Summary Report Landscape Vegetation Inventory 2014 West Williams Lake TSA Forest Analysis and Inventory Branch Ministry of Forests, Lands and Natural Resource Operations Revised May 5, 2014 Executive Summary1 Landscape Vegetation Inventory (LVI) was initiated in 2012 and completed in 2014 for the area of 1,753,913 ha in the west Williams Lake TSA. The LVI database provides low cost, current and interim spatial and attribute inventory products that could be used to support high level strategic planning and resource analysis applications in forest resource management decision-making. Inventory estimates from the Landscape Vegetation Inventory (LVI) were compared to estimates from ground samples in the Williams Lake TSA. The LVI estimates were derived through photo interpretation of 7,398 segments (area total: 38,564 ha, a sampling intensity of approximately 3%) based on the 2012 digital aerial photography. Ground estimates from two sources – 31 CMI/NFI plots and 84, 9-plot clusters – were combined to provide an independent ground sample. For a key inventory attribute, live net merchantable volume, the LVI estimate was 39.0 m3/ha compared to the ground estimate of 39.4 m3/ha, not a statistically significant difference. When the samples were stratified by BEC, the ground sample sizes within each stratum were lower, the differences between LVI and ground estimates were higher and the variability associated with the estimates also increased. The largest BEC zone, the SBPS, showed some evidence of overestimation of volume by LVI that was balanced by underestimation in the remaining BEC zones (IDF and MS zones). When stratified by leading species group, the differences between the LVI and ground estimates were quite variable although none of the differences were statistically significant. When stratified by age class, again, the differences were highly variable and none of the differences were statistically significant. The biggest differences were associated with the 120+ age class. Dead trees were significantly under estimated by the photo interpretation process by about 43%. Over the entire project area the total live net volume was estimated at 44.9 million m3 (standard error: +0.5 millions m3) and the total dead net volume was estimated at 50.7 million m3 (standard error: +0.6 millions m3). It is recommended that for future LVI projects; (1) the photo interpretation be improved for accuracy and consistency; (2) a better pre-stratification of Landsat segments is adopted; and (3) improvements be made to the kNN and classification. LVI is an alternative to the conventional VRI and it is recommended that it be implemented for the area where VRI is not justified and current strategic inventory is needed. 1. Contact Xiaoping Yuan of FAIB for more information, phone: 250-953-3626, e-mail: Xiaoping.Yuan@gov.bc.ca Page i Table of Contents EXECUTIVE SUMMARY1 ........................................................................................................................................... I TABLE OF CONTENTS .............................................................................................................................................. II 1. INTRODUCTION ............................................................................................................................................ 1 2. 2.1 2.2 2.3 BACKGROUND .............................................................................................................................................. 1 PROJECT AREA ..................................................................................................................................................... 1 VEGETATION RESOURCES INVENTORY 2012 .............................................................................................................. 2 PROJECT INPUT DATA ........................................................................................................................................... 4 3. 3.1 3.2 3.3 3.4 3.5 3.6 3.7 LVI METHODOLOGY ...................................................................................................................................... 5 LANDSAT SEGMENTATION AND CLASSIFICATION ......................................................................................................... 5 PHOTO SAMPLING ................................................................................................................................................ 5 THE K-NEAREST NEIGHBOR (KNN) AND FUZZY C-MEANS CLASSIFICATION ...................................................................... 5 CLASS ATTRIBUTION ............................................................................................................................................. 6 GROUND SAMPLING ............................................................................................................................................. 7 DEPLETION UPDATE .............................................................................................................................................. 8 QUALITY ASSURANCE ............................................................................................................................................ 8 4. 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 RESULTS ........................................................................................................................................................ 8 CLASSIFICATIONS AND KNN .................................................................................................................................... 8 LVI SPATIAL DATA................................................................................................................................................ 8 LVI TREE ATTRIBUTES DATA ................................................................................................................................... 9 OVERALL LIVE AND DEAD NET VOLUMES .................................................................................................................. 9 FINAL LVI DATA AREA SUMMARY ........................................................................................................................... 9 GROUND SAMPLING ASSESSMENT ......................................................................................................................... 10 GROUND ADJUSTMENT OF LIVE TREE BASAL AREA AND DEAD NET VOLUME ................................................................. 12 COST SUMMARY ................................................................................................................................................ 13 5. SUMMARY .................................................................................................................................................. 14 6. REFERENCES ................................................................................................................................................ 15 7. 7.1 7.2 7.3 APPENDIX ................................................................................................................................................... 16 GENERALIZED SITE SERIES CLASSES BY BEC SUBZONE ................................................................................................ 17 WILLIAMS LAKE LVI CLASSIFICATION AND DATABASES ............................................................................................... 20 LVI DATA DICTIONARY ........................................................................................................................................ 24 LVI Summary Report June 10, 2013 Page ii 1. Introduction The Landscape Vegetation Inventory (LVI) has been developed as a new, low cost, flexible, and statistically sound alternative inventory tool to provide spatial and attribute products at various spatial resolutions. The primary targets for the LVI are the areas of the province where a traditional photo interpretation inventory such as Vegetation Resources Inventory (VRI) is not justifiable or not cost effective, areas with less forest complexity, regions outside of the timber harvesting land base, parks, or areas heavily impacted by mountain pine beetle. The LVI data is appropriate for strategic planning and reporting, as opposed to operational purposes. The LVI design includes three basic components: (1) Landsat multispectral imagery for segmentation and classification; (2) low-level Digital Camera System (DCS) image sampling and interpretation; and (3) nearest neighbor classification for extrapolation to provide broad spatial and attribute products. Other auxiliary data layers: such as the existing Vegetation Resources Inventory (VRI), Biogeoclimatic Ecosystem Classification (BEC), Terrain Resource Inventory Management (TRIM), Predicted Ecosystem Mapping (PEM) and Site Productivity (SP), are also used in the classification. Forest Analysis and Inventory Branch (FAIB) designed and pilot tested the LVI in 2010, and since then has conducted and implemented two LVI operational projects, one in Quesnel Timber Supply Area (TSA) in 2011/2012 and the other in Williams Lake TSA in 2012-2014. This report presents a final summary of the LVI project for Williams Lake TSA. 2. Background 2.1 Project Area The area of interest is the western part (Chilcotin Forest District) of the Williams Lake TSA. The total project area is 1,753,913ha, covering 113 full 1:20,000 NTS maps (Figure 1). The area was first stratified into treed and not treed areas by stand age greater than 40 years old using the existing VRI. The treed area (1,152,631ha) was further stratified by British Columbia Beogeoclimatic Ecosystem Classification (BEC) [Meidinger and Pojar 1991]. Figure 1. The location of the project area in the Williams Lake TSA. The area breakdowns into 3 BEC zones: Sub-Boreal Pine Spruce (SBPS), Montane Spruce (MS), and Interior Douglas-Fir (IDF). The biggest stratum, SBPS accounted for more than half (57%) of total treed area, followed by MS (26%) and IDF 1. Contact Xiaoping Yuan of FAIB for more information, phone: 250-953-3626, e-mail: Xiaoping.Yuan@gov.bc.ca Page 1 (14%). There was a small area of Engelmann Spruce Subalpine Fir (ESSF) accounted for 3% in the area of interest, and for simplicity it was to lumped into the MS stratum. Area Distribution by BEC Zones 700 Area (x 1000ha) 600 500 400 300 200 100 0 IDF MS SBPS ESSF BEC ZONE Figure 2. Area distributions by BEC zones in the treed stratum of the LVI project area. 2.2 Vegetation Resources Inventory 2012 The majority of the inventory data available in the current VRI database (2012) was created during 1980-2000 using the old FIP/FC1 standards. Figure 3 provides an inventory vintage by the project area (Figure 3) . Area (1000 ha) Total Area by Reference Year 800 700 600 500 400 300 200 100 0 NA Prior 1980 1980-1989 1990-1999 2000-2011 Reference Year Figure 3. Total areas of forest inventory by reference year (NA: Not Available). Predominant tree species was lodgepole pine (PL: Pinus contorta Dougl.), followed by Spruce (SX: Picea A. Dietr.) , Douglas Fir (FD: Pseudotsuga menziesii (Mirb.) Franco), and several deciduous species (A: Aspen, Birch etc.). A very small amount of Balsam Fir (B: Abies balsamea (L.) Mill.) was lumped into the SX stratum (Figure 4). LVI Summary Report June 10, 2013 Page 2 Area Distribution by Leading Species 1,200 Area (1000 ha) 1,000 800 600 400 200 0 NA A B FD PL SX Leading Species Figure 4. Area distribution by leading species (NA: no species data available). Figure 5 provides an area distribution by age classes (<40 years old, 41-60, 61-80, 81-100, 101-120, 121-140, 141-160 and > 160). 350 Area Distribution by Stand Age 300 250 Area (x 1000 ha) 200 150 100 50 0 1-40 41-60 61-80 81-100 101-120 121-140 141-160 >160 Age Class Figure 5. Area distribution by stand age class. An epidemic outbreak of Mountain Pine Beetle (MPB) occurred throughout the project area during 2000-2010 peaking in 2005, and by 2012 it had killed 60% of the mature Lodgepole Pine volume in the Williams Lake TSA according MFLNR MPB LVI Summary Report June 10, 2013 Page 3 survey. FAIB has made an annual calculation of live and dead stand volumes based on the MPB survey and the total live and dead volumes by BEC zones for 2012 are shown in Figure 6. 35 Total Net LIve and Dead Volumes 30 Vol;ume (Millions m3) 25 20 15 10 5 0 ESSF IDF MS SBPS BEC Zone Live Volume Dead Volume Figure 6. Total live and dead volumes (source: the 2012 VRI). This MPB infestation followed by intensive salvage logging and burns, has made the existing inventory obsolete, particularly in pine dominant area. However, in some areas, dominated by non-pine species, the VRI may still be valid and useful. Since a new VRI was not justified right after MPB infestation due to rapid and dynamic changes in species composition and stand structure, an interim and current inventory product such as the LVI was required particularly for high level strategic planning and analysis such as TSR. 2.3 Project Input Data The following datasets were prepared and assembled for the project: (1) The primary project input data was 30m Landsat 5 Thematic Mapper (TM) imagery. Three scenes (48/24, September 8, 2009, 49/23, August 30, 2009, and 49/24, August 30, 2009) were selected for the project area. In addition to raw band spectral data, various spectral indices, principal components, ratios, normalized difference vegetation index, etc. were created. (2) The existing Vegetation Resources Inventory (VRI) 2012 data was extracted from the VRI database and served both for the initial stratification as well as an ancillary data layer in the LVI process. (3) The Predictive Ecosystem Mapping (PEM) was completed for the Cariboo region in 2007 and provided a seamless, 25 m grid digital ecological and site specific data for the vegetation/trees in the region [Moon et al 2008]. For the purpose of the LVI project, the site series values were grouped into several classes based on the similarity and closeness of the sites in terms soil and nutrient regimes in each of BEC subzones (Appendix 6.1). LVI Summary Report June 10, 2013 Page 4 (4) Site Productivity (SP) data was created by Ministry of Forests, Lands and Natural Resources Operations (FLNRO) in 2013. It was intended to provide site index estimate province-wide for commercial tree species for strategic purposes. It is a raster based 100m by 100m grid cell data [FLNRO 2013]. (5) Terrain Resource Inventory Management (TRIM) was used for extracting elevation, slope, and aspect data. (6) Disturbance data was created from several sources and it was used for updating the final LVI data. The first source of disturbances was from the consolidated disturbance layer created by FAIB in 2013, which included all harvested change captured in VRI, RESULTS and Landsat change detection mapping. The second source of disturbance data was the forest fire layer created by the Wildfire Management Branch of FLNRO. All the disturbances from the sources above were extracted if they occurred after 2000. For the year of 2013, cut and fire disturbances were created from Landsat 8 2014 change detection. 3. LVI Methodology 3.1 Landsat Segmentation and Classification The Landsat 5 TM scenes were segmented into small (1-5ha) spectrally homogenous segments using object based software Definies eCognition for the project area, resulting in a total of 340,057 segments. These segments were further stratified into treed and no treed based on Landsat classification and other auxiliary data such as the existing VRI and PEM. The final number of treed segments was 250,221 after masking out cuts, burns, age younger than 40 years, water, swamps, clouds/shadows etc. 3.2 Photo Sampling A systematic strip sampling, 20km spacing in south-north direction, was selected for the sampling design, and digital color photography with an average of 0.15m ground sampling distance (GSD) was acquired using the LEICA ADS80 line camera on August 5/6, 2012. The corridor of the strips covered with the photographs was about 1500m in width. The treed segments within 1km radius at each 5km grid point along each strip were selected, resulting in a total 7,926 segments (called the “photo sample” hereafter). Each photo sample segment was interpreted manually by certified VRI photo interpreters using the color/stereo photography and full tree stand attributes were estimated as per VRI standards. There were 28 segments excluded from the photo sample as they were either no tree segments or were logged or burned after the photography was taken. The total number of the final photo sample was 7,398 (area total: 36,276 ha, about 3% of the total treed area). The final attributes for each segment after photo interpretation were then projected to the year 2013 using the VDYP7. A dead tree layer was created for standing dead trees. For the purpose of this project, dead trees were photo interpreted with tree species as 100% Lodgepole Pine, basal area, crown closure, and stems per ha and then these attributes were compiled as live trees using VDYP7 for dead volume (net and gross). 3.3 The k-Nearest Neighbor (kNN) and Fuzzy C-Means Classification While the photo sample provided statistical estimates of the variables of interest for the population, a spatial map (i.e. spatial distribution of the variables of interest) was also required. The LVI process began with two sources of data: (1) wallto-wall spectral homogenous segments with various spectral measures (as well as other measures from the auxiliary data such as VRI and PEM) which we call the "target dataset", and (2) a subset of these segments that are photo interpreted with full tree attributes to the VRI standards, we call the "reference dataset". The objective of LVI was to estimate the VRI attributes for the target dataset by the way of "matching" (or classifying) them to most similar segments in the reference dataset. One way to achieve this was through the k-Nearest Neighbor (kNN) method. kNN is basically a classification process in which the missing variables in the target dataset are referred to as the Y-variables and the variables available commonly in both the target and reference datasets are referenced to as the X-variables. For each observation in the target dataset, one or more (k) nearest neighbors are found in the reference dataset based on the magnitudes of differences in the X-variables. Once the minimum difference between a target observation and a given reference observation (summed up across all X-variables in relation to each of the remaining observations in the reference dataset) is found, this particular reference observation is identified as the nearest neighbor to the target observation. This process may be extended to identify the next nearest neighbor, or to the k nearest neighbors as defined by the user. The LVI Summary Report June 10, 2013 Page 5 missing Y-variables could be assigned by the nearest neighbor (i.e. k=1) or estimated as the mean values of the k nearest neighbors. The photo sample data was classified by tree species composition, stand structure (crown closure, basal area, trees per ha, etc.), and dead tree attributes, using fuzzy c-means classification developed by Moss (2013). Multiple Discriminant Analysis was used for X-variable selection and determining weights by maximize the ratio of the between to within class sum of squares by applying weights to the X-variables dataset. These discriminant functions were then applied to the target dataset to identify nearest neighbors. For version 1 of the LVI data that was completed in December 2013, the Y-variables were estimated based on k=1, i.e. the nearest neighbor method. Various classification schemes and kNN imputations were tested and evaluated (Moss 2014a). For example, instead of selecting the 1st neighbor, averages of the first 2, 3, 4, and 5 nearest neighbors were also generated for tree attributes. Instead of applying a “global” classification, separate classifications and kNN assignments were done by BEC zones; or global classifications but applying to each BEC zone separately. A post classification was developed based on spectral properties, BEC subzones, PEM classification, and spatial distance between the target segments to reference segments. This post classification was applied to the 50 nearest neighbors identified from the kNN process. 3.4 Class Attribution Once the final classification of the Landsat segments was determined, polygon dissolving was applied to all the segments by the class. All the matched segments in each class were grouped together. For the numerical variables such as basal area, volume, age, height etc., a simple mean value of all the photo sample segments was calculated weighted by segment area. The standard deviations of classes were stored in a separate table which could be examined later to evaluate the precision of the final attribute estimates. Species composition for a given class was derived based on the following formula: π π ππ = 100 ∗ ∑π π=1(πππ ∗ π΅π ∗ π΄π) /[∑π=1 ∑π=1 (πππ ∗ π΅π ∗ π΄π)] (1) Where: m: number of photo segments in the class n: number of all possible tree species (n = 4, i.e. the 4 highest percentage proportions) ππ : percentage of species i πππ : percentage of species i in the jth photo segment π΅π: basal area in the jth photo segment π΄π: area of the jth photo segment The resultant percentage for a given species for a given class calculated from equation (1) is weighted by basal area. The final species composition is calculated only for the 4 tree species (PL, SX, FD, AT) that occurred commonly in the project area as per VRI species composition standards. Because age, height, and site index are related to the species, they are computed based on the following formulas: π AGEi = ∑π π=1(π΄πΊπΈππ ∗ π΄π) / ∑π=1 (π΄π) (2) π HTi = ∑π π=1(π»πππ ∗ π΄π) / ∑π=1 (π΄π) (3) π SIi = ∑π π=1(ππΌππ ∗ π΄π) / ∑π=1 (π΄π) (4) AGEi : Age for species i HTi : Height for species i SIi : Site Index for species i AGEij : Age for species i in the jth photo segment LVI Summary Report June 10, 2013 Page 6 HTij : Height for species i in the jth photo segment SIij : Site Index for species i in the jth photo segment Depending on the leading and the second leading species determined by equation 1, the age, height, and site index computed from equations 2-4 were assigned to the corresponding species. 3.5 Ground Sampling A separate ground sampling component of this project was designed and carried out to: 1) 2) 3) 4) To assess the LVI estimates over the population area; To assess LVI photo interpretation; To evaluate the overall final LVI product; and To adjust/improve the LVI estimates where appropriate. The approach used to select ground plots locations was broken into the following three sequential steps: 1) Select as many samples on the “Treed” photo plots as possible. [Result=67] 2) Next, select as many “Treed” samples as possible on the 10km X 10km National Forest Inventory (NFI) grid and stratify these points by BEC zone to only include SBPS and MS zones. [Result=24] 3) Lastly, select as many “Treed” samples as possible on the 5km X 5km NFI grid and stratify these points by BEC zone to only include SBPS and MS zones. Randomly select a number of these locations to insure we have the intended number of ground sample points. {Result=18] This process identified 109 candidate locations to measure ground sample data for the Williams Lake LVI project. At each candidate location, a 9-points cluster, each point is a fixed radius 5.64m radius plot with a nested 3.99m plot at 50m spacing (3 x 3 over a 100m by 100m square), was established. All the live/dead trees greater than 4cm within the 5.64m plot were measured using a modified timber cruising method. For small trees less than 15cm in height and 4cm in diameter, a quick estimate of number of trees was conducted. This 9-plot cluster dataset was used for population estimation as well was broken down by number of plots falling in corresponding segments/polygons for evaluation of LVI products. In addition, there was another ground sampling scheme adopted from two other projects, Change Monitoring Inventory (CMI) and NFI. These CMI/NFI plots (fixed radius 0.04ha plots at the NFI 20km grid points) falling into the LVI treed area were combined with the 9-plot clusters (see Penner 2014). Both the 9-plot clusters and CMI/NFI plots were compiled at FAIB using non standard procedures in regards to the taper equation and the accounting of decay, waste and breakage. The CMI samples were compiled by the RDW compiler. The LVI samples were compiled by a standalone compiler customer built specifically for the task. These datasets went through the following processing steps: ο· A GIS query to determine the status of each plot of the 9 plot LVI cluster was done to determine the placement in the LVI population and if so, LVI polygon. ο· The gross volume was calculated using the 1994 taper equation. ο· The DWB deductions were calculated using the BEC loss factors. The risk grouping of dead trees used the live tree criteria. ο· The blowup factor used to calculate per hectare values was based on plot radius: 11.28 m for the CMI samples and 5.64 m for the 9-plot samples. ο· Population boundaries were not recognized for both plots, i.e. the entire plot was established regardless of dissection by a population boundary. LVI plots were split into halves or quarters depending on excessive tree counts. LVI Summary Report June 10, 2013 Page 7 3.6 ο· Heights were measured for all trees, which allowed for the calculation of the average heights for all tallied trees and subsets of trees by crown classes. ο· In the 9-plot clusters, tree age was captured only for the two major species that were considered representative of the plot. The leading species age was used as the basis for the loss factor calculation for dead trees. ο· For CMI, tree age was captured for the site trees. Loss factor age is a variable named age_dwb and may be obtained from the leading species. ο· The compilation calculated per hectare values for each dead and live tree species separately for the 9 plot sample and for the LVI polygon. Summaries were also produced by live vs dead for each sample and polygon. Depletion Update The LVI data, once completed, was updated for depletion changes caused by harvesting and wild fire. The FAIB consolidated disturbance layer 2013 was used first to update all the cut blocks occurred between 2000 and 2012. Wild fire data was from Wild Fire Management Branch and all the fires occurred in the same time period 2000-2012 were extracted and used to update the LVI. For the final step, all the cut/fire changes occurred in 2012 to 2014 were automatically captured from the Landsat 8 TM imagery acquired on September 10, 2013, and were used for updating the LVI. 3.7 Quality Assurance Three critical components were subjected to quality assurance (QA): photo interpretation, nearest neighbour matching and classification. About 3% of photo interpreted segments were selected for QA by another certified VRI photo interpreter. The nearest neighbour matching was checked visually as compared to the orthophoto and Landsat. Statistical estimates from the photo sample data were used to check the total summation of matched segments and classified polygons. 4. Results 4.1 Classifications and kNN A total of seven classifications were tested to the photo sample data and a detailed summary report is provided by Moss (2014a, 2014b). One of them was deemed as the best and was selected for the final processing for generating nearest neighbors. This classification, based on combined species, structure, and dead trees, was created from the entire photo data set; subsequently 50 nearest neighbors was determined separately for each of the 3 BEC zones; and then a final post classification was applied to determine the best nearest neighbor for a given unknown target segment. 4.2 LVI Spatial Data After the best nearest neighbor was determined and photo sample data was populated to the unknown segments, a final grouping process was applied to produce a generalized spatial coverage. The argument for doing this is that LVI process is affected by uncertainties originated from photo interpretation and the kNN process. Generalizing from small segments to coarser resolution will reduce these uncertainties, while still providing a reliable inventory data product for strategic and high level planning and analysis applications. Six generalizations were produced as described in detail in Appendix 6.2. The first 4 (WL_CLASS1 –WL_CLASS4) were provided for the Williams Lake District Office according to the leading species and live/dead tree ratios. The fifth (WL_CLASS5) was created based on the best classification/kNN process, i.e. fuzzy C-means classification of the entire photo sample data and kNN was applied separately by BEC zones with a post classification. The last generalization (LANDSAT_POLY) was created from Landsat imagery. Segments were combined into coarse resolution big polygons based on the similarity of spectral properties. This is the dataset used to replace the old VRI in the VRI database in order to be consistent with the structure, format, and resolution of the VRI database. LVI Summary Report June 10, 2013 Page 8 4.3 LVI Tree Attributes Data The LVI products consist of three basic databases (both with spatial and attribute data): (1) the segment database; (2) the classification database; and (3) the photo sample database. The segment database is the primary LVI product for FAIB internal, and other, power users who understand the LVI process and have the ability to re-process it into other relevant products. The classification database is the inventory product for general public access and use. The photo sample database is a snapshot and map-less inventory of the population of interest and it can be recompiled and re-processed for statistical analysis and estimations. The data dictionaries of the above three kinds of LVI attributes in the LVI databases are given in Appendix 7.3. 4.4 Overall Live and Dead Net Volumes The overall estimates of total live net volume and dead gross volume in the LVI tree area are summarized in Table 4. Table 1. The average and total live and dead net volumes at 12.5cm estimated from the photo sample (means or total followed by standard errors). Average Volume (m3/ha) Live Dead1 39.0 + 0.4 1. 4.5 44.0 + 0.5 Total Volume (millions m3) Live Dead1 44.9 + 0.5 Photo Sample 50.7 + 0.6 36,276 Area (ha) Tree Population 1,152,631 1,753,913 Dead volume estimates derived from the photo sample was ratio adjusted using the ground sampling data (see 4.7). Final LVI Data Area Summary Table 4 provides an area summary of the final LVI data. Out of a total area of 1.754 million ha, new LVI data (with Inventory Standard as “L”) accounts for 1.419 million ha, with a breakdowns by the tree area of 1.06 million ha, new cuts of 0.039 million ha and new fires of 0.119 million ha occurred between 2000 and 2014, and the no tree (opens, roads, swamps, water etc.) of 0.201 million ha. The remaining 0.335 million ha was filled in with the VRI 2012. Table 2. The area summary by inventory standard code. Inventory Standard Total Area (x 1000 ha) LVI generated (L) Tree 1,419 VRI fill in (V, I, F) 335 Project total area LVI Summary Report June 10, 2013 1,060 Cut 39 Fire 119 No tree 201 Cloud/shadow 10 Young stand 34 Other 291 1,754 Page 9 4.6 Ground Sampling Assessment A total of 84 9-plots clusters and 31 CMI/NFI plots were resulted from the ground sampling. The detail analysis of ground sampling data in comparison with the LVI photo estimates is given by Penner (2014). The statistical estimates of some key attributes from the LVI photo sample, ground sample (pooled together the 9-plots clusters and CMI/NFI plots), and VRI extracted segments are summarized in Table 4 and Figure 7. Live net volume is a very important attribute and the estimates from all three sources are very close. Age and height from the LVI and ground are very close, while they are much higher from VRI. Basal area and stand density (trees per ha) from the LVI are higher than those from the ground. The differences are attributed to some differences in utilization levels between the data sources and basal area and trees per hectare are particularly sensitive to utilization. Dead tree volume was significantly under estimated by photo interpreters by 43%, primarily due to underestimated basal area and trees per ha. Table 3. The averages of some selected attributes are given by sources (the mean followed by the standard error). 9 plot Ground1 Attribute LVI CMI clusters Sample size 38,5643 ha 31 plots 84 plots 115 plots Live net volume (m3/ha) 39.0 ± 0.4 47.7 ± 10.2 38.0 ± 4.2 39.4 ± 5.1 Live whole stem volume (m3/ha) 50.6 ± 0.6 62.1 ± 12.8 49.2 ± 5.2 51.1 ± 6.3 Live whole stem volume (m3/ha) Dbh > 4.0 cm 62.9 ± 0.6 93.0 ± 14.5 79.9 ± 6.2 81.8 ± 7.4 Dead net volume (m3/ha) 24.8 ± 0.3 20.0 ± 4.3 47.8 ± 4.9 43.8 ± 4.8 Leading species age (years) 104.4 ± 0.5 102.8 ± 10.4 89.0 ± 3.1 91.0 ± 4.2 Basal area (m2/ha) 12.9 ± 0.1 9.4 ± 1.8 7.5 ± 0.7 7.8 ± 0.9 Leading species Height (m) 12.3 ± 0.1 13.3 ± 0.7 12.2 ± 0.3 12.4 ± 0.4 Live trees per hectare 940.7 ± 12.9 419.6 ± 75.4 304.2 ± 26.3 320.8 ± 33.4 Dead trees per hectare 220.3 ± 2.2 206.2 ± 41.4 325.9 ± 30.1 308.6 ± 31.7 1. Combined CMI/NFI and 9 plot cluster datasets with the weights given by Penner (2014) 2. The existing VRI database 2012 3. A total of 7398 photo segments with a total area of 38,564ha LVI Summary Report June 10, 2013 Page 10 VRI2 36,295 ha 40.3 ± 0.5 43.7 ± 0.5 124.2 ± 0.7 19.1 ± 0.1 14.8 ± 0.1 524.2 ± 7.4 551.9 ± 5.5 140 120 100 80 LVI 60 40 CMI 20 Nine plot cluster VRI 0 Net volume (m3/ha) Whole stem Whole stem Dead net volume volume (m3/ha) volume Dbh > 4.0 (m3/ha) cm (m3/ha) Attribute Age (years) 25 20 15 LVI 10 CMI Nine plot cluster 5 VRI 0 BA (m2/ha) Ht (m) TPH (x 100 stems/ha) Dead TPH (x 100 stems/ha) Attribute Figure 7. The data in Table 3 are plotted. Standard error bars are given. More than half of the area is in the SBPS BEC zone. The MS and SBPS are both dominated by pine-leading polygons (> 70% of the area) followed by spruce-leading. The IDF is dominated by Douglas-fir leading polygons (50% of the area). Overall, the LVI and ground volumes were very close (Table 4 and Figure 8). The overestimation of net live volume by the LVI compared to the ground plots in the SBPS zone is compensated by the overestimation in the other zone so the overall difference is -0.5 ± 4.5 m3/ha, not a statistically significant difference (α < 0.05). Note the increase in standard error associated with the volume/ha at the stratum level compared to the population, due to the smaller sample size. Table 4. The average net live volume is summarized by BEC and source. The mean is followed by the standard error. The data are graphed in Figure 8. BEC ESSF & MS IDF CMI 9-plot N N 12 3 Net live volume (m3/ha) LVI 26 46.1 ± 0.8 Ground Difference Net live volume (millions m3) Area LVI (ha) Ground Difference 60.7 ± 8.4 -14.6 ± 8.4 14.1 ± 0.3 22.5 ± 3.1 -4.5 ± 2.6 330,682 13 39.7 ± 1.1 43.3 ± 10.7 -3.7 ± 10.8 7.8 ± 0.2 7.3 ± 1.8 -0.7 ± 2.1 165,260 SBPS 16 45 35.4 ± 0.6 25.5 ± 3.5 9.9 ± 3.6 23.0 ± 0.4 15.7 ± 2.2 6.4 ± 2.3 656,690 All 31 84 39.0 ± 0.4 39.4 ± 3.9 -0.4 ± 3.9 44.9 ± 0.5 45.4 ± 4.5 -0.5 ± 4.5 1,152,631 LVI Summary Report June 10, 2013 Page 11 80 Live net volume (m3/ha) 70 a) 60 50 40 LVI 30 Ground 20 10 0 ESSF & MS IDF SBPS Total BEC zone Live net volume (m3) 60,000,000 b) 50,000,000 40,000,000 30,000,000 LVI 20,000,000 Ground 10,000,000 0 ESSF & MS IDF SBPS Total BEC zone 10,000,000 Live net volume (m3) 8,000,000 c) Difference (LVI - Ground) 6,000,000 4,000,000 2,000,000 0 -2,000,000 ESSF & MS IDF SBPS Total -4,000,000 -6,000,000 -8,000,000 BEC zone Figure 8. The average (a), total (b) and difference (c) between net live volumes are given by BEC zone and source. Standard error bars are given 4.7 Ground Adjustment of Live Tree Basal Area and Dead Net Volume As shown in Table 4 and Figure 8, although the overall live net volume estimate from the LVI agrees with that from the ground data, there were some differences in BEC zones, under estimate in ESSF/MS and over estimate in SBPS. A further analysis showed that basal area could reasonably adjusted in different BEC zones while keeping the overall population live LVI Summary Report June 10, 2013 Page 12 volume unchanged (Moss, 2014c). Accordingly, the decision was made to adjust the photo interpreted basal area as follows: ο· No adjustment in IDF and ESSF, and for SX/FD/AT leading stands ο· Multiply ratios of 0.5607 applied to SBPS and 1.30 to MS to the photo basal area, to the stands with PL leading, 1 < photo basal area > 30 The ground assessment as mentioned in 4.6 clearly identified under estimation of dead volume by LVI photo interpreters. Twenty-seven 9-plot clusters were selected (with more than 4 plots falling in the segment) to pair with segments as shown in Figure 9 and an adjustment curve was derived and applied to the photo interpreted dead net and gross volumes. Notmslized Dead Net Volume (Weighted 4plots+) 9 8 7 Ground 6 y = 1.8231x - 0.1081 R² = 0.5462 5 4 3 2 1 0 -1 0 0.5 1 1.5 2 2.5 3 3.5 Photo Figure 9. Normalized dead net volumes of ground 9-plot clusters (with more than 4 plots within the segment) vs the photo segments. 4.8 Cost Summary The total cost and breakdowns are summarized in table 5. Average cost per ha was $0.11. The cost for kNN analysis incorporated some research and testing components and as such, the operational cost could be much lower. Table 5. Project cost summary. Items Photo acquisition Cost Unit Cost (/1,753,913 ha) $4,102 $0.002 Photo interpretation $79,992 $0.046 kNN analysis $90,000 $0.051 Statistical analysis $25,000 $0.014 LVI Summary Report June 10, 2013 Page 13 Subtotal $199,094 $0.11 Ground sampling $180,885 $0.11 Total $379,979 $0.22 It should be pointed out that cost for internal staff working on Landsat segmentation, data preparation, data processing, etc. was not included in Table 5. If these items were contracted out to a typical operational consultant setup, it could cost about $18,000 (10 days for segmentation/classification and 20 days for data processing, at $600/day). So the unit cost, if the entire project was contract out, could be $0.12 or less. Total cost for the ground sampling is $180,885, or $0.11/ha. 5. Summary Landscape Vegetation Inventory (LVI) was initiated in 2012 and completed in 2014 for the area of 1,753,913 ha in west Williams Lake TSA. The LVI database provides low cost, current and interim spatial and attribute inventory products that could be used to support high level strategic planning and resource analysis applications in forest resource management decision-making. This project has demonstrated a design, execution, and implementation of a typical operational LVI project in the area where conventional VRI is not justified and current strategic inventory is needed. LVI could and should be considered as an alternative to VRI, and be implemented where appropriate. The LVI data is provided to potential users through the operational VRI database, however, more LVI products can be created and provided to specific users for specific requirements. Communications and consultations are encouraged to be made with FAIB as the LVI data is not exactly equivalent to the conventional VRI data that most people are familiar to. The concept of providing multi-resolution inventory data for various information needs is an important shift from traditional “one size fits all” idea. There will be varieties of inventory products ranging from higher resolution LiDAR inventory, operational inventory, to strategic VRI and LVI, and users should be aware of the differences so that inventory data is used appropriately. There are many lessons learned from this project and it is recommended that: (1). Consistency and accuracy on tree attribution must be improved in order to minimize measurement bias. For example, stand density (trees per ha) can be measured and basal area can be modeled instead of interpreting; (2). Attention should be paid to utilization, projection, and compilation of both photo interpretation and ground data collection so that consistent and relevant comparisons and adjustments can be made; and (3). There are many aspects in the LVI segmentation, classification and kNN that can and should be improved for future LVI projects. LVI is an alternative to the conventional VRI and it is recommended that it be implemented for the area where VRI is not justified and current strategic inventory is needed. LVI Summary Report June 10, 2013 Page 14 6. References Meidinger, D. and Pojar, J. 1991. Ecosystems of British Colombia. B.C. Ministry of Forests, Victoria, British Columbia. FLNRO. 2014. Williams Lake TSA Timber Supply Analysis Public Discussion Paper. Forest Analysis and Inventory Branch, Ministry of Forests, Lands and Natural Resources Operations, Victoria, British Columbia. Ministry of Forests, Lands and Natural Resource Operations. 2013. FLNR Provincial Site Productivity Layer. Version 4.0, FLNRO, March 31, 2013, Victoria, British Columbia. Moss, Ian. 2013. LVI Linear Discriminant Analyses. Prepared for Forest Analysis and Inventory Branch, BC Ministry of Forests, Lands and Natural Resources Operations, Victoria, British Columbia. Moss, Ian, 2014a. A Comparison of Alternative Landscape Vegetation Inventory (LVI) Nearest Neighbour Assignments, Prepared for Forest Analysis and Inventory Branch, BC Ministry of Forests, Lands and Natural Resources Operations, Victoria, British Columbia. Moss, Ian, 2014b. The Landscape Vegetation Inventory (LVI) Nearest Neighbour Analysis Process. Prepared for Forest Analysis and Inventory Branch, BC Ministry of Forests, Lands and Natural Resources Operations, Victoria, British Columbia. Moss, Ian, 2014c. Williams Lake LVI Basal Area Adjustment Ratios. Prepared for Forest Analysis and Inventory Branch, BC Ministry of Forests, Lands and Natural Resources Operations, Victoria, British Columbia. Penner, Margaret. 2014. Landscape Vegetation Inventory 2014 Williams Lake Sample Results. Prepared for Forest Analysis and Inventory Branch, BC Ministry of Forests, Lands and Natural Resources Operations, Victoria, British Columbia. LVI Summary Report June 10, 2013 Page 15 7. Appendix LVI Summary Report June 10, 2013 Page 16 7.1 Generalized Site Series Classes by BEC Subzone PEM/LVI Classification BEC Zone Subzone Site Series SBPS dc SBPS mc SBPS mk SBPS xc MS xv Modifier Nutrient/ Moisture Lead Species LVI_ID LVI_class Description 01 MM Pl 8101 101 Sub-Boreal Pine-Spruce . 02 MD Pl 8102 102 03 MM Pl 8103 102 04 PM Pl 8104 101 05 MM Sx 8105 101 Dry cold, Elev 1000-1280 South and 900-1225 North. Small portion at north-east corner. 01 and 04 most common, some 03 05 08, and little 02 06 07. Could combine 02 03, 01 04 05 and 06 07 08 06 RW Sx 8106 103 07 PW Sb 8107 103 08 MW Sx 8108 103 01 MM Pl 8201 201 Sub-Boreal Pine-Spruce . 02 MD Pl 8202 202 Moist cold, Elve 900-1250 03 MM Sb 8203 202 04 RW Sx 8204 202 Only small portion at north-west corner. Most common 01 and very little 02-07 05 RW Sx 8205 202 06 RW Sx 8206 202 07 RW Sb 8207 202 01 MM Pl 8301 301 Sub-Boreal Pine-Spruce . 02 PD Pl 8302 302 03 MD Fd 8303 302 04 MM Pl 8304 303 05 MM Sx Fd 8305 303 Moist cool, Elv 1000-1350 South, 9501250 North. Small portion at the north-east corner. Most common 01, some 04 06 07 and little 02 03 05 08. Could combine 02 03, 04 05, and 06 07 08. 06 RW Sx 8306 304 07 RW Sx 8307 304 08 PW Sx 8308 304 01 LK MD Pl 8401 401 Sub-Boreal Pine-Spruce . 02 LC PD Pl 8402 402 03 SB MW Sx 8403 403 Very dry and cold, Elev 1100-1500 South and 850-1300 North. 03 and 04 could be combined 04 SF MW Sx 8404 403 05 SH MW Sx 8405 404 06 SM RW Sx 8406 405 01 MM Pl 8001 801 Montane Spruce. 02 PD Pl 8002 802 Very dry very cold. 1400-1700 in the LVI Summary Report June 10, 2013 Page 17 IDF dk(4) IDF xm ESSF xv (1) Not-treed 03 MD Pl 8003 803 04 MM Pl 8004 801 05 MM Pl 8005 801 06 RW Sx 8006 804 07 PW Sx 8007 805 08 RW Sx 8008 804 09 PW Sx 8009 805 01 MM FdPl 6601 601 02 MD Fd 6602 602 03 PD Fd 6603 602 04 RD Fd 6604 602 05 RD Fd 6605 603 06 PM Pl 6606 603 07 MM Fd 6607 603 08 PW Sx 6608 601 09 RW Sx 6609 601 10 RW Sx 6610 604 01 MM Fd 6901 901 02 MD Fd 6902 902 03 PD Fd 6903 903 04 MD Fd 6904 902 05 MM Fd 6905 904 06 MM Fd 6906 905 07 RM Fd 6907 905 08 RW Sx 6908 905 09 RW Sx 6909 906 01 MM Bl 5501 501 02 MD Bl 5502 502 03 PD Pl 5503 503 04 MM Bl 5504 502 05 MM Bl 5505 502 06 PM Bl 5506 501 07 MW Bl 5507 504 08 MW Bl 5508 504 09 RW Bl 5509 504 100 10 00 OW LVI Summary Report June 10, 2013 south and 1250-1500 in the north. Most common 01 06 08 09, some 04 05 and little 02 03 07. Could combine 01 04 05, 06 08, and 07 09 Interior Douglas-Fir dry cool Chilcotin. Elev 1050-1350. Most common 01 and 09. Very little 02 03 04 and 07. Some 05 06 and 10. Could combine 01 08 09, 02 03 04, and 05 06 07. Very little dk(3) Interior Douglas-Fir very dry mild. Elev. 800-1200. Most common 01 and some 06. Very little others. Could combine 06 07 08, 02 04 Engelmann Spruce-Subalpine Fir very dry very cold. Elev. 1650-2100. Most common 01 and some 06 07 08 and little for the rest. Could combine 02 04 05, 01 06, and 07 08 09. WA (Water) Page 18 (include Alpine Tundra ATun and ESSF xvp) 00 WE 101 11 WL (Wetland) 00 BR DG FB FM FC GL HT ME PA RO SF 102 12 VNT (Vegetated and no treed) 00 AF AV CS GB RO RU TA 103 13 NV (Non-vegetated) 00 DL 104 14 Disturbed (e.g. road) Nutrient: P (poor), M (medium), R (rich) Moisture: D (dry), M (medium), W (wet) LVI Summary Report June 10, 2013 Page 19 7.2 Williams Lake LVI Classification and Databases Williams Lake LVI Classification and Databases The following 6 classification schemes were used for the LVI data for Williams Lake TSA west. The first 4 classification schemes were defined by the staff in the Williams Lake district office, while the schemes 5 and 6 were defined by FAIB. The attribute tables for each of the first 5 classifications and data dictionary are 7.3. Classification 6 is defined by homogeneous spectral polygons created from Landsat 8 imagery. All the classification maps (with the attributes) above are included in the ArcGIS personal geodatabase. 1. Age Williams Lake Classification 1 (WL_CLASS1) Lead species % Net Volume (m3/ha) <60 WL_CLASS1 Comment 1 Young stands >=60 PL >=50 <80 2 Pl leading >=60 PL >=50 80-100 3 Pl leading >=60 PL >=50 100-120 4 Pl leading >=60 PL >=50 >=120 5 Pl leading >=60 SX >=50 <50 6 Sx leading >=60 SX >=50 50-100 7 Sx leading >=60 SX >=50 >100 8 Sx leading >=60 FD >=50 <50 9 Fd leading >=60 FD >=50 50-100 10 Fd leading >=60 FD >=50 >100 11 Fd leading >=60 AT >=50 12 At leading 0 Unclassified LVI Summary Report June 10, 2013 Page 20 2. Age Williams Lake Classification 2 (WL_CLASS2) Lead species % Dead Net Volume (m3/ha) Net Volume (m3/ha) <60 WL_CLASS2 Comment 1 Young stands >=60 PL >70 <30 2 Pl leading >=60 PL >70 30-50 3 Pl leading >=60 PL >70 50-70 4 Pl leading >=60 PL >70 >70 5 Pl leading >=60 PL 50-70 <30 6 Pl leading >=60 PL 50-70 30-50 7 Pl leading >=60 PL 50-70 50-70 8 Pl leading >=60 PL 50-70 >70 9 Pl leading >=60 SX >=50 <50 10 Sx leading >=60 SX >=50 50-100 11 Sx leading >=60 SX >=50 >100 12 Sx leading >=60 FD >=50 <50 13 Fd leading >=60 FD >=50 50-100 14 Fd leading >=60 FD >=50 >100 15 Fd leading >=60 AT >=50 16 At leading 0 Unclassified WL_CLASS3 Comment 1 Young stands 3. Age Williams Lake Classification 3 (WL_CLASS3) Lead species % Dead Net Volume %* <60 Net Volume (m3/ha) >=60 PL >70 <30 2 Pl leading >=60 PL >70 30-50 3 Pl leading >=60 PL >70 50-70 4 Pl leading >=60 PL >70 >70 5 Pl leading >=60 PL 50-70 <30 6 Pl leading >=60 PL 50-70 30-50 7 Pl leading >=60 PL 50-70 50-70 8 Pl leading >=60 PL 50-70 >70 9 Pl leading LVI Summary Report June 10, 2013 Page 21 >=60 SX >=50 <50 10 Sx leading >=60 SX >=50 50-100 11 Sx leading >=60 SX >=50 >100 12 Sx leading >=60 FD >=50 <50 13 Fd leading >=60 FD >=50 50-100 14 Fd leading >=60 FD >=50 >100 15 Fd leading >=60 AT >=50 16 At leading 0 Unclassified * Percentage of dead net volume = 100 * dead net volume/(live net volume + dead net volume) 4. Age Williams Lake Classification 4 (WL_CLASS4) Lead species % Dead Net Volume (m3/ha) Net Volume (m3/ha) <60 WL_CLASS4 Comment 1 Young stands >=60 PL >50 <30 2 Pl leading >=60 PL >50 30-50 3 Pl leading >=60 PL >50 50-70 4 Pl leading >=60 PL >50 >70 5 Pl leading >=60 SX >=50 <50 6 Sx leading >=60 SX >=50 50-100 7 Sx leading >=60 SX >=50 >100 8 Sx leading >=60 FD >=50 <50 9 Fd leading >=60 FD >=50 50-100 10 Fd leading >=60 FD >=50 >100 11 Fd leading >=60 AT >=50 12 At leading 0 Unclassified 5. Williams Lake Classification based on Fuzzie C-Means Classification (WL_CLASS5) The classification was developed based on a clustering analysis of leading species, stand structure and age attributes, resulting in a total of 21 classes. LVI Summary Report June 10, 2013 Page 22 6. Spectral Generalized Polygons (LANDSAT_POLY) Unique spectrally homogeneous polygons were created from Landsat 8 segments and attribute values were computed from all the segments within polygons weighted by segment areas. LVI Summary Report June 10, 2013 Page 23 7.3 LVI Data Dictionary 1. The LVI Photo Database Field Name OBJECTID Shape FOREST_COVER_OBJECT_ID SEGMENT_AREA Class FEATURE_ID TEIS_ID LANDSAT_ID BEC_ZONE BEC_SUBZONE LAYER_ID MAPSHEET_ID POLYGON_NUMBER CROWN_CLOSURE_PCT TREE_SPECIES_CODE_1 SPECIES_PCT_1 TREE_SPECIES_CODE_2 SPECIES_PCT_2 TREE_SPECIES_CODE_3 SPECIES_PCT_3 TREE_SPECIES_CODE_4 SPECIES_PCT_4 TREE_SPECIES_CODE_5 SPECIES_PCT_5 TREE_SPECIES_CODE_6 SPECIES_PCT_6 PCT_check AGE_1 HEIGHT_1 AGE_2 HEIGHT_2 BASAL_AREA BASAL_AREA_ADJ VRI_LIVE_STEMS_PER_HA LVI_1 LVI_2 DATA_SOURCE INTERPRETER DATE_INPUT SITE_INDEX PROJ_HT PROJ_AGE PROJ_DIA Format integer shape integer float string integer integer integer string string integer string integer integer string integer string integer string integer string integer string integer string integer integer integer float integer float float float integer string integer string string date float float float float LVI Summary Report June 10, 2013 Definition Object ID Shape field Segment ID Segment area Landsat classification VRI Feature ID PEM polygond ID Landsat ID BGC zone ID BGC subzone ID Layer ID ECORA Map ID Polygon ID Crown closure The first tree species ID The first tree species percentage The second tree species ID The second tree species percentage The third tree species ID The third tree species percentage The fourth tree species ID The fourth tree species percentage The fifth tree species ID The fifth tree species percentage The sixth tree species ID The sixth tree species percentage Total percentage Age of the first species Height of the first species Age of the second species Height of the second species Basal area Adjusted basal area Live trees per ha Treed=t, no treed=n Uniform=1, mixed=2 Data source Name of the photo interpreter Date of interpretation Site index Projected height Projected age Projected diameter Comment ArcGIS internal format ArcGIS internal format Unique ID from Ann Grid site ID Same as FOREST_COVER_OBJECT_ID For QA check Original projection in 2013 Page 24 PROJ_BA PROJ_TPH PROJ_NET_VOL PROJ_GROSS_VOL D_TREE_SPECIES_CODE_1 float float float float string D_SPECIES_PCT_1 D_TREE_SPECIES_CODE_2 integer string D_SPECIES_PCT_2 D_TREE_SPECIES_CODE_3 integer string D_SPECIES_PCT_3 D_CROWN_CLOSURE_PCT D_AGE_1 D_HEIGHT_1 D_AGE_2 D_HEIGHT_2 D_BASAL_AREA D_BASAL_AREA_ADJ D_STEMS_PER_HA D_PROJ_HT D_PROJ_TPH D_PROJ_NET_VOL D_PROJ_GROSS_VOL D_PROJ_NET_VOL_ADJ integer integer integer float integer float float float integer float float float float float D_PROJ_GROSS_VOL_ADJ R2_TREE_SPECIES_CODE_1 R2_SPECIES_PCT_1 R2_TREE_SPECIES_CODE_2 R2_SPECIES_PCT_2 R2_TREE_SPECIES_CODE_3 R2_SPECIES_PCT_3 R2_AGE_1 R2_HEIGHT_1 R2_CROWN_CLOSURE_PCT R2_BASAL_AREA R2_VRI_LIVE_STEMS_PER_HA PROJ_4CM_HT PROJ_4CM_AGE PROJ_4CM_DIA PROJ_4CM_BA PROJ_4CM_TPH PROJ_4CM_NET_VOL PROJ_4CM_GROSS_VOL CLASSSPL5 CLASSSTAL5 CLASSDL5 float string integer string integer string integer integer float integer float integer float float float float float float float integer integer integer LVI Summary Report June 10, 2013 Projected basal area Projected stems per ha Projected net volume Projected gross volume Dead the first tree species ID Dead the first tree species percentage Dead the second tree species ID Dead the second tree species percentage Dead the third tree species ID Dead the third tree species percentage Dead tree crown closure Dead age of the first species Dead height of the first species Dead age of the second species Dead height of the second species Original dead tree basal area Adjusted dead tree basal area Dead trees per ha Dead projected height Dead projected stems per ha Dead projected net volume Dead projected gross volume Dead projected net volume adjusted Dead projected gross volume adjusted Layer 2 first species ID Layer 2 first species percentage Layer 2 second species ID Layer 2 second species percentage Layer 2 third species ID Layer 2 third species percentage Layer 2 projected age Layer 2 projected height Layer 2 projected crown closure Layer 2 basal area Layer 2 live trees per ha Projected height at 4cm Projected age at 4 cm Projected diameter at 4 cm Projected basal area at 4 cm Projected stems per ha at 4 cm Projected net volume at 4 cm Projected gross volume at 4 cm Species classification Structure classification Dead tree classification Final dead volume for reporting and analysis Final dead volume for reporting and analysis Page 25 SPLSTAL WLCLASS1 WLCLASS2 WLCLASS3 WLCLASS4 WLCLASS5 integer integer integer integer integer integer Species and structure classification Williams Lake classification 1 Williams Lake classification 2 Williams Lake classification 3 Williams Lake classification 4 Williams Lake classification 5 RA1_BA RA1_PROJ_BA RA1_NET_VOL125 float float float Adjusted basal area 1 Projected basal area 1 Projected net volume 1 RA2_BA RA2_PROJ_BA RA2_NET_VOL125 float float float Projected basal area 2 Projected basal area 2 Projected net volume 2 RA3_BA RA3_PROJ_BA RA3_NET_VOL125 RA4_PROJECTION_YEAR float float float integer Projected basal area 3 Projected basal area 3 Projected net volume 3 Year of projection RA4_BA_ADJ RA4_PROJ_BA RA4_PROJ_AGE RA4_PROJ_HT RA4_PROJ_DIA RA4_PROJ_TPH RA4_PROJ_NET_VOL RA4_PROJ_GROSS_VOL RA4_PROJ_WSV PROJECT_NAME PROJECT_MANAGER INFOR_DATE Shape_Length Shape_Area float float float float float float float float float string string date float float Adjusted basal area Projected basal area 2014 Projected age 2014 Projected height 2014 Projected diameter 2014 Projected trees per ha 2014 Projected net volume 2014 Projected gross volume 2014 Projected whole stem volume 2014 Project name Project manager Information date Length of the polygon Area of the polygon LVI Summary Report June 10, 2013 Refer to the classification definition table Refer to the classification definition table Refer to the classification definition table Refer to the classification definition table Refer to the classification definition table Multiplicator 1.49 for MS, 0.401 for SBPS from Ian Multiplicator 1.391 for MS, 0.7204 for SBPS, from Margaret Multiplicator 1.6058 for MS, 0.77 for SBPS, from Xiaoping 2014 Multiplicator 1.30 for MS, 0.5607 for SBPS, for final The date of creation of this data ArcGIS internal format ArcGIS internal format Page 26 2. The LVI Segment Database Field Name OBJECTID Shape FOREST_COVER_OBJECT_ID SEGMENT_AREA Format integer shape integer float Object ID Shape field Segment ID Segment area ALL_BEC_POST_ID ALL_BEC_POST_NN FEATURE_ID TEIS_ID BEC_ZONE BEC_SUBZONE DEM_MEAN SLOPE_MEAN ASPECT_MEAN centX centY Class LAYER_ID CROWN_CLOSURE_PCT TREE_SPECIES_CODE_1 SPECIES_PCT_1 TREE_SPECIES_CODE_2 SPECIES_PCT_2 TREE_SPECIES_CODE_3 SPECIES_PCT_3 TREE_SPECIES_CODE_4 SPECIES_PCT_4 TREE_SPECIES_CODE_5 SPECIES_PCT_5 TREE_SPECIES_CODE_6 SPECIES_PCT_6 AGE_1 HEIGHT_1 AGE_2 HEIGHT_2 BASAL_AREA VRI_LIVE_STEMS_PER_HA LVI_1 LVI_2 DATA_SOURCE INTERPRETER DATE_INPUT SITE_INDEX PROJ_HT PROJ_AGE PROJ_DIA integer string integer integer string string float float float float float string integer integer string integer string integer string integer string integer string integer string integer integer float integer float float integer string integer string string date float float float float Matched DCS ID Nearest Neighbor ID VRI Feature ID PEM polygond ID BGC zone ID BGC subzone ID Average elevation Average slope Average aspect UTN easting of centroid UTM northing of centroid Landsat classification Layer ID Crown closure The first tree species ID The first tree species percentage The second tree species ID The second tree species percentage The third tree species ID The third tree species percentage The fourth tree species ID The fourth tree species percentage The fifth tree species ID The fifth tree species percentage The sixth tree species ID The sixth tree species percentage Age of the first species Height of the first species Age of the second species Height of the second species Basal area Live trees per ha Treed=t, no treed=n Uniform=1, mixed=2 Data source Name of the photo interpreter Date of interpretation Site index Projected height Projected age Projected diameter LVI Summary Report June 10, 2013 Definition Comment ArcGIS internal format ArcGIS internal format Unique ID Global classification, BEC zone applied with post classification From Ann KNN matched, DCS interpreted, 8, 18 Original projection in 2013 Page 27 PROJ_BA PROJ_TPH PROJ_NET_VOL PROJ_GROSS_VOL D_TREE_SPECIES_CODE_1 D_SPECIES_PCT_1 D_TREE_SPECIES_CODE_2 D_SPECIES_PCT_2 D_TREE_SPECIES_CODE_3 D_SPECIES_PCT_3 D_CROWN_CLOSURE_PCT D_AGE_1 D_HEIGHT_1 D_AGE_2 D_HEIGHT_2 D_BASAL_AREA D_BASAL_AREA_ADJ D_STEMS_PER_HA D_PROJ_HT D_PROJ_TPH D_PROJ_NET_VOL D_PROJ_GROSS_VOL D_PROJ_NET_VOL_ADJ D_PROJ_GROSS_VOL_ADJ R2_TREE_SPECIES_CODE_1 R2_SPECIES_PCT_1 R2_TREE_SPECIES_CODE_2 R2_SPECIES_PCT_2 R2_TREE_SPECIES_CODE_3 R2_SPECIES_PCT_3 R2_AGE_1 R2_HEIGHT_1 R2_CROWN_CLOSURE_PCT R2_BASAL_AREA R2_VRI_LIVE_STEMS_PER_HA PROJ_4CM_HT PROJ_4CM_AGE PROJ_4CM_DIA PROJ_4CM_BA PROJ_4CM_TPH PROJ_4CM_NET_VOL PROJ_4CM_GROSS_VOL CLASSSPL5 CLASSSTAL5 CLASSDL5 SPLSTAL WLCLASS1 WLCLASS2 float float float float string integer string integer string integer integer integer float integer float float float integer float float float float float float string integer string integer string integer integer float integer float integer float float float float float float float integer integer integer integer integer integer LVI Summary Report June 10, 2013 Projected basal area Projected stems per ha Projected net volume Projected gross volume Dead the first tree species ID Dead the first tree species percentage Dead the second tree species ID Dead the second tree species percentage Dead the third tree species ID Dead the third tree species percentage Dead tree crown closure Dead age of the first species Dead height of the first species Dead age of the second species Dead height of the second species Original dead tree basal area Adjusted dead tree basal area Dead trees per ha Dead projected height Dead projected stems per ha Dead projected net volume Dead projected gross volume Dead projected net volume adjusted Dead projected gross volume adjusted Layer 2 first species ID Layer 2 first species percentage Layer 2 second species ID Layer 2 second species percentage Layer 2 third species ID Layer 2 third species percentage Layer 2 projected age Layer 2 projected height Layer 2 projected crown closure Layer 2 basal area Layer 2 live trees per ha Projected height at 4cm Projected age at 4 cm Projected diameter at 4 cm Projected basal area at 4 cm Projected stems per ha at 4 cm Projected net volume at 4 cm Projected gross volume at 4 cm Species classification Structure classification Dead tree classification Species and structure classification Williams Lake classification 1 Williams Lake classification 2 *1.87 Final dead volume for reporting and analysis Final dead volume for reporting and analysis Refer to the classification definition table Refer to the classification definition table Page 28 WLCLASS3 WLCLASS4 WLCLASS5 RA4_PROJECTION_YEAR RA4_BA_ADJ RA4_PROJ_BA RA4_PROJ_AGE RA4_PROJ_HT RA4_PROJ_DIA RA4_PROJ_TPH RA4_PROJ_NET_VOL RA4_PROJ_GROSS_VOL RA4_PROJ_WSV PROJECT_NAME PROJECT_MANAGER INFOR_DATE Shape_Length Shape_Area integer integer integer integer float float float float float float float float float string string date float float LVI Summary Report June 10, 2013 Williams Lake classification 3 Williams Lake classification 4 Williams Lake classification 5 Year of projection Projected basal area 2014 Projected basal area 2014 Projected age 2014 Projected height 2014 Projected diameter 2014 Projected trees per ha 2014 Projected net volume 2014 Projected gross volume 2014 Projected whole stem volume 2014 Project name Project manager Information date Length of the polygon Area of the polygon Refer to the classification definition table Refer to the classification definition table Refer to the classification definition table 2014 Multiplicator 1.30 for MS, 0.5607 for SBPS The date of creation of this data ArcGIS internal format ArcGIS internal format Page 29 3. The LVI Class Database Field Name OBJECTID Shape POLYGON_ID POLYGON_AREA MAP_ID BEC_ZONE BEC_SUBZONE VRI_FEATURE_ID Disturbance_Start_Date Disturbance_End_Date VRI_SP1_CD VRI_SITE_INDEX INVENTORY_STANDARD REFERENCE_YEAR PROJECTION_YEAR DATA_SOURCE Format integer shape integer float string string string integer date date string float string integer integer string Definition Object ID Shape field Polygon ID Polygon area in ha NTS20K Map ID BGC zone ID BGC subzone ID VRI Feature ID Disturbance start date Disturbance end date Leading species ID in VRI Site index in VRI Inventory Standard ID Reference year Projection year Data source CLASSID CLASS_NO CLASS_GROUP CLASS_AREA NO_DCS_SEGMENTS string integer string float integer Classification ID Class ID Class group name Class area Number of segments in the polygon SP1_CD SP1_PCT SP2_CD SP2_PCT SP3_CD SP3_PCT SP4_CD SP4_PCT SP5_CD SP5_PCT SP6_CD SP6_PCT AGE1 HEIGHT1 AGE2 HEIGHT2 string integer string integer string integer string integer string integer string integer integer float integer float The first tree species ID The first tree species percentage The second tree species ID The second tree species percentage The third tree species ID The third tree species percentage The fourth tree species ID The fourth tree species percentage The fifth tree species ID The fifth tree species percentage The sixth tree species ID The sixth tree species percentage Age of the first species Height of the first species Age of the second species Height of the second species SITE_INDEX CROWN_CLOSURE BASAL_AREA LIVE_STEMS_PER_HA NET_VOLUME float integer float integer float Site index Crown closure Basal area Live trees per ha Net volume LVI Summary Report June 10, 2013 Comment ArcGIS internal format ArcGIS internal format unique for no LVI tree area Same start and end dates, for cut and fire L, V, F, I Year of photo or Landsat For VDYP7 projected attributes LVI, LANDSAT, PEM, WFMB, VRI, etc LANDSAT_PO for Landsat polygon, WLCLASS3 (15) for classification, CUT, OPEN, FIRE(LOW, MEDIUM, HIGH), WATER same class ID has same attribute values TREE, NOTREE For LVI, only 4 species, and for VRI it could be up to 6 Weighted average Weighted average Weighted average Weighted average Derived by leading species, age and height, weighted average Weighted average Weighted average Weighted average Primary utilization at 12.5cm, weighted average Page 30 WSV_4CM DEAD_STEMS_PER_HA DEAD_NET_VOLUME DEAD_GROSS_VOLUME sCROWN_CLOSURE sBASAL_AREA sLIVE_STEMS_PER_HA sNET_VOLUME float float float float float float float float sWSV_4CM sDEAD_STEMS_PER_HA sDEAD_NET_VOLUME sDEAD_GROSS_VOLUME float float float float Whole stem volume Dead tree stems per ha Dead tree net volume Dead tree gross volume Strandard Error (SE) of croan closure SE of basal area SE of live stems per ha SE of net volume SE of whole stem volume, gross at 4cm SE of dead tree stems per ha SE of dead tree net volume SE of dead tree gross volume PROJ_CLASS_AGE PROJ_CLASS_HT PROJ_CLASS_DIA PROJ_CLASS_BA PROJ_CLASS_TPH PROJ_CLASS_NET_VOL PROJ_CLASS_GROSS_VOL PROJECT_NAME PROJECT_MANAGER INFOR_DATE Shape_Length Shape_Area float float float float float float float string string date float float Projected age of the class Projected height of the class Projected diameter of the class Projected basal area of the class Projected stems per ha of the class Projected net volume of the class Projected gross volume of the class Project name Project manager Information date Length of the polygon Area of the polygon LVI Summary Report June 10, 2013 Gross volume at 4cm, weighted average All projected are based on the attribute values at reference year The date of creation of this data ArcGIS internal format ArcGIS internal format Page 31 4. The LVI Polygon Database Field Name OBJECTID Shape POLYGON_ID POLYGON_AREA PIXELS BEC_ZONE BEC_SUBZONE FEATURE_ID VRI_MAX_POLY_AREA VRI_POLY_AREA DATA_SOURCE REFERENCE_YEAR PROJECTION_YEAR INVENTORY_STANDARD Format integer shape integer float integer string string integer float float string integer integer string Definition Object ID Shape field Polygon ID Polygon area in ha Number of Landsat pixels BGC zone ID BGC subzone ID VRI Feature ID Area of VRI portion in the LVI polygon VRI original polygon area Data source Reference year Projection year Inventory Standard ID CLASSID CLASS_NO CLASS_GROUP NO_SEGMENTS string integer string integer Classification ID Class ID Class group name Number of segments in the polygon SP1_CD SP1_PCT SP2_CD SP2_PCT SP3_CD SP3_PCT SP4_CD SP4_PCT SP5_CD SP5_PCT SP6_CD SP6_PCT AGE1 HEIGHT1 AGE2 HEIGHT2 BASAL_AREA CROWN_CLOSURE string integer string integer string integer string integer string integer string integer integer float integer float float integer The first tree species ID The first tree species percentage The second tree species ID The second tree species percentage The third tree species ID The third tree species percentage The fourth tree species ID The fourth tree species percentage The fifth tree species ID The fifth tree species percentage The sixth tree species ID The sixth tree species percentage Age of the first species Height of the first species Age of the second species Height of the second species Basal area Crown closure SITE_INDEX LIVE_STEMS_PER_HA float integer Site index Live trees per ha NET_VOLUME float Net volume GROSS_VOLUME WSV_4CM float float Gross volume Whole stem volume PROJ_AGE float Projected age LVI Summary Report June 10, 2013 Comment ArcGIS internal format ArcGIS internal format unique for no LVI tree area LVI, LANDSAT, PEM, WFMB, VRI, etc Year of photo or Landsat For VDYP7 projected attributes L, V, F, I LANDSAT_PO for LVI data, CUT, OPEN, FIRE(LOW, MEDIUM, HIGH), WATER TREE, NOTREE For LVI, only 4 species, and for VRI it could be up to 6 Weighted average Weighted average Weighted average Weighted average Weighted average Weighted average Derived by leading species, age and height, weighted average Weighted average Primary utilization at 12.5cm, weighted average Primary utilization at 12.5cm, weighted average Gross volume at 4cm, weighted average All projected are based on the attribute values at reference year Page 32 PROJ_HT PROJ_DIA PROJ_BA PROJ_TPH PROJ_NET_VOL PROJ_GROSS_VOL D_SP1_CD D_SP1_PCT D_AGE D_HT D_BA D_CC D_TPH D_NET_VOL D_GROSS_VOL sBASAL_AREA sCROWN_CLOSURE sLIVE_STEMS_PER_HA sNET_VOLUME sGROSS_VOLUME sWSV_4CM sD_AGE sD_HT sD_BA sD_CC sD_TPH sD_NET_VOL sD_GROSS_VOL PROJECT_NAME PROJECT_MANAGER INFOR_DATE Disturbance_Start_Date Disturbance_End_Date VRI_SP1_CD VRI_SI EST_SITE_INDEX_SPECIES_CD EST_SITE_INDEX SP_SI Shape_Length Shape_Area float float float float float float float float float float float float float float float float float float float float float float float float float float float float string string date date date string float string float float float float LVI Summary Report June 10, 2013 Projected height Projected diameter Projected basal area Projected stems per ha Projected net volume Projected gross volume Dead tree first species Dead tree first species percentage Dead tree age Dead tree height Dead tree basal area Dead tree crown closure Dead tree stems per ha Dead tree net volume Dead tree gross volume Strandard Error (SE) of basal area SE of crown closure SE of live stems per ha SE of net volume SE of gross volume SE of whole stem volume at 4cm SE of dead tree age SE of dead tree height SE of dead tree basal area SE of dead tree crown closure SE of dead tree stems per ha SE of dead tree net volume SE of dead tree gross volume Project name Project manager Information date Disturbance start date Disturbance end date Leading species ID in VRI Site index in VRI Estimated site index species in VRI Estimated site index in VRI Site index from the Site Productivity 2014 Length of the polygon Area of the polygon PL 100 The date of creation of this data Same start and end dates, for cut and fire ArcGIS internal format ArcGIS internal format Page 33