Forest Stewardship Analysis Project - Pilot Overlay Study Methodology for the State of Connecticut Outline • Project Summary • Base layers and computing requirements • Factors of Influence and Data layer development Fire Protection Assessment: Risk of Insects/Pests: Risk of Development (change in census block households): Private Forested Lands: Wetlands: Forest Patches: Riparian Corridors: Natural Heritage Priority Habitats (Natural Diversity Database): Proximity to Publicly Protected Lands: Slopes: Public Water Supply Areas: Priority Watersheds: Analysis Mask: Private Forest Analysis Layer: Non-forest – Non-developed (NFND) Mask: • The Overlay Model Model Overview Data Aggregation Weighting Numbering system • Web site development Site purpose State participation Project Summary The Spatial Analysis Project (SAP) has as one its objectives the creation of datasets for states that highlight sites where there would be potential benefit from, or suitability for inclusion within, the Forest Stewardship Program. Such layers could assist Sate Forestry agencies with the identification of potential sites and help the US Forest Service assess the effectiveness of its programs. As stated by the Missouri participant, private land program and GIS staff from the four states involved in the pilot SAP effort (Connecticut, Maryland, Massachusetts and Missouri), along with Forest Service program and GIS staff, identified 12 factors which help identify the “Stewardship potential” of a given piece of land. The pilot states worked together to develop layer source materials, common and available for each state, designed to lead to this final set. The factors were then separated into two groups, resource potential and resource threats, and later combined using weighted ranking within a GIS model to create the final analysis maps. CT_SAPMethodology.doc 3/30/2006 Page 1 of 8 Factors of Influence and Data layer development Fire Protection Assessment: This data layer is spatially organized by local watershed basins, highlighting moderate to high potential for fire and the potential of those fires as a risk to property. The layer was derived using combinations of slope, aspect, and land cover. The results then summarized over local watershed basins. Risk evaluation was generalized to rank low, medium, and high potential. Medium and high were combined to create the fire risk layer. Risk of Insects/Pests: This layer shows areas of Connecticut that have had measured insect infestations within the last 10 years. Data provided by the USFS. The working set used for the pilot study only included small areas of gypsy moth defoliation within that period. Other impacts such as frost damage and drought were not considered. Risk of Development (change in census block households): This layer shows areas of Connecticut were change in households per square mile between 1990 and 2000 where the increase in the number of households has been 0 to 20. Change in this range is considered high enough to highlight the need for Stewardship, yet low enough where changes in attitude can still provide an alternative to urban sprawl. For Connecticut the Census block group data was collected and combined by the USFS. Private Forested Lands: This datalayer describes forested land in Connecticut that are not within the extent of the Analysis Mask. The forested land coverage was derived by querying fields from the MRLC National Land Cover Datalayer (NLCD) for forested uplands and woody wetlands. Shrublands were not considered for CT because of the high correlation with Powerline right of ways in the MRLC dataset within the State. Other types of stewardship may be appropriate for powerline areas; however, they were removed from the Private Forest layer because the intent is to highlight forest activities. Wetlands: This layer shows wetland areas in Connecticut. Statewide coverage was derived from two sources, US Topographic data (DRG’s) and NRCS Wetland soils data. Soils have been transferred to GIS for all but six State quadrangles; wetlands derived from this data are preferred because CT regulations are soils based. National Wetlands Inventory data does not exist in electronic format for CT. CT_SAPMethodology.doc 3/30/2006 Page 2 of 8 Forest Patches: The source of the Forested related information in the pilot study has been MRLC data. Our forest cover layer was derived from the MRLC layer for Connecticut by reselecting forest land cover classes (see below), then running several modeling functions to create patch features. Applications used were Patch Analyst and ArcView 3.3 Spatial Analyst. Note: The other three pilot states formed the patch routines by first converting the data to Vector format. This was done to overcome the edge proximity links associated with Grid pixels that only touch at their corners. Within Connecticut, all grouping operations were prepared using the data in the Grid format; the proximity issue was overcome using the Patch Analyst extension (see below). An advantage is computation speed and reduced data size. The operations used by Connecticut should work for other states, but a recent ArcGIS 9.0 model designed by the Colorado State Forest Service is probably the best way to go. This model uses raster formats and does not require stepping out of the model to create patch elements. For Connecticut’s model, the Forest features were reselected from the MRLC data. Forest data layers features were selected and a new layer created emphasizing the forest classes. Note: Shrublands were included in this data set. While not forested, for this project shrubland was considered as a non-fragmenting entity. This resulting layer was later used with Private Lands, minus the shrubland layer (see reason above). After fusing the layer with 30 meter grid of major roads (available through CT DEP), the layer was further reclassed to 1/0, forested or non-forested. This resulting output was run through the Patch Analyst program, using tools designed to determine patch statistics for a given layer. A byproduct of the operation is a new layer with unique group IDs for each fully enclosed patch. The user has a choice of weather or not to use the diagonals (pixels touching at their corners) as a route of transfer between blocks. With a unique ID for each cluster the areas were determined by pixel count. All areas with pixel counts greater than 50 acres were reselected and placed into the final data layer. At this point the final patch layer is complete; however, another product of the patch analyst operation is the perimeter of each group. Because of the convoluted nature of the forest blocks in New England it can be hard to consider a block unfragmented simply because it’s large. Many fragmenting features, like subdivisions, extend into a forest block, but do not completely bisect the region. The result is a block with holes and folds that modify the integrity of the forest. To provide a measure of this integrity one can divide the perimeter of the forest block by its area and measure the resulting perimeter to area ratio; the larger the number, the more fragmented the block. Essentially more surface per unit area means more compromise when considering interior species or applications that require non-encroachment. While not used for this project, the product may be of interest. CT_SAPMethodology.doc 3/30/2006 Page 3 of 8 Riparian Corridors: This raster data layer shows 300 foot riparian corridors encompassing perennial stream and river features from the Digital Raster Graphics (DRG) quad based hydro data set from the CT DEP. Areas near or touching stream features are given a weighted value within the model. All linear streams were selected from the DEP vector data set and buffered to 300 feet on either side (a distance settled on by the SAP team). All river polygon features were also selected and buffered to 300 feet, the original polygons then removed. The two results were combined to create one buffered dataset which was then converted to a 30 meter Grid format consisting of a 1/0 relationship – within buffer equal to one, all other areas equal to zero. Natural Heritage Priority Habitats (Threatened and Endangered Species): This raster data layer depicts estimated priority habitats of rare species in Connecticut. The State of CT DEP provides these data as a vector Natural Diversity Database file. The layer consists of generalized circles and corridors, providing a proximity to the areas of concern, yet protecting the resource by not compromising the exact location. The vector data was converted as is to a 1/0 raster dataset. Proximity to Publicly Protected Lands: This raster layer represents areas touching or in proximity to existing Public (protected) lands. The CT DEP provides three property ownership layers within their vector data sets: Federal, State, and Municipal Open Space. Only those attribute features considered as natural resources were selected, schools, cemeteries, and other public yet non-natural features were removed. These three vector layers were then buffered to 1000 feet. The buffer provides a mathematical tool for measuring proximity to the protected resource. The centers were removed and the resulting buffered layers converted to 1/0 grid layers. These were then combined into one final dataset consisting of the buffers around publicly protected lands. Slopes: This raster data layer describes areas of Connecticut where the slope is more than 15% and less than 30%. Percentage slope was derived from a statewide NED 30 meter Digital Elevation Model (DEM) raster file using the Spatial Analyst ‘Surface Analysis’ tool. The final slope layer was then queried to the desired slope range of 15% to 30% slope, providing a 1/0 Grid dataset. Public Water Supply Areas: This raster data layer is a composite derived from the following 3 CT DEP vector data layers: Wellhead Protection Areas – Zone II, Wellhead Protection Areas – Interim, and Surface Water Supply Protection Areas. The final grid is a 1/0 layer that describes areas falling into the above categories. Watersheds: This raster data layer highlights the Major HUC watersheds considered under stress as rated by the USGS. In Connecticut, all but the Farmington River Watershed are included in this bracket. The classified watersheds cover most of the State. The vector data from the HUC USGS set were converted to Raster and given the 1/0 Grid format. Analysis Mask: The analysis mask represents urban, open water, and protected public lands. Features in this layer are considered outside the areas where the private CT_SAPMethodology.doc 3/30/2006 Page 4 of 8 stewardship will be applied; to be removed from the full analysis. Urban and open water features were selected from the MRLC data set. The Connecticut hydrography vector data provides more detail for lake features, but the MRLC data is consistent across state boundaries. Protected public lands were selected from the DEP public lands data sets. Unlike the proximity to protected lands layer, where the interests are natural resource features, public schools, cemeteries, and park features were included in this mask. These are areas outside of the stewardship program. As a layer the mask features were assigned a value of 1, all others a value of zero. During model applications the 1 was converted to Null, canceling all calculations in those areas removing them from the final summaries. Private Forest Analysis Layer: Part of the final analysis is a layer depicting potential stewardship high, medium, low analysis (HML) on Private lands with existing forests. During the model analysis the Private Forested Lands layer was given a priority coding and combined within the model to create the final HML layer. In this analysis the private layer acts as an inverse mask, all features outside the layer assigned a value of Null, those within a 1. Multiplying this layer with the final HML layer creates an HML layer depicting on these private forested lands (no agriculture or municipal lands). Non-forest – Non-developed Mask (NFND): As an inverse to the Private Forest Analysis, there was in interest in summarizing the total potential stewardship analysis over lands that were not forested but still have potential for stewardship activity. This layer depicts the mask used for this final summary by selecting Agricultural lands (opposite of forest selection) and non-urban features (those not included in full analysis mask) from the MRLC dataset. The mask features were assigned a value of 1, all others a value of 0. Multiplying this layer with the final HML data layer creates an HML layer depicting stewardship potential outside the forested areas. CT_SAPMethodology.doc 3/30/2006 Page 5 of 8 The Overlay Model Model Overview ArcView 3 Model Builder has an upper limit on the number of pixels it can handle in a given operation. To maintain the integrity of the 30 meter resolution the state was divided into three blocks, each with a pixel count under the Model limit. The model computations were run three times (see diagram below) and the final results combined to the statewide layer using spatial analysis tools. For a state this small the inconvenience was minor. As far as I know the ArcGIS 9.0 Model Builder (built in) does not have this size limitation. ArcView 3 Model Builder also had limitation with Null values, erasing features from the dataset if Null was used instead of zero (0). To compensate all non-weighted and mask values were given a value of zero (0). Data Aggregation – Each of the 12 layers was assigned its multiplier (weighted values) then combined through simple spatial addition within model builder. The resulting output layer contained values from 0 to 24, the maximum for the given layers and overlay patterns for this Connecticut model. If all layers were weighted as 3 and overlaid directly the possible maximum would be 36. For the final three level output the values had to be grouped according to the natural breaks technique. The result was the final analysis map with values designated as low, medium, and high priority. A mask was then used to remove lands considered outside the stewardship program designation (urban area, open water, public lands). The overall process was repeated for the resource layers and the threat layers, providing separate low, medium, high analysis maps for the two types. In addition to the primary layers, the forested lands that had earlier been separated as a unique layer were used as a mask overlay to summarize and display only those priority CT_SAPMethodology.doc 3/30/2006 Page 6 of 8 areas covered by lands where trees currently exist. While Stewardship is viable and encouraged on non-forested lands having potential for growing trees, those lands with existing forested systems should be identified for their immediate forest benefits and resource protection. With just one harvest, characteristics uniquely developed over hundreds of years can be destroyed, or timber values set back to unsustainable levels, if resource protection procedures are not followed. Weighting –The Connecticut procedures for weighting ended up more qualitative than the quantitative (fractional) techniques used by Massachusetts and Missouri. Like the other States our CT Forest Stewardship committee reviewed the list of data layers for priority considerations, ranking them according to the techniques described by Massachusetts. The result was a fractional list of weights for each layer. In addition, each member was asked to rank the individual layers as high, medium, or low. Of the three methods for weighting identified in the Massachusetts report, the third did not return conclusive results so it was removed from consideration. For Connecticut the 1-12 method identified in their report did not have the required significant trends for analysis, so it was also removed. The 0-2 by decimal (fractional result) technique did show some trends. The final fractional list provided by the committee was then grouped (a type of natural breaks determination prior to the final mapping summary). The conclusion was to summarize the high, medium, low evaluation into weighted values of 3, 2, or 1 for each layer. With each layer assigned its ranking the layers were then added together to create a final map - resulting in values of 0 for no data to 24 for the maximum recorded summary (if all layers were assigned the value of 3 the theoretical maximum would be 3 x 12 or 36). For the final output map these were then grouped into the high, medium, low categories using equal interval breaks. Natural breaks is unavailable as a summary choice when precategorized (non-continuous) data is used - another reason for using the fractional methods described in the Massachusetts and Missouri reports. Numbering System - Throughout the project various names were being used to identify the data layers for the pilot project. Unique fixed names or identifying codes are essential to provide common data sets across state boundaries. We settled on a numbering system for each unique layer in part because a computer database (in our case a database driven web site) worked most efficiently with a number based system. The names could change, but if states provided reference to the number code the other states could easily match layers within the database. A four digit code was settled on. Initially the base data layers were grouped by categories of specific resources (water, forest, urban resources, etc.). The number scheme was developed based on these initial groupings; the first two numbers representing the group, the last two the layer within that group (two digits to allow for expansion). CT_SAPMethodology.doc 3/30/2006 Page 7 of 8 There is less of a tie to these initial groupings, but the numbers stayed the same. In the future, if the need arises, a new numbering set can be developed. When dealing with partner states, computer database tools, and web based applications, the importance of a unique number based code for each layer cannot be overstated. Web site development Site Purpose - To highlight progress for each pilot state and provide a mechanism for viewing techniques and output results. The feedback helped identify differences between states before analysis progressed too far. Site Participation – A web based form was provided to allow each State to update their individual progress records as they completed the layers. Map samples were provided to my office for processing as thumbnail images for display on the site. The update form was rarely used by the individual states, requiring updates from this central office. While not automatically updated as hoped, the resulting web page with image captures of each state layer did prove very useful for the project. Credits The layout and content flow for this report was based on the structure of the methodology report from our project partner in the State of Massachusetts. Supplemental information on the project summary was stolen from the Missouri report. The material content was authored and edited by Joel Stocker from the University of Connecticut, Cooperative Extension System, Middlesex County Extension Center, Haddam, Connecticut with extensive project support by Tom Worthley, Associate Educator, Cooperative Extension System. CT_SAPMethodology.doc 3/30/2006 Page 8 of 8