Geog350 Mallory

Title A Mapping Update of Vegetation in the American River Parkway
Author Jeffrey Mallory American River College, Geography 350: Data Acquisition in GIS; Fall 2006 Contact Information: mallorj@imail.losrios.edu
Abstract The American River Parkway is a fantastic diverse greenbelt that traverses through Sacramento. The vegetation in the Parkway was mapped around 1993 but the layer that was produced had inherent errors that needed to be updated. With the use of object oriented spatially based software a segementation process was run to create a new set of polygons for the Parkway encompassing the area maintained be the Sacramento County Department of Parks and Recreation. A regionalization process was the run to populate the new polygons with the existing vegetation data. Manual on screen eediting of vegetation types was done to complete the map.
Introduction The American River Parkway within the confines Sacramento is a fantastic oasis amoungst the buzzing population. The American River Parkway was purchased by the County of Sacramento as a way to perserve its rich natural habitat and stop urban encroachment. The 5000 acre 23-mile stretch of the American River form the mouth at Discover Park to the Nimbus Dam is managed by the Sacramento County Department of Parks and Recreation. This stretch is home to many resident and migratory birds,several egret rookeries, deer, otters, beavers, and more. The river itself is host to a rich fisheries habitat that includes salmon, trout,steelhead, striper, and shad. Recreational atictivies in the Parway abound with up to 5 million visitors a year enjoying its beauty and grandure. An accuarate and precise vegetation map of the parkway curently does not exist. There is a partial vegetation map that was created sometime around 1993. An updated map with more preicsion will go a long way to help with the management native and non-native species as well as provide for a way to monitor the gain or loss of habitat within the confines of the Parkway. My approach was to create a new set of polygons using 2005 NAIP (National Agriculture Imaging Program) as the base imagery. The existing vegetation data from the Sacramento County Parks and Recreation was then regionalized to the new set of polygons. The existing data was also cross-walked to the the US Forest Sevice Region 5 CALVEG vegetation type which was subseqently cross-walked to the California Wildlife Habitat Relationships (CWHR) system. There was also extensive field work done to idenify vegetation types by polygon that were ultimately used to populate new polygons where data from the existing product does not exist or where new polygons have created a change in vegetatiion type due to the finer precision of the new polygons. Extensive editing of the new database for updates was done to create a fianlized database. Finally a metadata file was created.
The American River Parkway from Discovery Park to Nimbus Dam.
Background The American River Parkway was mapped sometime around 1993. The exact date was not know by natural resource specialist, Trevor Burwell, at the Sacramento County Department of Parks and Recreation.To further complicate matters the person who originally did the mappping no longer worked there.The format of the existing map was one of hand delineated polygons saved as a shapefile. I used this as the basis for the map I created, but there were inherent problems that had to be solved before the layer could be of use. When I recieved the shapefile there was no projection file with it and When I brought it into ArcView to view it over imagery the two layers would not allign. The layer is composed of 9700 polygons but 2000 of the polygons did not have data. There was simply an "x" in the field for vegetation type, so the layer was also incomplete. There were vegetation types but no acompanying vegetation descriptions. There also was no metadata file associated with shapefile.
Methods Software Used ArcGIS 9.1 Erdas Imagine ARC Command Line AML ArcView 3.3 Definiens eCognition Solving the Projection and Vegetation Descriptions Problems The problem with the projection was solved determining what projection other county data was in. So the first one I tried with the data I had was NAD_1983_StatePlane_California_II_FIPS_0402 GCS_North_American_1983, and it worked. I posed the problem of vegetation descriptions to US Forest Service Ecologist Hazel Gorden, and once she saw the vegetation types and knew the year of the data she immediatly suggested "Preliminary Descriptions of the Terrestrail Natural Communities of California" by Robert F. Holland. She was correct. Most of the vegetation types list with the county data were described there. I had overcome two major hurdles in a short period of time. The Segmentation Process Segmnetation is the process of creating polygons over imagery. The imagery I had was 2005 NAIP with a resolutuion of 1 meter in MrSID format.With this type of imagery and resolution the typical spectral segmentation process is not appropriate. For this type of imagery a pattern recognition nad object oriented process is what is used. That is what the Definiens eCognition software does. It looks at features and their spatial relationship in the evirronment. But the Deginiens eCognition software does not support MrSID images as an input, so I had to use Erdads Imagine to converrt the MrSID image into Imagine image ( a .img). This is the image that was used in the eCognition software. Only three parameters need to be set in eCognition: Scale, smoothness, and compactness. After several trial runs I found that scale=65, smoothness=5, and compactness=5 gave me the best results. The output from eCognition is a tiff image. This needs to be converted into a coverage. An AML has been written to do this process. In ARC command line the segcreate AML is run and the tiff is used as an imput. The output from the AML is a coverage of polygons that have unique id's as well as acres which the segcreate.aml calls from another AML that computes acres for each polygon.
Segmentation polygons over 2005 NAIP imagery
The Regionalization Process Regionalization is the process of taking existing county vegetation data and populating the new polygons with it. The first thing that needed to be done was to create a field in the existing county vegetation data set that converted each vegetation type to an integer. This is done so that the existing data can be coverted to a GRID, where each GRID cell, of 1m resolution, is populated with the integer value that represents vegetation type. Now the melding together ot the polygon coverage and the existing vegetation GRID can be done. This has been found to be most proficient using ArcView 3.3 Spatial Analyst. An avenue script was developed to do this process. What the script does is look at the shapes of the polygons over the GRID and based on the vegetation integer value that occurs the most in the in the polygon (ZonalMajority) a vegetation type is assigned to that polygon. The output is a coverage of polygons that have existing county vegtation type assigned to each polygon.
Regionalization--New polygons (red oulines) get the label of the old polygons (green outlines)
Fieldwork and Editing Fieldwork consisted of taking out maps of polygons drawn over 2005 NAIP imagery and labeling them with vegetation types the were observed. I did this a several locations along the American River Parkway. Field observations were done at the Nimbus basin, Sailor Bar, Sunrise, ElManto, Arden Ponds, Watt Avenue, Howe Avenue, Paradise Beach, and Discovery Park. These map notes were than entered into the database for each poygon to assign vegetation types. At the same time image interpretation was done on the map noted polygons so that polygons that were similar looking could be labeled in areas that were not visited. This process was done for the polygons that were labeled with a unknown vegetation type as well as those polygons that had changed due to the finer resolution of the new polygons.
Editing vegetation type from imagery
Results The segmentation process crated polygons that captured the spatial distribution of vegetation with much more precision then the origianl vegetation product. The regionalization process also worked well, but with the finer detail in the polygons there were new polygons that had bad labels. These labels, along with the unknown labels in the original data, had to be edited on screen to attribute the correct vegetation type.
Analysis The origianl sacramento county vegetation layer has 9700 polygons and the new layer has 13500 polygons. Approximately 3500 polygons now had bad labels because there was a change in lifeform (ex-old polygons of trees were now a shrub type). Along with the 2000 original polygons that had labels of unknown vegetation type this made for an intensive manual editing task that was the most time consuming portion of the process.
Conclusions The segmention and regionalization processes worked quite well for this project. The amount of manual editing that had to be done was not originally anticipated to be such a time consuming portion. An alternate approach in the future might be to do a lifeform edit (ie edit to shrub, hardwood, grass) instead of going straight to the vegetation type edit. The use a nearest neighbor routine could then be used to populate vegtation type by lifeform. The only attribute mapped for this project was vegetation type. To be more complete tree diameter and tree crown closure should also be mapped. CALVEG vegetation type is still being worked on and the cross-walk to WHR still needs to be done. There are also vegetation communities in the project that did not fall under any current vegetation discriptions and so descriptions of these types need to be generated. Finally until the above situations are addressed and completed the meatdata file is also currently incomplete.
References Burwell, Trevor. Natural Resource Specialist with the Sacramento County Department of Parks and Recreation Gorden, Hazel. Ecologist at the US Forest Service Remote Sensing Lab Holland, Robert F., 1986. "Preliminary Descriptions of the Terristrial Natural Communitiew of Claifornia." USDA Forest Service - Remote Sensing Lab, Ecosystem Planning http://www.fs.fed.us/r5/rsl/clearinghouse/gis-download.shtml