplify some of the rock and barren covertypes, 7800 was recoded to 7300.
Since canopy and size classes did not use the same classes, the combined datasets were recoded in a way to keep all the original west- and east-side classes.
In earlier runs of the k-nearest neighbor (KNN) process, western Continental Divide areas where potential lifeform was labeled 'clouds' and 'cloudshdw' were not assigned Potential Natural Vegetation (PNV) types correctly; the erroneous PNV classification stood dramatically out. 'Clouds' and cloudshdw' did not affect PNV classification east of the Divide as dramatically, possibly due to the fact that nearly all clouds occurred in grassland areas. To fix the error for the west-side, 'clouds' and 'cloudshdw' cells were eliminated using the ArcGrid's NIBBLE function, allowing all lifeform's except 'water' and 'urban' to fill in 'clouds' and 'cloudshdw'.
The east-side area had a substantial amount of 'snow/ice' in the Absaroka Mountains. Using the earlier East-side Potential Vegetation Type (EAST PVT) grid, 'snow/ice' in the potential lifeform grid was recoded to 'rock/barren' anywhere the east-side PVT grid had 'rock/barren'.
The grid was aggregated to 90 meter cell size. The potential lifeform grid from this step was only used in the KNN process.
The data was clipped to the Forest Service Northern Region boundary, reprojected and clipped down further to the analysis area and changing the cell size from 1000 meter to 90 meter. The floating point grids were converted to integer and recoded to reduce the number of classes to 32 based upon natural breaks (Jenks).
For grasslands, an individual grid of potential grassland lifeform was created. With the grassland grid and strata polygons in the background, 1-3 on-screen points were added in locations where large blocks of grassland lifeform occurred and only in those strata lacking grass PNV plot data. 'Drygrass' PNV was assigned to these points.
For shrublands and using the covertype grid, individual grids of sagebrush, saltshrub, mesic shrubland, and xeric shrubland were created. Following the same procedure used for grasslands, points were added for those strata polygons lacking shrubland PNV plot data. 'Bigsage', 'saltshrub', 'messhrub', or 'dryshrub' PNV was assigned to each point, based upon which shrub grid was in the background.
The forested area southeast of Lewistown, ID and north of the Salmon River, did not have any plot data. This area is described as predominately Douglas-fir and Grand Fir PNV types, although earlier runs of KNN assigned Western Red Cedar and Western Hemlock to the area since these were the predominate PNV types for this particular strata. Dummy points of psme2 and abgr2 PNV types were added for this area using covertypes of 4212 (Douglas-fir) and 4207 (Grand Fir) in the background. The final run of KNN had re-delineated the strata polygon for this area to include the problem area in the Salmon River strata polygon which also helped eliminate KNN assignment errors.
Due to the large number of potential classes, the 6 input grids were combined into 2 grids: one with strata and climate data; and other with strata, topographic, and potential lifeform data. Convert the raster grids into point coverages bringing all attributes across. With the analysis area covering a large extent, the remaining steps had to be processed by individual strata polygons due to software limitations. For each strata the following was completed: 1) select the individual strata polygon; 2) buffer the polygon 2km; 3) clip the plot data, climate input, and topography & lifeform input to the buffered strata polygon; 4) using Arc/Info NEAR and JOINITEM, combine the attributes of the climate data onto the topography and lifeform input point coverage; 5) using Arc/Info NEAR and JOINITEM programs attribute the plot data with the climate, topographic, potential lifeform and strata polygon information; 6) clip the combined climate, topography, lifeform, and strata input point coverage to the exact strata polygon boundary; 7) on the plot data input, correct the potential lifeform attribute where lifeform is forest, shrub, or grass to match the PNV type, i.e. abgr1, abgr2, abla1, etc. should be 'forest' potential lifeform; 8) add the x,y coordinates to the attribute table using Arc/Info ADDXY; 9) create a DBF file for each of the two inputs.
The continuous input layers were standardized by scaling all input points and grid data for each strata unit between zero and one, where zero and one represented the minimum and maximum values, respectively. The categorical input layers aspect and lifeform were defined as n Boolean attributes where n represented the number of categories.
An expert then weighted the point and grid attribute values where each PNV type was guaranteed to fall within its potential lifeform type. The weighting then assumed that aspect, slope, and elevation received 70% of the weighting equally distributed. The x and y coordinates received 10% equally distributed. The remaining 20% was distributed amongst the three climate variables with precipitation and growing degree days each receiving 40% and solar radiation receiving 20%.
With the standardized and weighted data, the KNN algorithm inferred a PNV type for each cell in the strata unit by finding the PNV point closest to the cell. Each cell was compared against all the PNV points by computing the Euclidean distance. The PNV point with the minimum distance to the given cell was selected as the PNV type for that cell. If there was a tie, all points with that minimum distance were compared with one another and the dominant PNV type was selected. If there was a tie selecting the dominant PNV type, the PNV type was assigned randomly amongst the remaining PNV types.
Reference for KNN algorithm: Mitchell, Tom M. 1997. Machine Learning. McGraw Hill College Division. 432p. ISBN: 0070428077.
The individual strata point coverages were appended together and subsequently re-rasterized into a 90 meter grid.
Clean-up of the vegetation and PNV attributes involved looking at the tabular data to fix where the combination of the attributes do not match. For example PNV shows 'water', but the covertype is not 'water'. To facilitate this, a DBF file was pulled from the grid VAT file and brought into Microsoft Access. In order, the type of clean-up that was accomplished included: 1) where the PNV label was 'water', 'urban', 'rock/barren', or 'agriculture', the cover, canopy, size and lifeform items were changed to match; 2) within 'forest' potential lifeform and for each forested PNV type, ensuring that the covertypes would occur for that PNV type, and if not, changing the covertype to an appropriate generalized code such as 4260, 4270, 4280, or 4290; 3) for 'pial' PNV type where the covertype did not match, it was discovered that some of the mapping had 'pial' at too low of elevation, so the PNV was changed for those non-matching covertypes to 'abla3'; 4) for 'pico' PNV type where the covertype did not match, the PNV value was changed to 'pifl', 'abla3', or 'psme1' as appropriate; 5) if the PNV was 'alpine' and the covertype was a forested type, the PNV value was changed appropriately to either 'pial' or 'abla3', and if needed the covertype was modified as well; 6) for grassland and shrubland PNV types, the covertype, canopy, size, existing and potential lifeform items were changed to match; 7) where potential lifeform was 'water', 'urban', 'rock/barren', or 'agriculture', the PNV type was changed to match.
For areas with little or lacking plot data, 'ripdecid' became difficult to model. To fix this problem as best as possible, the following was completed. Where PNV was 'ripdecid', existing and potential lifeform were 'forest' and covertypes were pure conifer types between value 4000 and 4290, the PNV value was changed appropriately to 'juniper', 'pifl', 'psme1', 'psme2', 'pipo', abla1', 'abla3', or 'thpl1' depending on whether the strata polygon was west or east of the Continental Divide. Additionally, where PNV was 'ripdecid' and the covertype was a grassland or shrubland type, the PNV value was changed to the appropriate grass or shrub PNV.
Based upon the east-side SILC3 data not having information for shrub height or canopy cover and the west-side SILC1 data not having grassland covertypes split by percent cover, all grassland and shrubland covertypes where assigned 'non-conifer' on east and 'no information' on west for canopy and size items. Essentially, only the forested PNV and covertypes would have canopy and size infomation.
Code Descrip Written Description
1 abgr1 Abies grandis (Grand Fir) dry type 1 2 abgr2 Abies grandis (Grand Fir) moist type 2 3 abla1 Abies lasiocarpa (Subalpine Fir) wet type 1 4 abla2 Abies lasiocarpa (Subalpine Fir) moist type 2 5 abla3 Abies lasiocarpa (Subalpine Fir) dry type 3 6 abla4 Abies lasiocarpa (Subalpine Fir) cold type 4 7 juniper Juniperus (Juniper) species type 8 laly Larix lyallii (Subalpine Larch) type 9 pial Pinus albicaulis (Whitebark Pine) type 10 picea Picea (Spruce) species type 11 pico Pinus contorta (Lodgepole Pine) type 12 pifl Pinus flexilis (Limber Pine) type 13 pipo Pinus ponderosa (Ponderosa Pine) type 14 poptre Populus tremuloides (Aspen) and upland deciduous type 15 psme1 Pseudotsuga menziesii (Douglas-fir) warm dry type 1 16 psme2 Pseudotsuga menziesii (Douglas-fir) moist type 2 17 psme3 Pseudotsuga menziesii (Douglas-fir) cool dry type 3 18 thpl1 Thuja plicata (Western Red Cedar) wet type 1 19 thpl2 Thuja plicata (Western Red Cedar) moist type 2 20 tshe Tsuga heterophylla (western Hemlock) type 21 tsme1 Tsuga mertensiana (Mtn. Hemlock) without Whitebark Pine type 1 22 tsme2 Tsuga mertensiana (Mtn. Hemlock) with Whitebark Pine type 2 23 agriculture Agricultural lands 24 agrsmi Agropyron smithii (Western Wheatgrass)grassland type 25 fesida Festuca idahoensis (Idaho Rescue) grassland type 26 fessca Festuca scabrella (Rough Fesceu) grassland type 27 drygrass Dry species grassland type 28 bigsage Artemesia (Big Sagebrush) species shrubland type 29 potfru Potentilla fruticosa (Shrubby Cinquefoil) shrubland type 30 rhus Rhus (Sumac) species shrubland type 31 saltshrub Atriplex (Saltshrub) species shrubland type 32 dryshrub Dry species shrubland type 33 messhrub Mesic species shrubland type 34 ripdecid Riparian deciduous forest type 35 alpine Alpine and alpine grassland type 36 rock/barren Rock, talus/scree, barren areas, and mines 37 water Water, lakes, reservoirs, and wide rivers 38 urban Urbanized areas (cities and towns) 39 no information
For each PNV type, the habitat types are listed
abgr1: abigra/phymal; abigra/spibet
abgr2: abigra/asacau; abigra/cliuni; abigra/linbor; abigra/sentri; abigra/vacglo; abigra/xerten
abla1: abilas/calcan; abilas/descae; abilas/galtri; abilas/ledgla; abilas/oplhor; abilas/salix; abilas/stramp
abla2: abilas/alnsin; abilas/cliuni; abilas/linbor; abilas/menfer
abla3: abilas/arncor; abilas/calrub; abilas/cargey; abilas/clepse; abilas/vaccae; abilas/vacglo; abilas/vacsco; abilas/xerten; pincon/vacglo;
abla4: abilas/luzhit; abilas/pinalb; abilas/ribmon
juniper: junost/agrspi; junsco/agrspi; junsco/arttsv
laly: larlay/ ; larlya/abilas
pial: pinalb/ ; pinalb/abilas; pinalb/calcan; pinalb/cargey; pinalb/fesida; pinalb/ribmon; pinalb/vacsco
picea: picea/ ; picea/calcan; picea/carex; picea/cliuni; picea/corsto; picea/equarv; picea/galtri; picea/linbor; picea/phymal; picea/rosaci; picea/salix; picea/salwol; picea/senstr; picea/smiste; picea/thaocc; picea/vacces
pico: pincon/ ; pincon/carhel; pincon/purtri; pincon/vacsco; pincon/calrub; pincon/linbor; pincon/vaccae; pincon/xerten
pifl: pinfle/ ; pinfle/agrspi; pinfle/fesida; pinfle/junsco; pinfle/juncom; pinfle/symalb
pipo: pinpon/ ; pinpon/andx; pinpon/agrspi; pinpon/carhel; pinpon/fesida; pinpon/juncom; pinpon/phymal; pinpon/prubir; pinpon/purtri; pinpon/symalb
poptre: poptre/ ; poptre/arttsv; poptre/berrup; poptre/calcan; poptre/calrub; poptre/carex; poptre/equarv; poptre/junhor; poptre/osmocc; poptre/poapra; poptre/salix; poptre/salboo; poptre/salgey; poptre/shecan; poptre/symore;
psme1: psemen/agrspi; psemen/arttsv; psemen/fesida; psemen/fessca; psemen/junsco; psemen/symore
psme2: psemen/calrub; psemen/corsto; psemen/phymal; psemen/psemen; psemen/vaccae; psemen/vacsco; psemen/vacglo; psemen/xerten
psme3: psemen/arcuva; psemen/arncor; psemen/berrep; psemen/cargey; psemen/juncom; psemen/pincon/ psemen/spibet
thpl1: thupli/adiped; thupli/athfil; thupli/gymdry; thupli/oplhor
thpl2: thupli/asacau; thupli/cliuni
tshe: tsuhet/ ; tsuhet/asacau; tsuhet/athfil; tsuhet/cliuni; tsuhet/gymdry; tsuhet/menfer
tsme1: tsumer/cliuni; tsumer/stramp
tsme2: tsumer/luzhit; tsumer/menfer; tsumer/xerten
agriculture:
agrsmi: agrsmi/carfil; agrsmi/sticom; agrsmi/stivir
fesida: fesida/ ; fesida/agrcan; fesida/agrsmi; fesida/agrspi; fesida/balsag; fesida/carex; fesida/carfil; fesida/cargey; fesida/carhel; fesida/carhoo; fesida/danint; fesida/descae; fesida/koecri; fesida/poapra; fesida/stiric
fessca: fessca/ ; fessca/agrspi; fessca/fesida
drygrass: agrspi/ ; agrspi/agrsmi; agrspi/balinc; agrspi/balsag; agrspi/boucur; agrspi/carfil; agrspi/fesida; agrspi/lomtri; agrspi/poasan; agrspi/poasec; agrspi/sticom; agrspi/wyeamp; andsco/carfil; arilon/poasan; balsag/agrspi; balsag/bromar; callon/carex; danuni/balinc; spocry/poasan; sticom/ ; sticom/carfil; sticom/carhel
bigsage: artarb/ ; artarb/agrspi; artarb/fesida; artcan/agrsmi; artfri/fesida; artnov/agrspi; arttri/ ; arttri/agrspi; arttri/cargey; arttri/fesida; arttri/fessca; arttrp/fesida; arttsv/ ; arttsv/agrspi; arttsv/fesida; arttsv/fessca; arttsv/gervis; arttsw/agrsmi; arttsw/agrspi
potfru: potfru/ ; potfru/descae; potfru/fesida; potfru/fessca
rhus: rhutri/agrspi
saltshrub: sarver/
dryshrub: cerled/ ; cerled/agrspi; cerled/celret; chrnau/ ; chrnau/agrspi; junhor/andsco; purtri/ ; purtri/agrspi; purtri/fessca; rhutri/fesida
messhrub: celcocc/ ; celocc/agrspi; celocc/rhugal; celret/agrspi; celtis/ ; phymal/acegla; phymal/amealn; phymal/osmocc; rhugla/ ; rhugla/agrspi; symalb/ ; symalb/agrspi; symalb/balsag; symalb/fesida; symalb/galapa; symocc/
ripdecid: popang/corsto; poptri/ ; poptri/corsto; poptri/poapra
alpine: casmer/fesovi; dryoct/ ; kalmic/
rock/barren:
water:
urban:
no information:
Habitat types from 3 sources: Cooper, Stephen V., Kenneth E. Neiman, and David W. Rev. 1991. Forest habitat types of northern Idaho: a second approximation. General Technical Report INT-236. Ogden, UT: U.S. Department of Agriculture, Forest Service, Intermountain Research Station. 143p.
Pfister, Robert D., Bernard L. Kovalchik, Stephen F. Arno, and Richard C. Presby. 1977. Forest habitat types of Montana. USDA Forest Service General Technical Report INT-34. Ogden, UT: U.S. Department of Agriculture, Forest Service, Intermountain Research Station. 174p.
Mueggler, Walter F. and William L. Stewart. 1980. Grassland and shrubland habitat types of western Montana. USDA Forest Service General Technical Report INT-66. Ogden, UT: U.S. Department of Agriculture, Forest Service, Intermountain Research Station. 154p.