Sunteți pe pagina 1din 84

EVALUATION OF LOCAL LAND COVER / LAND USE MAPPING APPROACHS FOR THE CHESAPEAKE BAY WATERSHED

FINAL REPORT 11-10-2005

Stewart Bruce Rick L. Day The Pennsylvania State University

Table of Contents Introduction..........................................................................................................................5 Land Use..............................................................................................................................5 OBJECTIVES......................................................................................................................6 Objective #1: Evaluation of methods to produce local land use/land cover data...............6 Methodologies.................................................................................................................6 County-Level Survey...................................................................................................6 Selection of Pilot Counties...........................................................................................8 Land use/Land Cover Classification System...............................................................8 Land use/Land Cover Data Collection Methodologies................................................9 Results and Discussion..................................................................................................16 County-Level Survey ................................................................................................16 Locally Generated Land Use/Land Cover.................................................................18 It is very important to note that these costs are for first time data collection. Once a county has digital land use/land cover data, subsequent updating costs would be significantly reduced resulting in lower long-term maintenance costs......................22 Parcel Based Land Use Codes...................................................................................23 Other Options: Combining County Level Data with Satellite-Derived Data............23 Objective #2: Comparison of locally produced land use/land cover data to satellite derived data........................................................................................................................28 Methodologies...............................................................................................................28 Land Cover Comparisons..........................................................................................28 Impervious Surface Comparisons..............................................................................31 Results and Discussion..................................................................................................35 Land Cover Comparisons..........................................................................................35 Impervious Surface Comparisons..............................................................................53 Appendix B: Anderson Land Use/Land Cover Classification Code............................61 Appendix C: New York State Assessment codes translation to Anderson Code.........69 Appendix D: QC Procedure for Land Use Digitizing Product Quality Control ..........78 Appendix E: Procedure for creating confusion matrix to compare photo interpretation land use and satellite derived land use...........................................................................81

Table of Figures Figure 1: Breakdown of GIS Survey Access tables............................................................7 Figure 2: Henrico County Zone Map................................................................................12 Figure 3: Henrico County Residential and Transportation...............................................13 Figure 4: Henrico County Null Space...............................................................................14 Figure 5: Henrico County Tile Index Map.......................................................................15 Figure 6: Number of Dedicated GIS Workers by County................................................17 Figure 7: Northern Henrico County parking lots (in grey)...............................................22 Figure 8: Sample parcel from Mifflin County, PA...........................................................23 Figure 9. RESAC land cover and residential parcel boundaries........................................24 Figure 10. RESAC classification of residential parcels.....................................................25 Figure 11: Harford County image showing planimetric impervious surface over color aerial imagery.....................................................................................................................31 Figure 12: Resample of Harford County planimetric to 5 meter cell raster.....................32 Figure 13: Aggregate of Harford County 5 meter to 30 meter impervious data layer......32 Figure 14: Rural sample area............................................................................................33 Figure 15: Suburban sample area......................................................................................34 Figure 16: Urban sample area...........................................................................................34 Figure 17: Henrico County 10 meter LAL land cover data..............................................37 Figure 18: Henrico County 30 meter RESAC land cover data.........................................37 Figure 19: Henrico County 10 meter LAL land cover data..............................................38 Figure 20: Henrico County 30 meter RESAC land cover data.........................................38 Figure 21: RESAC land cover non-simplified..................................................................39 Figure 22: RESAC simplified land cover showing major cloverleaf intersection ...........39 Figure 23: LAL land cover data for Baltimore County....................................................42 Figure 24: RESAC simplified land cover data for Baltimore County..............................42 Figure 25: RESAC land cover data showing urban classifications .................................43 Figure 26: LAL land cover for Treasure Lake area..........................................................46 Figure 27: RESAC land cover for Treasure Lake area.....................................................46 Figure 28: LAL land cover showing reclaimed strip mine area.......................................47 Figure 29: RESAC land cover in reclaimed strip mine area.............................................47 Figure 30: LAL land cover for York County, PA.............................................................50 Figure 31: NLCD land cover for York County, PA..........................................................50 Figure 32: Impervious Surface areas not mapped by RESAC..........................................54 Figure 33: Impervious Surface overestimated urban areas...............................................55 Figure 34: Planimetric overlaid onto RESAC data shown in Figure 31...........................55

Table of Tables Table 1 Comparison of Centre County, Pa local land cover data with 2000 NLCD land cover data ..............................................................................................................................................5 Table 2: Aerial Imagery Sources......................................................................................11 Table 3: Time Study for Digitizing Land Use..................................................................19 Table 4: Tioga County , Pennsylvania. Comparison of land use acreage total differences between 1:2400 digitizing versus digitizing on one meter resolution 1:4800 DOQQ for Antrim Quad......................................................................................................................20 Table 5: Tioga County , Pennsylvania. Comparison of land use acreage total differences between 1:2400 digitizing (NAD83) versus digitizing on one meter resolution DOQQ(UTM) for Knoxville Quad....................................................................................21 Table 6: Cost Estimates to Digitize Land Use for Entire Bay..........................................22 Table 7: List of Counties That Supplied GIS Datasets.....................................................27 Table 8: UMD RESAC code translation to Anderson code.............................................30 Table 9: Henrico County versus RESAC Confusion Matrix............................................36 Table 10: Baltimore County versus RESAC Confusion Matrix.......................................41 Table 11: Clearfield County versus RESAC Confusion Matrix.......................................45 Table 12: NLCD code translation to Anderson code........................................................48 Table 13: York County, PA, LAL versus NLCD Confusion Matrix................................49 Table 14: UMD RESAC and NLCD 2000 Pasture and Cropland acreage Totals............52 Table 15: Pennsylvania Department of Agriculture 2002 Pasture and Cropland Acreage Totals..................................................................................................................................52 Table 16: Impervious Surface Comparison between RESAC and Harford County.........54

Introduction
The Penn State Cooperative Extension Program and the Cooperative Extension GIS Program (CEGIS) provide education and technical support to counties throughout the Chesapeake Bay watershed and as such have developed close relationships with sponsoring county governments. Technical support to county governments includes the use of geospatial technologies such as remote sensing, GIS, and GPS, for environmental assessment, land use planning and agriculture. Through these efforts CEGIS has come to recognize that many local governments routinely collect and maintain high-quality GIS information for their own internal purposes. Furthermore, we have realized the potential for these data to be integrated into regional-scale land cover assessment efforts to provide a vastly improved land cover and land use database that will be valuable both locally and regionally. Conventional land cover data bases produced for the Bay such as EMAP, EMAP2MRLC, and NLCD are often inadequate to address land use issues associated with urban sprawl and farmland protection and may lead to erroneous results for modeling efforts dealing with pollutant transport and sources. For example, a comparison of detailed land use information collected for Centre County, PA showed significant differences compared to NLCD land cover information (Table 2).
Table 1 Comparison of Centre County, Pa local land cover data with 2000 NLCD land cover data Local Land Use (acres) Aggregated NLCD (acres) Difference (acres) Difference %

Land Use
Agriculture Residential & Other Developed Forest Quarries/Mined Lands/Transitional Open Water

126071 65511 499674 16402 4608

155775 10924 528692 13997 4089

29704 -54587 29019 -2405 -519

+24 -83 +6 -15 -11

The NLCD database significantly overestimated the amount of agricultural and forest land and conversely underestimated the amount of developed land. This is largely due to the inability of Landsat satellite imagery to distinguish the utility of the land from the surface cover. Many residential developments in forested areas were mistakenly mapped as forest cover and many low-density residential areas were mistakenly mapped as agriculture. These problems render such databases of limited utility for local and regional planning activities, especially those addressing urban sprawl. Additionally, an agricultural non-point source pollution model will be detrimentally impacted when such large discrepancies exist in the land use database, especially when such models are so

closely dependent on land use practice and pollutant load relationships. Therefore more detailed information is needed. Most counties in the Chesapeake Bay watershed, and especially those that are experiencing pressures from urban sprawl, are producing high-quality digital databases that include data such as road networks, digital orthophotos, parcel boundaries, land use, and building locations etc. The local uses of these data within county governments are largely for planning, tax assessment, and emergency communications (E911). Typically these data meet 1=400 National Map Accuracy Standards (NMAS) or better. In addition to their local usage, we feel these data provide a valuable resource that can be integrated for use in land use/ land cover mapping for the Bay watershed. The major limitation to usage of locally produced databases is the lack of uniformity in quality, scales, timeliness, and legends.

OBJECTIVES
CEGIS has already worked with several counties to produce low-cost but detailed land use data from their existing datasets. These local land use databases can enhance land cover data produced using currently available satellite systems. The goal of this effort was to select methodologies that would be acceptable to local governments and provide cost-effective methods to produce improved data for use in regional and Bay-wide efforts and also provide data useful for local-level purposes such as urban sprawl mapping, planning, and farmland protection. Specific objectives for this project were to: o Evaluate methods to produce local land use/land cover data. o Compare locally produced land use/land cover data to satellite derived data. Each objective will be discussed in detail in the following sections.

Objective #1: Evaluation of methods to produce local land use/land cover data.
Methodologies
County-Level Survey
In order to understand what local GIS resources might be available to help with the project, and to be a guide to future projects, a comprehensive survey of county level governments throughout the Chesapeake Bay Watershed was conducted. The 6

comprehensive survey was conducted between 2002 and 2004. A total of 178 counties were surveyed in the Bay. Some counties where less then 5 % of the land area of the county was in the Bay watershed were excluded from the survey. For example, in New York State, Ontario and Yates counties were not surveyed. The Pennsylvania survey included every county in the state regardless of whether or not it was in the Bay watershed. Other organizational partners assisted with the survey. In Maryland, the Geographic Information Services Division, Maryland Department of Natural Resources conducted the entire survey for Maryland Counties. In Pennsylvania, the Pennsylvania Department of Conservation and Natural Resources, the U.S. Census Bureau, and PAMAGIC assisted with the survey. The remaining counties in New York, Virginia, West Virginia, and Delaware were surveyed entirely by staff at Penn State. With the assistance of the partnering organizations, a survey design was completed to insure that sufficient information was collected about the use of geographic information systems to satisfy the data requirements of each organization. A complete data dictionary of the questions asked is contained in Appendix A. The survey was conducted by a mail survey to each individual county followed by phone calls, and in some cases, site visits to the respective counties. Survey responses were entered into an Access database. Due to the complexity of the survey it was divided into separate tables within an Access database. As shown in Appendix A, very detailed questions were asked of each county surveyed. Each state has two Access database tables that contain the responses as shown in Figure 1.

Figure 1: Breakdown of GIS Survey Access tables.

Selection of Pilot Counties


Several study areas were selected to test methodologies for this project. These areas were selected based on several criteria. One of the important criteria was that study sites had to be geographically distributed across the entire Chesapeake Bay Watershed and representative of the range of GIS capabilities found in county governments. Availability of locally-produced GIS data such as cadastral mapping was also an important consideration. The completed GIS survey was especially useful in determining which counties were selected. Finally, access to aerial imagery for the study area was also a factor in choosing the pilot county areas. The study areas that were selected are as follows: New York o Broome County o Steuben County o Tioga County Pennsylvania o Tioga County o Sullivan County o Clearfield County o Centre County o Mifflin County o Lancaster County o York County Maryland o Baltimore County Virginia o Henrico County

Land use/Land Cover Classification System


There are a large number of existing land use and land cover classification systems that were considered for use in this project. The American Planning Association (APA) has perhaps the most definitive list of land use/land cover classification schema on their web site. http://www.planning.org/lbcs/OtherStandards/

Several of these classification schemes were reviewed but it became apparent that only the LBCS and the Anderson classification scheme were being widely used by local, state, and federal government organizations. The other systems were outdated and not being used by many organizations, or were historical schemas no longer in use at all. The APA is promoting the Land Based Classification System (LBCS). To quote from their Executive Summary: LBCS provides a consistent model for classifying land uses based on their characteristics. The model extends the notion of classifying land uses by refining traditional categories into multiple dimensions, such as activities, functions, building types, site development character, and ownership constraints. Each dimension has its own set of categories and subcategories. These multiple dimensions allow users to have precise control over land-use classifications. The analysis of classification schemes conducted indicated this system to be far too complicated and unwieldy for the project purposes. The reasoning for this decision is as follows: 1. Maryland RESAC, NLCD, MRLC, and other remote sensed land cover systems all use some variation of the Anderson land cover classification system. 2. There was a need to be able to directly compare the projects developed land use/land cover datasets with these products. 3. The Anderson classification system is relatively easy to use in a production environment and is widely utilized by counties within the Chesapeake Bay Watershed. 4. The use of the LBCS would have introduced problems with translation and complicated the production of data. These reasons provided a compelling argument to use the Anderson system. Therefore, the Anderson classification scheme was adopted. The exact specifications of the Anderson classification that was used for this project is found in Appendix B. One area where the Anderson code was modified for use in this project was the introduction of the code 201 for residential grass and the code 401 for residential forest. These were for areas that were of a residential land use but were mostly either grass or forest land cover.

Land use/Land Cover Data Collection Methodologies


This project tested three primary methods for deriving land use/land cover data from locally produced GIS datasets. Aerial imagery interpretation Aerial imagery interpretation mixed with attribute information stored within planimetric GIS datasets

Derived land use/land cover from attribute information stored within planimetric GIS datasets

Method 1: Aerial Imagery Interpretation The primary mapping methodology used for developing the land use/land cover data was aerial imagery interpretation and on-screen digitizing of land use/land cover boundaries.. Utilizing ArcView 3.3 software in a production digitizing mode various aerial images were used as a base map. Using interpretive skills, operators determined what the land use/land cover was from observing the base aerial images, and either digitized polygons around specific land use/land covers or split polygons from an index base polygon file. Upon completion, each polygon was appropriately assigned an Anderson classification code. The imagery that was utilized for this project is indicated in Table 2. To aid in interpretation of what was observed on the aerial imagery, county cadastral data was utilized when available. By examining information such as property ownership, assessment land use code, and building information it was possible to increase the accuracy of the interpretive process. Several operators were involved in the digitizing process and were trained in the proper procedures to create the GIS datasets. During the initial training phases 100 percent of their work was checked for accuracy and any problems were brought to the operators attention in a Continuous Quality Improvement process. Operators were taken into the field so they could see first-hand what was actually on the ground so they could improve their interpretive skills. While time did not permit for 100 percent field verification of every dataset produced, spot checks were conducted. Clearfield County was 75 % field verified with the cooperation of the Clearfield County GIS Department The following quality control general procedures were utilized: The land use polygon dataset was viewed without using .avl (ArcView legend file which shows land use classification). This allowed the verifier to see the aerial imagery beneath the land use tile. Then the verifier scrolled through the entire tile zoomed to 1=4800 foot scale starting in NW corner, moving east until the NE corner, then moving down one level and moving back to the west. This scan was continued, evaluating every polygon for land use. During this scan, the verifier changed land use codes if a misinterpretation was observed and created new polygons, if required. Once the entire tile was finished (a bulk of the QC time), the .avl was applied so that the land use classifications could be observed on the screen. The verifier scrolled through the entire tile again and checked for oddities and neighboring polygons that have the same classification that should be combined, or further checked for accuracy. Once the second scan was completed, the verifier then checked the tile by using the attribute table. First all LU Codes were

10

Table 2: Aerial Imagery Sources State County New York Broome County Steuben County Tioga County Pennsylvania Tioga County Sullivan County Clearfield County Centre County Mifflin County York County Maryland Baltimore County Virginia Henrico County

Source New York State New York State New York State

Date 2002 2002 2002

Type Color/Panchromatic Color/Panchromatic Color/Panchromatic

Resolution 1/2 foot 1/2 foot 1/2 foot 2 foot/1 meter 1 Meter 2 foot 2 foot 2 foot 2 foot 2 foot 1 foot

USGS/County USGS County County County PA DCNR County County

1998 1996 1997 2002 1996 2003 2001 1998/2002

Panchromatic Panchromatic Panchromatic Panchromatic Panchromatic Color Color Color

checked for mistypes. Then all 300 (rangeland) level codes were checked for accuracy and to ensure that all strip mined areas were marked as such. Then 111 (single-family residential) were checked for codes for 0 or large R densities. Then all polygons were checked with small acreages. All polygons with less than 0.5 acres and odd polygons with less than 1.0 acres (forests, agricultural land, etc.) were then reviewed. Any significant errors were noted and the information was passed back to the operator as part of our continuous quality control improvement process. The full written procedure for quality control checking can be found in Appendix D. Time records were kept for various operators so an estimate of the total time necessary to digitize land use/land cover via this method would be recorded. The time impact of using different resolution aerial imagery was also obtained by keeping time records. Method 2: Aerial imagery interpretation mixed with attribute information stored within planimetric GIS datasets In some cases, such as Henrico County, Virginia, and Baltimore County, Maryland, cadastral data were used as a seed file to start the interpretive process. A seed file contains polygons derived from the cadastral GIS dataset that are coded based upon a county-assigned assessment land use code classification. Henrico County and Baltimore County had an advantage in that these counties have very good, spatially accurate parcel data that aligned well with the aerial imagery. It was expected that by using accurate

11

planimetric parcel based land use data it would be possible to reduce the time it took to digitize the land use from aerial imagery interpretation. It was determined that residential land use that occurs within a subdivision could be easily extracted. Residential coded parcels of less then 2 acres could also be easily extracted from the cadastral data. Although the parcel map did not include any of the road right-of-way information as a polygon feature it was still possible to use this null data to determine transportation land use. The procedure used in Henrico County will be discussed to explain the overall methodology. In order to expedite geoprocessing the county was split into four arbitrary zones (Figure 2). Each of these zones was derived by selecting tiles from the county tile index shapefile and then performing a combine operation to form a single polygon shapefile. The software used to create a master land use layer was ArcView 3.3.

Figure 2: Henrico County Zone Map

The next operation one of the zone files (the southern most zone was tried first) to clip the parcel shapefile. Once this operation was completed the clipped parcel shapefile was unioned with the corresponding zone shapefile. The primary reason for doing this was to create a polygon for the road right-of-way since this was null space within the parcel shapefile. Next the query builder was used to select all use description where the use was residential single family subdivision (Res Subd(1 family) and the acreage was less than or equal to 2 acres. The road right-of-way polygon was then added to the selection. At this point a new shapefile was created using the Convert Theme to Shapefile command. This file was named Residential_and_Roads.shp and consisted of 2,138 polygons (Figure 3). Excel was then used to fill out the Lu_code field to 111 for single family residential and a

12

1 was entered for the density figure. The road space, one single polygon, was coded 141 for transportation.

Figure 3: Henrico County Residential and Transportation

The query builder was then used again to select all use description where the use was NOT residential single family subdivision (Res Subd(1 family). Then the NOT road right-of-way polygon was added to the selection. At this point a new shapefile was created using the Convert Theme to Shapefile command (Figure 4). In order to simplify the Not_Residential_and_Not_Roads shapefile, this file was used to clip the zone file so that all the mixed use descriptions would be merged into single polygons. The resultant clipped file contained 244 polygons.

13

Figure 4: Henrico County Null Space

The next step was to Merge the two shapefiles together using the merge command in the Geoprocessing Wizard. The resultant merged file had 2,382 which equals the sum of the two input shapefiles indicating that no slivers polygons were created. From this merged shapefile it was then possible to edit the shapefile for those areas not already assigned a land use code. This exact process was used for the other southern zones. In the northern section of Henrico County the density of parcels did not lend itself to this procedure as the cadastral layer was too complex for ArcView 3.3 to handle efficiently. A different process was used for the northern section and will be discussed later. Once the merged files were created a tile index was derived from Henrico County original tile index. Each original Henrico tile covered 144 million square feet. To create the land use tiles 16 of the original tiles were put together and then combined to form one new tile that covered 2,304 million square feet. Then the new tile was used to clip the merged files created in the previous step. In this way it was possible to digitize one tile at a time. Each tile was numbered starting in the upper right corner. The diagram shown in Figure 5 indicates numbering scheme (note that the northern portion of the county is not shown)

14

32

1 45

Figure 5: Henrico County Tile Index Map

Upon completion of the individual tiles they were appended together and then imported into a personal GeoDatabase feature class. A topology validation process was then run to make sure the data was topologically correct. In the northern section of Henrico County the utilization of parcel data was expanded beyond just residential and included commercial, offices, and industrial sections 5 acres or less in size. The parking lot layer was intersected with the resulting cadastral file. The remaining non-classified polygons were identified through the use of aerial imagery interpretation. In areas where the imagery did not show parking lots (planimetric data were from 2004 and the imagery was from 1998), these areas were classified these areas as future development. In Baltimore County a similar methodology was used but since a tile grid from the county did not exist it was decided to just do one bigger polygon.

15

Derived land use/land cover from attribute information stored within planimetric GIS datasets Many county governments have excellent cadastral data and often the county assessment department will assign a land use code to each individual parcel. Typically these codes are assigned by a field operator when the property in question is being assessed for county tax purposes. Using GIS it is then possible to classify the parcels by the land use attribute. For many counties this is an easy way for them to determine their land use although there are some concerns with this approach which will be discussed later. In order to translate the county assigned codes to an Anderson based classification system is necessary to derive a translation table to convert the county assigned codes to the Anderson Code. For this project three counties in the state of New York were chosen due to the fact that New York has a uniform assessment code so the same translation could be applied to all three counties. The translation code table that was utilized can be found in Appendix E.

Results and Discussion


County-Level Survey
Out of the 178 counties surveyed, less then 2% failed to respond to the survey for a 98% survey completion rate. For some counties, with no GIS capability at all, the survey was very easy to complete. Contact information for key personnel in these counties was obtained for future reference. Figure 6 shows an example of a map produced with the data from the survey showing the numbers of dedicated GIS workers with the unit of government. The survey indicated that there is a direct relationship to the amount of GIS data available and the number of GIS workers so the map below is indicative of where GIS data can be found. When examining this map it is clear that some of the major metropolitan areas were not surveyed. For example, Washington D.C was not surveyed. Other major metropolitan areas such as Baltimore and Richmond were actually surveyed but are difficult to map. The problem lies in the use of the county FIPS code to join the survey database to the spatial feature of county boundaries. Cities do not use a county FIPS code so when the join was accomplished these data did not transfer over. Due to this reason cities were excluded from any further discussion of some of the results. Nearly 68% of the counties who responded indicated they would share their data with state and federal government agencies. This is deceiving to some degree as many counties do not have much data to share even if they wanted to share it. During the process of acquiring GIS data from these counties over 70 percent of those contacted released the data at no charge. Several counties did have policies that required them to collect reasonable costs for duplication of data. And a few counties with good data had

16

unreasonable costs so this data was not acquired. Only one county refused to sell or share their data when asked.

Figure 6: Number of Dedicated GIS Workers by County

Out of the 175 counties who responded, 54% indicated they had GIS cadastral data layers. 65 % of respondents indicated they used GIS for planning purposes while only 17% indicated they used GIS for environmental purposes. In regards to land use and land

17

cover GIS datasets, only 6 % of the respondents indicated they had developed their own land use/land cover that was not generated from the cadastral GIS attribute data. During the phone interview process, the use of high quality land use/land cover was discussed, and without exception, all of the counties who were using GIS for planning purposes indicated they would like to have access to good land use/land cover data. It became apparent during the survey that many organizations that had GIS staff and GIS data within the Chesapeake Bay Watershed were not being surveyed. This project survey was designed to focus strictly on county level government. Other local level governmental groups such as regional planning agencies, municipal authorities, cities, towns, townships, boroughs, and economic development agencies in many cases have significant staff and GIS data holdings. If a new survey is conducted for the Chesapeake Bay Watershed these groups need to be included. The only major stumbling block is there are a lot of these other organizations. In Pennsylvania alone there are over 2,600 municipal governments.

Locally Generated Land Use/Land Cover


One of the sub-objectives was to test various land use/land cover creation methodologies to gauge accuracy of the final product and production time. Time records were kept for the operators who did the production work for this project. Table 3 indicates the actual time it took to digitize three of our pilot areas. All three of these areas were digitized by the same highly qualified operator who had several years of digitizing experience. All three study areas used the digitizing method of starting with a polygon seed file and then cutting out polygons from the seed file. In York County the aerial imagery interpretation method was utilized. This method took on average 1.63 hours to digitize the land use for one square mile. In Baltimore County the mixed aerial/planimetric interpretation method was used. This method took on average 2.55 hours to digitize the land use for one square mile. This comparison shows that it did not benefit the production rate to try to use cadastral GIS data. In fact it took almost an hour longer per square mile. Upon further review, and discussion with the operator, it became apparent that while the accuracy of the cadastral data was excellent, the additional polygons added to the master polygon seed file caused additional digitizing work rather then reduce it. In the case of Baltimore County the additional polygons had a dramatic impact on the time it took ArcView to redraw the map after panning or zooming to a new location. These extra seconds add up to a significant amount of time. One of the key problems was the attempt to utilize the road right of way data. The created polygons in some cases covered the entire county as a single polygon. When the operator tried to do an edit and one of these polygons was involved it would require the software to verify the entire boundary of the road right of way polygon. In York County the actual digitizing went quicker even though more polygons were created by the operator. It is clear that the digitizing would have gone much faster if the road right-of-ways were not used. This would have created many island polygons within the seed file that could have been easily split off without the software having to verify boundary topology over

18

huge adjacent rod polygons. At the completion of the digitizing the road right-of-ways could have then been created by intersecting the land use layer with a county boundary polygon and then simply attribute the one large polygon that would be created through this process. Another time saver would have been to dissolve the small single-family residential lots (code 111) to reduce the total count of polygons. The difference in time needed to digitize if different resolution aerial images were used was also analyzed. This comparison revealed that when lower resolution aerial images were used the time was reduced to digitize the land use (Table 3). This method only took 0.73 hours to digitize one square mile of land use. A comparison of the differences in the resultant land use determination does not show any significant difference; especially when looking at Level 1 Anderson coding (see Table 3). The biggest overall change was in determining residential areas. This is because isolated residential areas were more difficult to determine at the lower resolution imagery. Table 5 shows more detailed breakdowns at the Anderson Level 2/3 for another quad study area in Tioga County, Pennsylvania.
Table 3: Time Study for Digitizing Land Use

______________________________________________________________________ Baltimore County 1:2400 digitizing scale, 2 foot pixel color aerials, cadastral seed of two acre residential and roads used. o 73 square miles o It took 186 hours o 2.55 hours per square mile o York County 1:2400 digitizing scale, 2 foot pixel color aerials, no cadastral seed used. o 11 tiles @ 43 square miles per tile. o Average time of 70 hours per tile o 1.63 hours per square mile Various USGS Quadrangles 1:4800 digitizing scale, one meter resolution black and white, no cadastral seed o 10 quads @ 55 square miles per quad o Average time of 39 hours per quad o 0.71 hours per square mile

19

Table 4: Tioga County , Pennsylvania. Comparison of land use acreage total differences between 1:2400 digitizing versus digitizing on one meter resolution 1:4800 DOQQ for Antrim Quad. Comparison of 1" = 200' scale base map versus DOQQ basemap - Antrim Quad

Lu_code 100 200 300 400 500 600 700

1:2400 2637 9605 5413 17790 231 31 172

1:4800 2275 9737 5265 17610 250 24 365

Difference 362 -132 148 180 -19 7 -193

% Change 13.7% -1.4% 2.7% 1.0% -8.2% 22.6% -112.2%

20

Table 5: Tioga County , Pennsylvania. Comparison of land use acreage total differences between 1:2400 digitizing (NAD83) versus digitizing on one meter resolution DOQQ(UTM) for Knoxville Quad. 1:2400 1:4800 Difference in LUCODE ACRES LUCODE_UTM ACRES Acres 100 16.3 100 13 3.2 110 148.6 110 143 5.2 111 527.0 111 537 -9.6 114 13.8 Code not used 0 13.8 120 0.9 Code not used 0 0.9 123 4.3 Code not used 0 4.3 130 36.0 130 17 19.2 141 461.7 141 210 252.0 150 36.0 150 63 -27.1 186 16.4 Code not used 0 16.4 193 113.4 193 109 4.4 210 141.1 Code not used 0 141.1 211 8790.5 211 9911 -1120.1 211a 552.6 211a 126 426.2 212 19.0 212 68 -49.0 213 205.8 213 274 -67.8 240 115.8 240 131 -14.8 310 2820.5 310 1931 889.4 320 629.5 320 817 -187.1 330 1553.8 330 2587 -1033.6 400 18810.0 400 18271 538.5 411 9.0 Code not used 93 -84.3 421 1.4 412 66 -65.1 423 29.1 Code not used 110 -81.3 500 74.1 500 31 43.6 511 133.6 511 59 75.0 600 65.3 600 10 55.6 720 149.6 720 0 149.6 999 91.8 999 0 91.8 35566.6 35576 -9.7

In northern Henrico County the parking lots were burned into the seed file. In hindsight, the use of the parking lot layer caused more problems then it was worth especially as the parking lot layer excluded all the small grassy areas one finds in a typical parking lot. This incredibly complicated the planimetric land use layer when the parking lot data was burned into the seed file as shown in the figure 7. This could possibly be resolved by running a topology validation with a cluster tolerance slightly larger then the distance across the largest island within the parking lot such that the voids closed and could then be classified as parking lots.

21

Figure 7: Northern Henrico County parking lots (in grey).

Table 5 indicates the total estimated time needed to digitize the entire Chesapeake Bay watershed by the methods tested. These estimates assume that for high-density urban areas a simple 100 land use code is assigned.
Table 6: Cost Estimates to Digitize Land Use for Entire Bay

___________________________________________________________________ Chesapeake Bay Watershed Total Area = 64,000 square miles Aerial/Planemetric Method: 163,200 hours @ $15 per hour labor rate $2,448,000 Aerial Image Only Method: 104,320 hours @ $15 per hour labor rate $1,564,800 Reduced Resolution Method: 45,440 hours @ $15 per hour labor rate $681,600 ___________________________________________________________________

It is very important to note that these costs are for first time data collection. Once a county has digital land use/land cover data, subsequent updating costs would be significantly reduced resulting in lower long-term maintenance costs.

22

Parcel Based Land Use Codes Many counties have high quality land use codes assigned to their cadastral GIS data. The problem is that these codes only describe the predominant land use for the entire parcel. Figure 8 shows a 187 acre parcel that is coded as agriculture land use. 1/3 of the parcel is clearly forest land cover while at least 1-2 acres are residential or farmstead. Therefore the parcel based land use code, alone, describes the use and not the actual land cover information needed by the Bay Program. If parcels are less then 2 acres in size, the land use code can provide an accurate assessment of the actual land use. Although not tested, it might be possible to use other available data layers such as a digital elevation model or soils data to make decisions about the true land use/land cover. In the example provided the forested areas exist on steep slope and poor soils.

Figure 8: Sample parcel from Mifflin County, PA.

Other Options: Combining County Level Data with Satellite-Derived Data

23

The use of satellite derived land cover can provide valuable information. In the case of the RESAC land cover dataset, RESAC utilized planimetric data in the decision tree used to classify land cover. They utilized road centerline data to help make a determination of urban and impervious land covers. It would be possible to also use local county level GIS data in the decision tree. In cases where the GIS parcel acreage is less then 2 acres, the land use code carried in the attribute table for the cadastral dataset could be used in the decision tree. Another possibility is to combine information found in County-level planimetric data with satellite-derived data to improve the classification. Figure 9 illustrates the improvement in satellite data classification if local parcel information were incorporated into the process. Residential parcels are highlighted in Figure 9 (left image) for a township in Centre County, Pa based on the land use code assigned by the county. RESAC land cover data for the same area appears on the right image. Note that most of the residential parcels are classified incorrectly as agriculture. This is because the average lot size is approximately 1-2 acres in size and the satellite imagery interprets the nonimpervious area as agriculture rather than residential, thereby overestimating agricultural lands. This type of residential density is very typical throughout developing areas of Pennsylvania, and incidentally, is contributing to our urban sprawl problems.

Figure 9. RESAC land cover and residential parcel boundaries.

Figure 10 illustrates RESAC classification of just the residential parcels for the area with the legend adjusted so that low, medium and high density residential classes are all blue. The figures clearly show that almost none of the parcels were classified correctly as residential and that over half were incorrectly classified as agricultural lands. The use of parcel data with land use codes combined with RESAC data could be used to adjust the 24

RESAC classification or better yet, it could be used in the classification process at early stages of the project.

Figure 10. RESAC classification of residential parcels Data Acquisition During the initial phases of the project the Land Analysis Lab supplied MarylandRESAC with planimetric datasets from two counties in Pennsylvania, Lancaster County and Centre County. These planimetric datasets included such features as parcel boundaries, building outlines, and edge of pavement. In addition detailed land use data layers were provided for both of these counties. Land use and parcel data from Mifflin County was also provided but Mifflin County did not have building outline or edge of pavement data layers. RESAC needed these datasets for ground-truth purposes. Subsequent to the initial data deliveries to be used as ground truth data the Land Analysis Lab has acquired many different additional data sets from counties throughout the Chesapeake Bay Watershed. A total of 42 counties supplied data to the project as shown in Table 7. This represents about 1/5 of the counties within the bay watershed. These counties were selected based upon the quality of data that they might have. This data quality was indicated from the survey that was conducted. All received data will be delivered to the CBP. The counties that supplied their data were all very reasonable in responding to requests for their data. In some cases data agreements had to be signed indicating their data would not be released to a third party. There were also cases where a small fee ($50 or less) had to be paid to cover their costs in reproducing the data onto CD-ROM or DVD. Not all 25

counties charged these fees. The vast majority of counties were agreeable to letting the CBP have access to the data as long as it was not released to third parties without prior consultation with the respective county. Therefore it is specifically requested that the Chesapeake Bay Program respect their wishes and only use this data for internal purposes.

26

Table 7: List of Counties That Supplied GIS Datasets

County Kent Baltimore Co Carroll Harford Howard Broome Chemung Madison Steuben Tioga Tompkins Adams Cameron Centre Chester Clearfield Columbia Cumberland Lancaster Mifflin Schuylkill

Stat e DE MD MD MD MD NY NY NY NY NY NY Pa PA PA PA PA PA PA PA PA PA

County Sullivan Tioga Wyoming York Accomack Albemarle Botetourt Clarke Culpeper Hanover Henrico James City King William Loudoun Northampton Rockingham Shenandoah Spotsylvania Warren York Hampshire

State PA PA PA PA VA VA VA VA VA VA VA VA VA VA VA VA VA VA VA VA WV

One notable exception to this policy was aerial imagery from the state of Virginia. The Virginia Geographic Information Network (VGIN) does not readily distribute their data. An agreement with Henrico County was signed for use of their aerial imagery since Henrico County viewed the Land Analysis Lab (LAL) as a subcontractor and LAL was producing a land use data layer for them as part of our overall project. This agreement prohibits LAL from giving the Chesapeake Bay Program these aerial images. Subsequent attempts to gain permission to use aerial images from other Virginia counties were unsuccessful. Further information can be found at their website at http://www.vgin.virginia.gov/. Only one county refuse to provide available data to the project. This was Arlington County, Virginia, and the reason given was related to the fact that it was Federal project. Apparently that county has had some issues with data sharing with the Federal government. Some of the other counties wanted more then $50 to release their data and these fees were not paid, so no data was collected.

27

It was noted that some counties where there are Federal lands, such as Harford County, Maryland (Aberdeen Proving Grounds), the county data sets do not cover the Federal areas. This was also noted in York County, Virginia, and for some datasets that were downloaded from the District of Columbia (D.C.). The DC data was not included since any organization can download this data directly. The DC data has eliminated areas around the White House, the Congressional Buildings, and the Naval Observatory. To access these datasets go to http://dcgis.dc.gov/. Appendix F contains a more detailed listing of the data that was received and some of the observations on some of the datasets. Specific counties were selected where their datasets were used to help validate the impervious surface data that was produced by Maryland RESAC. Appendix G contains specific details on some of the issues related to acquiring the data. This Appendix includes information on counties where data was received and also where attempts were made to receive data. It may prove useful to someone who might try to acquire updated GIS datasets in the future.

Objective #2: Comparison of locally produced land use/land cover data to satellite derived data.
Methodologies
Land Cover Comparisons
One of the primary utilizations of the land use datasets developed by the Land Analysis Lab (LAL) was for reference datasets to assess the accuracy of remotely sensed land cover datasets. Comparisons were made with Maryland RESAC, NLCD 2000, and a Pennsylvania land cover dataset compiled by the Penn State University's Office for Remote Sensing of Earth Resources under contract with the Pennsylvania Department of Environment Protection. The RESAC dataset covers the entire Chesapeake Bay Watershed. The NLCD 2000 data was only available for Baltimore County, York County, and Lancaster County during the period of our research project. The Penn State dataset is a Pennsylvania statewide land cover map generated from Enhanced Thematic Mapper satellite data and three other ancillary data sources. It is an update to the MRLC data layer produced for the state in 1992. Metadata on Maryland RESAC can be found be found at http://gis2.pasda.psu.edu/Pasda/UCI_Metadata/chesapeakebaylanduselandcover2000.htm Metadata for the NLCD 2000 program is also readily available and can be found at http://landcover.usgs.gov/index.asp

28

Metadata describing the Penn State land cover dataset can be found at http://www.pasda.psu.edu/documents.cgi/orser/psu-palulc_2000.xml The three websites referenced can also be used to download these datasets. All the LAL reference datasets will be posted at www.pasda.psu.edu at a later date and are being provided directly to the Chesapeake Bay program office. The RESAC data were compared to eight county reference areas across the bay watershed as follows; Henrico, Baltimore, York, Lancaster, Mifflin, Centre, Clearfield, and Sullivan. Key to this comparison was how the UMD codes were translated into Level I Anderson Codes. Table 8 indicates the translations used.

29

Table 8: UMD RESAC code translation to Anderson code.

RESAC LU Codes 1 3 4 5 8 10 11 12 15 17 18 20 21 22 25 26 30 35 36 37 38

RESAC Description Open water Low intensity development Med intensity development High intensity development Transportation Urban/residential deciduous tree Urban/residential evergreen tree Urban/residential mixed trees Urban/residential/recreational grass Extractive Barren Deciduous forests Evergreen forest Mixed (deciduous-evergreen) forest Pasture/hay Croplands "Natural" grass Deciduous wooded wetlands Evergreen wooded wetland Emergent (sedge-herb) wetland Mixed wetland

Anderson Level 1 LU/LC Codes 500 100 100 100 140 100 100 100 100 700 700 400 400 400 200 200 200 600 600 600 600

Anderson Level 1 Description Water Urban or Built-Up Land Urban or Built-Up Land Urban or Built-Up Land Transportation/Communication Urban or Built-Up Land Urban or Built-Up Land Urban or Built-Up Land Urban or Built-Up Land Barren Lands Barren Lands Forestland Forestland Forestland Agriculture Agriculture Agriculture Wetlands Wetlands Wetlands Wetlands

Only the 100 (urban), 140 (transportation), 200 (agriculture), and 400 (forested) codes were considered in the analysis. The 500 (water), 600 (wetlands), and 700 (barren) codes were not considered to be reliable and also do not amount to significant percentages of the total land area in each study area. Appendix H contains the general procedure that was used to compare the datasets. The end result was that a confusion matrix (error matrix) was created for each comparison showing differences between the locally-produced data and the satellite-derived data.

30

Impervious Surface Comparisons


In order to compare Maryland RESAC Impervious surface data some local county governments were identified that had good, high-resolution planimetric data that was created at the same time as the Landsat TM images were captured. The best match to this requirement was Harford County, Maryland. Harford County captured several impervious surface layers from an aerial imagery flight that was flown in the spring of 2000. These layers included road pavement, driveways, building structures, and parking lots. They did not capture sidewalks which would have increased the impervious surface percentages. An example of the data is shown in Figure 11.

Figure 11: Harford County image showing planimetric impervious surface over color aerial imagery.

In order to compare the Harford County data to the RESAC data the following procedure was utilized. 1. All the relevant Harford County data layers were combined using the Union command. 2. The resultant Harford County planimetric impervious data layer was then reprojected into the same UTM projection as the RESAC data. 3. The Harford County data were then converted into a 5 meter raster data layer and classified as 100 percent impervious where the cells were created. The remaining cells were reclassified as 0 percent impervious (Figure 12).

31

Figure 12: Resample of Harford County planimetric to 5 meter cell raster

4. The 5 meter raster layer was then aggregated to the 30 meter cell resolution taking the mean of all 5 meter cells within a 30 meter neighborhood to assign an impervious surface percentage for each 30 meter cell (Figure 13).

Figure 13: Aggregate of Harford County 5 meter to 30 meter impervious data layer.

5. A boundary file for the Harford County data was created that excluded the Aberdeen Proving Grounds as the Harford County data did not include this area.

32

6. This boundary file was then used to clip the RESAC data so that it matched the same area as the Harford County data. 7. Using Spatial Analyst, the difference command was used to create a new raster layer so that each cell showed the differences between the two data layers. 8. The two impervious raster data sets were analyzed by looking at the attribute table and determining the actual percent impervious area that each raster data layer was recorded. 9. Three areas within Harford County were selected to represent rural, suburban, and urban areas to analyze the differences in the impervious surfaces for varying land uses. The procedure used to analyze the total impervious area was the same as described already. These areas are shown in Figures 14-16.

Figure 14: Rural sample area

33

Figure 15: Suburban sample area

Figure 16: Urban sample area

34

Results and Discussion


Land Cover Comparisons
Firstly, comparisons between RESAC and LAL will be discussed. Henrico County Table 9 shows the confusion matrix for Henrico County locally-produced data compared to RESAC. Overall RESAC underestimated total urban by 228 hectares, or a 1.3 % difference. While the total reported areas (hectares) were very close, the producers accuracy and the users accuracy, as shown in the confusion matrix results, were not as good. Producer accuracy indicate the portion of land area within a given land cover category that were classified correctly. For example, the producers accuracy was 49.9% for urban. This indicates that 49% of the urban areas were correctly classified by RESAC as urban. Therefore, 51.1 % of the urban areas were misclassified by RESAC as something other than urban. User accuracy indicates the portion of cells classified by RESAC within a certain land cover category that are correctly classified. For example, the users accuracy for urban was 53.3 %. This indicates that 53.3% of the cells that were classified as urban by RESAC were actually urban. The remaining 46.7% of the urban cells are really not urban according to LAL mapping. When the producers accuracy is higher then the users accuracy, RESAC has overestimated the amount of land area for that category and proportionately underestimated other categories. This is observed when looking at the agriculture land cover category. The producer accuracy is 79.2 % but the user accuracy is 61.4 %. In this case, RESAC overestimated agriculture by 2356 hectares or a 5.1 % difference. As can be seen in the confusion matrix, there was significant misclassification found in most categories. In particular, it is apparent that many agricultural and forested areas were misclassified as some type of urban category. An overall accuracy (produced by dividing the number of correctly classified cells by the total number of cells) of 72.1% was found for this study area.

35

Table 9: Henrico County versus RESAC Confusion Matrix Henrico South 1998 - Anderson Level 1 All figures in Henrico Hectares County (Urban) 100 3160 261 1309 1602 6332 (Transportation) 140 1186 793 232 351 2562 (Agriculture) 200 387 153 4245 574 5359 Users Accuracy (UMD/UMD) 100 140 200 400 Difference -404 -1201 1550 55 (Forest) 400 1194 155 1123 13851 16322 Grand Total (RESAC) 5928 1361 6909 16377 30576

RESAC Anderson Level 1

(Urban)100 (Transportation)140 (Agriculture)200 (Forest)400 Grand Total (Henrico) Producers Accuracy (UMD/LAL) (Urban)100 (Transportation)140 (Agriculture)200 (Forest)400 Land Use Code (Urban)100 (Transportation)140 (Agriculture)200 (Forest)400

49.9% 30.9% 79.2% 84.9% RESAC Total 5928 1361 6909 16377

(Urban)100 (Transportation)140 (Agriculture)200 (Forest)400 LAL Total 6332 2562 5359 16322

53.3% 58.2% 61.4% 84.6%

Examples from Henrico County illustrate common errors found in the residential RESAC category. Figures 17-19 show LAL 10 meter data and RESAC data, respectively, centered on a relatively large residential subdivision. The RESAC 30 meter data misclassify large portions of the subdivision as agriculture. As shown in Figures there is clearly a large amount of residential or urban (Anderson level 100) that is not captured. But overall in the entire study area the difference in urban between RESAC and LAL is only 404 hectares The obvious conclusion is that RESAC must be over reporting urban in another area of the county. In other words, on a cell-by-cell basis, the RESAC data are less accurate than they are on summary data produced from very large areas. For estimates of land cover percentages on a county basis, the numbers may be accurate but for small-area assessments, the numbers may be quite erroneous.

36

Figure 17: Henrico County 10 meter LAL land cover data

Figure 18: Henrico County 30 meter RESAC land cover data

Since RESAC urban forest categories were reclassified as urban (100) in this analysis, this would result in higher urban areas. For example Figure 19 and 20 show an urban forest area simplified in this analysis as urban. If RESAC data were not reclassified so that urban forest was considered urban then RESAC would have significantly underestimated urban even more and differences would have been much larger then is evident after the reclassification (see Figure 21).

37

Figure 19: Henrico County 10 meter LAL land cover data.

Figure 20: Henrico County 30 meter RESAC land cover data.

38

Figure 21: RESAC land cover non-simplified.

Another factor in RESACs reporting of urban areas is found by analyzing some e highways. RESAC utilized existing road networks as part of their decision tree and apparently assumed that urban development occurs along these roads. As shown in Figure 22, this is not always the case.

Figure 22: RESAC simplified land cover showing major cloverleaf intersection with grassy areas inside cloverleaf classified as urban.

39

Baltimore County Table 10 shows the confusion matrix for digitized data for Baltimore County compared to RESAC. Overall RESAC underestimated total urban by 623 hectares, or a 2.8 % difference. While these total reported hectares are close, the producers accuracy and the users accuracy were also relatively low. For urban lands the producers accuracy was 41.7% while the users accuracy was 52.3% and many of the same issues that affected Henrico County appled to Baltimore County. Agriculture was slightly overestimated by 407 hectares, or a difference of 3.5 %. The producers accuracy was 82.6 % while the users accuracy was 80.6 %. As in Henrico County, there was significant misclassification of urban lands as agriculture and forest. Apparently, the satellite imagery and classification algorithms are not able to discern low density residential development in open or forested lands.

40

Table 10: Baltimore County versus RESAC Confusion Matrix. Baltimore County 2002-2004 - Anderson Level 1 All figures are in Baltimore Hectares County RESAC (Urban) 100 (Transportation) 140 (Agriculture) 200 (Forest) 400 Grand Total (Baltimore) (Urban) 100 1092 162 898 447 2598 (Transportation) 140 96 83 60 71 310 (Agriculture) 200 213 194 5074 822 6303 (Forest) 400 574 100 678 7032 8384 Grand Total (RESAC) 1975 539 6710 8372 17595

Producers Accuracy (UMD/LAL) (Urban)100 (Transportation)140 (Agriculture)200 (Forest)400 42.0% 26.8% 80.5% 83.9%

RESAC Anderson Level 1

Users Accuracy (UMD/UMD) 100 140 200 400 55.3% 15.4% 75.6% 84.0%

Land Use Code (Urban)100 (Transportation)140 (Agriculture)200 (Forest)400

RESAC Total 1975 539 6710 8372

LAL Total 2598 310 6303 8384

Difference -623 229 407 -12

Figures 23 and 24 show the LAL and RESAC data , respectively, for an area of Baltimore County along I-83. Note that the urban areas to the west of the highway look like buffered zones versus the actual urban pattern. Also note that to the east of the highway smaller urban areas are largely misclassified by RESAC

41

Figure 23: LAL land cover data for Baltimore County

Figure 24: RESAC simplified land cover data for Baltimore County

42

An evaluation of original, non-simplified RESAC data (Figure 25), revealed that the majority of the urban area was classified as either urban forest (dark green) or urban recreational grass (light green). Developed areas (yellow) were small.

Figure 25: RESAC land cover data showing urban classifications (forested lands and agriculture are shown in white).

Clearfield County Table 11 shows the confusion matrix resulting from the comparison of RESAC data and detailed data mapped for Clearfield County. Overall RESAC underestimated total urban by -448 hectares, or a 0.2 % difference. While these total reported hectares are close, the producers accuracy and the users accuracy were also relatively low on a cell-by-cell basis. The producers accuracy was 34.3% while the users accuracy was 35.3%. When compared to the more developed counties like Henrico and Baltimore there are more rural areas in Clearfield. As shown for Henrico and Baltimore RESAC data misses a lot of rural development and this is why the producer and user accuracy percentages are low. The difference is very small because of the buffering of urban areas similar to what was explained in Baltimore County. An example of this will be discussed for Clearfield as well. Agriculture was overestimated by 5257 hectares, or a difference of 1.9 %. The producers accuracy was 57.8 % while the users accuracy was 46.6 %. The primary

43

reason for the lower accuracy may be the large number of reclaimed strip mines in Clearfield County that will be discussed later. As in other counties, there was much confusion between urban, forest, and agriculture. Many agricultural and urban lands were mapped as forest. Many urban lands were mapped as forest and agriculture.

44

Table 11: Clearfield County versus RESAC Confusion Matrix. Clearfield County, PA 1997/2003 - Anderson Level 1 All figures are in Clearfield Hectares County RESAC (Urban) 100 (Transportation) 140 (Agriculture) 200 (Forest) 400 Grand Total (Clearfield) (Urban) 100 5414 804 2073 7508 15798 (Transportation) 140 172 131 22 684 1008 (Agriculture) 200 1012 316 12580 7856 21764 (Rangeland) 300 1462 298 8384 15447 25591 (Forest) 400 7290 1137 3962 205267 217657 Grand Total (RESAC) 15350 2686 27021 236761 281817

Producers Accuracy (UMD/LAL) (Urban) 100 34.3% (Transportation) 140 13.0% (Agriculture) 200 57.8% (Forest) 400 94.3% Land Use Code (Urban) 100 (Transportation) 140 (Agriculture) 200 (Forest) 400 RESAC Total 15350 2388 27021 221314 LAL Total 15798 1008 21764 217657

UMD 2000 Anderson Level 1

Users Accuracy (UMD/UMD) 100 140 200 400 35.3% 4.9% 46.6% 86.7%

Difference -448 1380 5257 3657

The Treasure Lake area of Clearfield County illustrates the overestimation of urban areas probably resulting from RESACs buffering of street data. Figure 26 shows the LAL land cover data for Treasure Lake. Large portions of Treasure Lake are not developed yet streets do exist. Figure 27 shows the RESAC data where one can clearly observe the effects of buffering the streets. Note that many small areas of urban development are missed in RESAC. It should also be noted that most of the urban shown in RESAC is urban forest.

45

Figure 26: LAL land cover for Treasure Lake area.

Figure 27: RESAC land cover for Treasure Lake area.

In regards to the accuracy of the RESAC agriculture, the main issue here may be the large number of reclaimed strip mines in Clearfield County. The reclaimed land is not used generally for agriculture and was classified as rangeland, or Anderson code 300. Figure 28 shows a sample area with a high concentration of reclaimed strip mines (red). Figure 29 shows the same area classified by RESAC as agriculture and forest.

46

Figure 28: LAL land cover showing reclaimed strip mine area.

Figure 29: RESAC land cover in reclaimed strip mine area.

47

NLCD 2000 Data The NLCD 2000 land cover data has some of the same issues as the RESAC data. The data underestimates urban/residential development, especially in rural areas, and the data overestimates agricultural areas. As with RESAC, NLCD is reasonably accurate in mapping woodlands, especially large contiguous woodland areas. The NLCD codes were translated as shown in Table 12.
Table 12: NLCD code translation to Anderson code.
Anderson Level 2 LU/LC Codes 500 100 100 100 100 700 400 400 400 200 200 600 600

NLCD 2001 LU Codes 11 21 22 23 24 31 41 42 43 81 82 90 95

NLCD 2001 Description

Anderson Level 1 Description

Open Water Developed, Open Space Developed, Low Intensity Developed, Medium Intensity Developed, High Intensity Barren Land (Rock/Sand/Clay) Deciduous Forest Evergreen Forest Mixed Forest Pasture/Hay Cultivated Crops Woody Wetlands Emergent Herbaceous Wetlands

Water Urban or Built-Up Urban or Built-Up Urban or Built-Up Urban or Built-Up Barren Lands Forestland Forestland Forestland Agriculture Agriculture Wetland Wetland

Table 13 shows the confusion matrix for LAL versus NLCD. For the Level 1 Anderson 100 urban built-up land there is a difference of 10577 hectares, or 9.1 %. NLCD underreported urban/residential development. The larger difference in the two datasets as compared to RESAC versus LAL is explained by the fact that NLCD did not have an urban forest category nor does it appear that NLCD used planimetric street centerlines in their decision tree. A comparison between Figure 28 and Figure 29 shows some visual examples of the differences between LAL and NLCD in regard to urban/residential development. For Level 1 Anderson 200 agriculture there is a difference of 22,537 hectares, or 19.5 %. NLCD over-reported agricultural lands. Much of this difference can be explained by the underreporting of rural residential land. This land was classified as agriculture instead. The confusion matrix illustrates significant mapping of developed lands as agriculture. This region contains much residential development at relatively low densities. Satellite imagery is typically unable to differentiate low-density residential development from agricultural lands. Figures 30 and 31 show a portion of the county where NLCD misclassifies residential lands (yellow) as agriculture.

48

Table 13: York County, PA, LAL versus NLCD Confusion Matrix. York County, PA 2002-2003 - Anderson Level 1 All figures in York hectares County NLCD 2000 Anderson Level 1 NLCD 2000 (Urban) 100 (Agriculture) 200 (Forest) 400 Grand Total (York) (Urban) 100 3901 11616 1648 17165 (Trans) 140 1013 1273 287 2572 (Agriculture) 200 1077 50610 1405 53092 (Rangeland) 300 305 5994 1040 7339 (Forest) 400 292 6137 29253 35682 Grand Total (NLCD) 6588 75629 33634 115850

Producers Accuracy (NLCD/LAL) (Urban) 100 22.7% (Agriculture) 200 95.3% (Forest) 400 82.0%

Users Accuracy (NLCD/NLCD) 100 200 400 59.2% 66.9% 87.0%

Land Use Code (Urban) 100 (Agriculture) 200 (Forest) 400

NRLC Total 6588 75629 33634

LAL Total 17165 53092 35682

Difference -10577 22537 -2048

49

Figure 30: LAL land cover for York County, PA.

Figure 31: NLCD land cover for York County, PA.

50

Agricultural Comparisons The reported agriculture categories by both RESAC and NLCD were also examined for hay/pasture versus row crops. In Lancaster County it was observed that there was a large amount of classified hay/pasture which did not correlate with agricultural conditions in Lancaster County as described by the Pa Agricultural Statistics Service. Tables 14 and 15 show RESAC, NLCD and Ag Statistics for Lancaster County. The data show striking differences in the relative proportions of cultivated and uncultivated croplands. In fact, the relative proportions are nearly opposite.

51

Table 14: UMD RESAC and NLCD 2000 Pasture and Cropland acreage Totals. UMD Classifications 25 Pasture/hay 26 Croplands Total NLCD 2001 Classifications 81 Pasture/Hay 82 Cultivated Crops Total Acres 258,923 84,832 343,756 286,968 111,819 398,787

Table 15: Pennsylvania Department of Agriculture 2002 Pasture and Cropland Acreage Totals. PA Ag Stats (Year 2002) Wheat Hay Barley Haylage/Silage Rye Oats Pasture Total Hay/Pasture Corn/Grain Corn/Silage Soybeans Tobacco Vegetables Total Row Crops Total Agriculture Acres 10,830 73,264 9,071 35,222 2,449 591 25,574 157,001 94,421 69,829 28,223 4,496 5,606 202,575 359,576

52

Impervious Surface Comparisons


The results of comparisons of impervious surface areas mapped using county-level data and satellite-derived products (RESAC) indicate that the RESAC impervious layer may significantly underestimate total impervious surface area. Table 16 shows data for a comparison for Harford County.

53

Table 16: Impervious Surface Comparison between RESAC and Harford County. Overall Impervious 3.8 % 6.0 % 379.0 Rural Impervious 0.5 % 2.3 % 7.8 Suburban Impervious 4.0 % 7.8 % 7.8 Urban Impervious 19.5 % 23.3 % 7.8

Data Source RESAC Harford County GIS Area (sq miles)

Notes 1. Aberdeen Proving Grounds not part of study area 2. Harford County Planimetric data dated from Spring of 2000 3. Resac data cbw2000imperv_v1.2

The RESAC data was most accurate in urban areas and least accurate in rural areas. Figure 32 shows an example of missed impervious areas. The black cells represent where RESAC indicated an impervious surface area greater then 0 while the colored areas indicate where impervious surfaces actually exist.

Figure 32: Impervious Surface areas not mapped by RESAC.

In urban areas, RESAC data often overestimate specific cells percent impervious while overall slightly underestimating the actual impervious. In Figure 33 the blue cells represent overestimates of impervious surface areas while the yellow/red colors represent underestimates. Figure 34 shows the Harford County planimetric impervious data overlay for comparison.

54

Figure 33: Impervious Surface overestimated urban areas.

Figure 34: Planimetric overlaid onto RESAC data shown in Figure 31.

55

Appendix A: Data Dictionary of GIS survey questions


Part I ID of the jurisdiction Survey Status Date Survey Completed Name of Data Recorder Entity Coverage (FIPS) Name of jurisdiction Name of State Preferred Salutation First Name Last Name Preferred Suffix Business Title Organization Name Department or Unit Building Name/ Mail Stop Street Address City Zip Code Telephone Number FAX Number E-mail Organization Web Page URL Name of Planning Director Telephone Number E-mail Name of 911 Director Telephone Number E-mail Name of Public Works Director Telephone Number E-mail Are you responding to this questionnaire for Entire County including most Municipalities? Are you responding to this questionnaire for Single County Government Agency? Are you responding to this questionnaire for Single Municipal Government Agency? Are you responding to this questionnaire for State Agency Only? Are you responding to this questionnaire for Entire County Government Only? Are you responding to this questionnaire for Entire Municipal Government Only? Are you responding to this questionnaire for Regional Government Agency? Are you responding to this questionnaire for Other? Please estimate the total number of people working in jurisdiction Please estimate the number of Dedicated GIS Operators in the jurisdiction Please estimate the total number of GIS users working in the jurisdiction Are the GIS functions Centrally Managed? Are the GIS functions Split Among Different Departments?

56

Does the Planning Department have GIS functions? Does the Public Works Department have GIS functions? Does the Environmental Department have GIS functions? Does the Health Department have GIS functions? Does the Fire Department have GIS functions? Does the Police Department have GIS functions? Does the Information Technology Department have GIS functions? Do other departments have GIS functions? (list) Are your GIS Operating Systems UNIX? Are your GIS Operating Systems Windows NT? Are your GIS Operating Systems Windows (Yr.)? Is your GIS Software ArcInfo? Is your GIS Software ArcView? Is your GIS Software GenaMap? Is your GIS Software Imagine? Is your GIS Software Intergraph? Is your GIS Software MapInfo? Is your GIS Software Smallworld? Is your GIS Software IDRISI? Is your GIS Software Maptitude? Is your GIS Software Other (list)? Is your CAD Software AutoCAD? Is your CAD Software TurboCAD? Is your CAD Software ArcCAD? Is your CAD Software Microstation? Is your CAD Software Other (list)? Is your RDBMS Software Oracle? Is your RDBMS Software Informix? Is your RDBMS Software SQL Server? Is your RDBMS Software MS Access? Is your RDBMS Software dBase? Is your RDBMS Software Other (list)? Do you collect data with at GPS receiver? If you do collect data with a GPS receiver, what grade receiver do you use? If you do collect data with a GPS receiver, are your readings differentially corrected? Is your Map Coordinate System VA State Plane North? Is your Map Coordinate System VA State Plane South? Is your Map Coordinate System Pennsylvania State Plane North? Is your Map Coordinate System Pennsylvania State Plane South? Is your Map Coordinate System West Virginia State Plane? Is your Map Coordinate System New York State Plane West? Is your Map Coordinate System New York State Plane Central? Is your Map Coordinate System New York State Plane East? Is your Map Coordinate System Delaware State Plane? Is your Map Coordinate System Maryland State Plane? Is your Map Coordinate System Geographic? Is your Map Coordinate System UTM? Is your Map Coordinate System Other (list)? Is your unit of measure US Survey Feet?

57

Is your unit of measure Meters? Is your unit of measure Other (list)? Is your horizontal datum NAD27? Is your horizontal datum NAD83? Is your horizontal datum Other (list)? Is your vertical datum NGVD29? Is your vertical datum NAVD88? Is your vertical datum Other (list)? Do you maintain a current list of available data? If yes, is it available through Hardcopy? If yes, is it available through Website? If yes, is it available through NSDI Clearinghouse? If yes, is it available through Other (list)? Website URL: Do you distribute data in multiple formats? Do you support ESRI Coverage File Formats for Vector data distribution? Do you support ESRI Shape File Formats for Vector data distribution? Do you support ESRI Export File Formats for Vector data distribution? Do you support Other (list) File Formats for Vector data distribution? Do you support GeoTiff File Formats for Raster data distribution? Do you support TIFF/TFW File Formats for Raster data distribution? Do you support Other (list) File Formats for Raster data distribution? Do you support 4mm Tape as distribution media? Do you support 8mm Tape as distribution media? Do you support CD-ROM as distribution media? Do you support 3.5" Floppy as distribution media? Do you support Zip Disk as distribution media? Do you support DVD as distribution media? Do you support Internet Download as distribution media? Do you support FTP as distribution media? Do you distribute data in the same coordinate system, datum and unit that you use internally? Do you normally copyright your data? Do you normally license your data? Do you normally charge for your data? Do you restrict redistribution of your data? Will you allow your data to be used by State and Federal agencies in the Floodplain Mapping Program? If yes, will you provide the data with full charges? If yes, will you provide the data with no charges? If yes, will you provide the data with reduced charges? Will you allow these agencies to redistribute these data? Will you allow these agencies to post these data to a related web application or Internet Map Server that does not allow the raw data to be downloaded? Will you allow your data to be used by the Pennsylvania State University and the Chesapeake Bay Program in the Land Use and Land Cover Program? If yes, will you provide the data with full charges? If yes, will you provide the data with no charges? If yes, will you provide the data with reduced charges? Will you allow these agencies to redistribute these data?

58

Will you allow these agencies to post these data to a related web application or Internet Map Server that does not allow the raw data to be downloaded? Will you allow your data to be used by State and Federal agencies for Homeland Security Purposes? If yes, will you provide the data with full charges? If yes, will you provide the data with no charges? If yes, will you provide the data with reduced charges? Will you allow these agencies to redistribute these data? Will you allow these agencies to post these data to a related web application or Internet Map Server that does not allow the raw data to be downloaded? Will you allow your data to be used by State and Federal agencies to improve the MAF/TIGER data produced by the US Census Bureau? If yes, will you provide the data with full charges? If yes, will you provide the data with no charges? If yes, will you provide the data with reduced charges? Will you allow these agencies to redistribute these data? Will you allow these agencies to post these data to a related web application or Internet Map Server that does not allow the raw data to be downloaded? Can these agencies obtain a copy of your price schedules by hardcopy? Can these agencies obtain a copy of your price schedules by mail? Can these agencies obtain a copy of your price schedules by fax? Is it ok to list your contact information and data in a statewide GIS resource guide? Part II What was the population of your jurisdiction in the year 2000? Does your jurisdiction link building permit data to the parcel layer? Does your jurisdiction assign a land use code to the parcel layer? Does your jurisdiction assign a zoning code to the parcel layer? Does your jurisdiction produce CAMA data and link it to the parcels? If yes, what software do you use? Are your parcel boundaries mapped with COGO? Are your parcel boundaries mapped by aligning with orthophotos? Are your parcel boundaries mapped in another way? How many parcels are in your jurisdiction? How many parcel "splits" occur in your jurisdiction annually? How many months does it take on average to add the splits to your system? Do you maintain parcel information on publicly owned land? How confident are you in percentage of the spatial accuracy of your parcel boundaries? How confident are you in percentage of the spatial accuracy of your parcel attribute data? What is the approximate ground resolution per pixel of your orthophotos? While film emulsion was used for your orthophotography? Was this photography flown with airborne GPS and IMU? How are your data distributed? What is the best source for road locations in your jurisdiction? What is the best source for road names in your jurisdiction? Are road names given in that data?

59

Are route numbers given in that data? Are road types given in that data? Are road descriptions given in that data? Are road directions given in that data? Are street addresses given in that data? Are address ranges given in that data? Are zip codes given in that data? How were road locations generated? Do you use a linear referencing system? Has your jurisdiction been readdressed for enhanced 9-1-1? If no, do you have plans to readdress in the next two years? If your jurisdiction produces land cover information, how is it generated? If it is produced by Image Processing/ Remote Sensing , please list the exact source. What Classification System do you use? Did your jurisdiction create its own Hydrography layer? How were these data created? What features do you have? What post spacing was used in your DEMs? What contour interval does your data support? Do you have breaklines? Do you have control monuments set by the county? If you have control monuments are they established by GPS? What is the date range for when your monuments were set? Do you document your accuracy assessment methodologies? Do you keep your accuracy assessment studies? Does your jurisdiction currently provide Internet Mapping Access to its data? If you do not currently provide Internet Mapping Access to your data, do you plan to provide it? If you plan to provide it, what is your approximate time frame? What Internet Mapping Software do you use, or anticipate on using? What is the name of your regional planning agency? What is the name of a contact person at that agency? What is the phone number of this contact person?

60

Appendix B: Anderson Land Use/Land Cover Classification Code


The following is the version of the Anderson Land Use/Land Cover classification adopted by Penn State University (Land Analysis Lab) for the Chesapeake Bay Basin Land Cover/Land Use Monitoring and Assessment Study. The original Anderson code can be found in this published paper: A Land Use And Land Cover Classification System For Use With Remote Sensor Data By JAMES R. ANDERSON, ERNEST E. HARDY, JOHN T. ROACH, and RICHARD E. WITMER Geological Survey Professional Paper 964 A revision of the land use classification system as presented in U.S. Geological Survey Circular 671 A copy is available on the Internet in an Adobe PDF file format. http://landcover.usgs.gov/pdf/anderson.pdf The Anderson code that was utilized was derived from the scheme that was used by the Delaware Office of State Planning Coordination for their 1997 Land Use/Land Cover project. This same source was used to develop the initial Pennsylvania State GIS standard for land use and land cover. 100 Urban or Built-Up Land 110 Residential High to low density urban residential and linear development along roads and shorelines. Also includes residential areas on military bases and school grounds, when larger than minimum mapping unit. Includes associated structures, driveways and lawn areas. 111 112 113 114 119 Single Family Dwellings Multi-Family Dwellings Includes duplexes, 4-plexes, apartment buildings and high rises. Group Quarters Includes nursing homes, barracks and dormitories. Mobile Home Parks/Courts Other Residential

120 Commercial Areas used predominantly for the sale of products and services. 121 Retail Sales/Wholesale/Professional Services

61

122 123 125 129

Includes stores, shopping strips and malls, motels, restaurants, office buildings and associated driveways, sheds, landscaped areas and parking lots under the minimum mapping unit. Vehicle Related Activities Includes garages, vehicle sales, gas stations and auto paint and body shops. Junk/Salvage Yards Includes auto junkyards, scrap metal salvage operations, and related activities. Warehouses and Temporary Storage Other Commercial

130 Industrial 131 Primary Processing at Extraction Site and Extraction Equipment Includes processing facilities, loading devices, access roads, stockpiles, storage sheds and parking lots found at resource extraction sites, including oil, gas, ore, coal, timber. Light Industry Focused on design, assembly, finishing, processing and packaging of products. Generally have few to no lagoons, stockpiles of raw materials or waste disposal areas. Heavy Industry Uses raw materials. Stockpiles, lagoons, waste disposal areas, transportation facilities for bulk products are often present. C - Chemical Related - plastics, pharmaceuticals, etc. P - Petroleum Related - refining M - Metals - refining, shaping W - Wood Processing - pulp and paper, sawmills, wood treatment B - Brick, Concrete, Block Electrical Generation N - Nuclear 0 - Oil G - Gas C - Coal H - Hydro-Electric Tank Farms and Materials Storage Includes tank farms, stockpiles and bins of materials away from industrial facilities and/or larger than the minimum mapping unit. Industrial Waste Treatment/Recycling, Facilities Inactive/Dismantled Industrial Site Other Industrial

132

133

134

135 136 137 139

140 Transportation/Communication 141 Highways/Roads/Access Roads/Freeways/Interstates

62

142 143

144 145 146 149

Parking Lots Only those parking lots that exceed the minimum mapping unit. Railroads X - Dismantled Railroads Y - Rail yards, Repair Facilities Includes stations, parking lots, and repair and switching yards. Airports Includes runways, intervening land, terminals, service buildings, navigational aids, fuel storage, parking lots and a limited buffer zone. Communications - antennas Marinas/Port Facilities/Docks Other Transportation/Communication

150 Utilities Facilities involved in treatment, transportation, and delivery of water, gas, oil, and electricity. 151 152 153 186 187 188 189 Water Treatment Does not include reservoirs. Sewage Treatment Plants Power Substations Religious Municipal Services Includes police and fire stations. Road repair and Maintenance Yard Other Institutional/Governmental

190 Recreational 191 192 193 194 195 199 Large Single Structure Includes stadiums and race tracks. Multiple Structure Includes fairgrounds, amusement parks, zoos, and piers. Open Urban Includes parks, playgrounds, golf courses, athletic fields, drive-in theaters and swimming pools. Open Rural Includes ski areas and camp grounds. Public Beaches Includes only those beaches with associated boardwalks. Other Recreational

63

200 Agriculture 201 Residential Grass 210 Cropland and Pasture 211 Cropland Includes those fields which are previously being maintained as crop producing areas and have vegetation cover at the time of imagery acquisition. A Row Crops - Includes those fields with vegetation cover that show distinct striations. B Cover Crops - Includes those obviously tended fields with vegetation cover, but no striations. C Bog Crops - Includes cranberries. Pasture Includes onlv those fields or areas in which livestock or distinct signs of livestock grazing can be detected, such as trails, watering tanks, corrals, salt licks and/or supplemental feeding stations. Idle Fields Includes areas that were previously cultivated in the past, but do not show signs of current cultivation or grazing. If shrubs and trees are present, use Category 32. Plowed, Un-vegetated Fields Truck Crops Areas of intense cultivation with mixed crops in relatively small fields. Generally in agricultural areas and commercially oriented. Gardens Includes home and community gardens.

212

213

214 215 216

220 Orchards/Nurseries/Horticulture 221 222 223 224 225 Fruit and Nut Trees Fruit Bushes Vineyards Greenhouses Nurseries/Floriculture

230 Confined Feeding Operations/Feedlots/Holding Areas 231 232 233 234 235 239 Cattle Poultry Hogs Horses Sheep/Goats Other

64

240 Farmsteads and Farm Related Buildings Farm ponds are in Category 536. 290 Other Agricultural 291 Fish Hatcheries

300 Rangeland 310 Herbaceous Rangeland 320 Shrub-Brush Rangeland 330 Mixed Rangeland 400 Forestland 401 Residential Forest 410 Deciduous Forest 411 412 413 Deciduous Forest Logged - Partial/Select Cut Recently Replanted - Deciduous

420 Evergreen Forest 421 422 423 Evergreen Forest Logged - Partial/Select Cut Recently Replanted - Evergreen

430 Mixed Forest 431 432 433 Mixed Forest Logged - Partial/Select Cut Recently Replanted - Mixed

440 Clear-Cut Includes those areas obviously related to commercial timbering operations where all trees have been removed. 441 442 Barren Clear-cut Ground surface totally denuded of vegetation cover. Non-barren Clear-cut All trees removed; shrubs, forbs and/or grasses still present.

65

500 Water 510 Waterways, Streams, and Canals Includes rivers, creeks and other linear water bodies. The boundary between streams and other bodies of water is the straight line across the mouth of the stream, up to a width of 1.85 km. Where a watercourse is interrupted by a water control structure, the impounded area will be placed in the Reservoirs (531) Category 511 512 Rivers/Streams/Creeks/Bayous Canals

520 Natural Lakes and Ponds Unless collateral data is available, no distinction can be made between fresh and salt water lakes. Salt ponds are located near the sea coast, inland of a barrier dune or dunes. Salt lakes are formed by evaporation in basins of interior drainage. 521 522 523 Freshwater Lakes and Ponds Salt Lakes Salt Ponds

530 Man-Made Reservoirs and Impoundments 531 532 Reservoirs Includes any man-made impoundment, lake or pond that does not fall into one of the following categories. Mined Areas Reservoirs Includes only water captured by mined depressions and/or water present because the bottom of the excavation is lower than the groundwater level. Lagoons/Basins Includes any man-made impoundment used specifically for waste liquids and/or the cooling of liquids. Drainage Catchment Basin Includes acid mine and storm water runoff catchments. Trailing Ponds Revetted areas that receive a slurry of crushed or powdered rock and water. Associated with primary ore processing. Farm Ponds

533 534 535 536

540 Bays and Coves 590 Other 591 Open Ocean

66

600 Wetlands 610 Forested Wetlands 611 612 613 Deciduous Evergreen Mixed

620 Non-Forested Wetlands 700 Barren Lands 710 Salt Flats 720 Beaches and River Banks 721 722 723 Beaches Riverbanks/Flood Plains Mud Flats

730 Inland Natural Sandy Areas 731 732 Dunes Dry Water Courses

740 Bare Exposed Rock 750 Extraction 751 751 752 752 753 753A 754 754A 755 759 Strip Mines Abandoned StripMines Stone Quarries Abandoned Quarries Sand and Gravel Abandoned Sand/Gravel Non-Strip Mined Minerals (Open Pit Mining) Abandoned Non-strip Mined Minerals Slag/Spoil Piles Other

760 Transitional May include cleared, filled, and graded areas, and new road networks. 761 762 Future Residential Future Commercial/Institutional

67

763 764 765 766 768 769

Future Industrial Future Transportation/Communication Future Water Project Future Utilities Unidentifiable Other Transitional

68

Appendix C: New York State Assessment codes translation to Anderson Code


NYS Property Class Code none 0 100 Direct Translation Code to Anderson

New York State Land Use Description

AGRICULTURAL Agricultural land (productive) w/o living structure Livestock and products Poultry and poultry products Dairy products: milk, butter, cheese Cattle, calves, hogs Sheep, wool Honey and beeswax Other livestock: donkeys, goats Horse farms Field crops Acquired Development Rights Truck crops - muck lands Truck crops - not muck lands Orchard Crops Apples, pears, peaches, cherries, etc. Vineyards Other fruits Nursery and greenhouse Specialty farms Fur products: mink, chinchilla, etc. Pheasant, etc. Aquatic: oysterlands, fish and aquatic plants Livestock: deer, moose, llamas, buffalo, etc. Fish, Fame and Wildlife Preserves RESIDENTIAL One family year-round residence

200

105 110 111 112 113 114 115 116 117 120 129 130 140 150 151 152 160 170 180 181 182 183 184 190 200 210

210 230 232 210 210 210 200 210 210 211 999 211 211 220 221 223 220 220 200 200 200 291 210 200 110 111, 1

69

220 230 240

Two family year-round residence Three family year-round residence Rural residence with acreage Rural residence with 10 or more acres (residential and agricultural) Rural residence with 10 or more acres (recreational) Estate Seasonal residences Mobil home Multiple mobile homes Residential - multi-purpose/multistructure Multiple residences Residence with incidental commercial use VACANT LAND Residential Residential vacant land (no house, but near one?) Residential land including small improvement (not for living) Waterfront vacant lots Rural vacant lots of 10 acres or less Underwater Vacant Land

112, 2 112, 3 111

241

111

242 250 260 270 271 280 281 283 300 310 311

110 111 119 110 114 100 110 100 300 300 300

312 313 314 315

300 300 300 500

316 320 321 322 323 330 331 340 341

Waterfront vacant land including a small improvement (not for living) Rural Abandoned agricultural land (nonproductive) Residential vacant land over 10 acres Other rural vacant lands Vacant land located in commercial areas Commercial vacant land with minor improvements Vacant land located in industrial areas Industrial vacant land with minor improvements

100 400 213 300 300 343 120 343 130

70

350 380 400 410 411 414 415 416 417

Urban renewal or slum clearance Public utility vacant land COMMERCIAL Living accommodations Apartments Hotel Motel Mobile home parks Camps, cottages, bungalows Inns, lodges, boarding and rooming houses, tourist homes, fraternity and sorority houses Dining establishments Restaurants Diners and luncheonettes Snack bars, drive-inns, ice cream bars Night clubs Bar Fast food franchises Motor vehicle services Auto deals - sales and service Service and gas stations Auto body, tire shops, other related auto sales Automatic car wash Manual car wash Self-service car wash Parking garage Parking lot Small parking garage Storage, warehouse, and distribution Facilities Gasoline, fuel, oil, liquid petroleum storage and or distribution Bottled gas, natural gas facilities Grain and feed elevators, mixers, sales outlets Lumber yards, sawmills Coal yards, bins Cold Storage facilities Trucking terminals Piers, wharves, docks and related facilities

760 100 120 110 112 121 121 114 119

418 420 421 422 423 424 425 426 430 431 432 433 434 435 436 437 438 439 440

121 121 121 121 121 121 121 121 120 122 122 122 120 120 120 142 142 142 120

441 442 443 444 445 446 447 448

120 120 121 130 130 125 125 146

71

449 450 451 452 453 454 455 460 461 462 463 464 465 470 471 472 473 474 475 480 481 482 483 484 485 486 500 510 511 512 513 514 515 520

Other storage, warehouse and distribution facilities Retail services Regional shopping centers Area or neighborhood shopping centers Large retail outlets Large retail food stores Dealerships - sales and service (other than auto) Banks and office buildings Standard bank/ single occupant Drive in branch bank Bank complex with office building Office building Professional building Miscellaneous services Funeral home Dog Kennels, Veterinary Clinics Greenhouses Billboards Junkyards Multiple use or multipurpose Downtown row type (with common wall) Downtown row type (detached) Converted residence One story small structure One story small structure - multi occupant Minimart RECREATIONAL AND ENTERTAINMENT Entertainment Assembly Legitimate theater Motion picture theater Drive-in theaters Auditoriums, exhibition and exposition halls Radio, TV, and motion picture studios Sports assembly

125 121 121 121 121 121 121 121 121 121 121 121 121 100 121 121 121 100 123 100 100 100 100 100 100 121 190 190 190 121 193 100 121 190

72

521 522 530 531 532 533 534 540 541 542 543 544 545 546 550 551 552 553 554 555 556 557 560 570 580 581 582 583 590 591 592 593 600 610 611 612 613 614 615 620 630 631 632 633

Stadiums, arenas, armories, field houses Racetracks Amusement Facilities Fairgrounds Amusement Facilities Game farms Social Organizations Indoor sports facilities Bowling centers Ice or roller skating rinks YWCA, YMCA, etc. Health Spas Indoor swimming pools Other indoor sports Outdoor spots activities Ski centers Public golf courses Private golf courses Outdoor swimming pools Riding stables Ice or roller skating rinks (may be covered) Other outdoor sports Improved beaches Marinas Camps, camping facilities and resorts Camps Camping facilities Resort Complexes Parks Playgrounds Athletic fields Picnic grounds COMMUNITY SERVICES Education Libraries Schools Special schools and institutions Colleges and universities Other educational facilities Religious Welfare Orphanages Benevolent and moral associations Homes for the aged

191 191 190 192 192 190 189 190 121 121 190 121 190 190 190 194 193 193 190 190 190 190 195 146 190 190 190 190 190 493 193 190 100 189 189 189s 189s 189s 189s 186 189 189 189 189

73

640 641 642 650 651 652 653 660 661 662 670 680 681 682 690 691

Health Hospitals All other health facilities Government Highway garage Office building Parking lots Protection Army, navy, air force, marine and coast guard Police and fire protection, electrical signal Correctional Cultural and Recreational Cultural facilities Recreational facilities Miscellaneous Professional associations Roads, streets, highways and parkways, express or otherwise (including adjoining land) Indian reservations Animal Welfare Shelters Cemeteries INDUSTRIAL Manufacturing and processing Mining and quarrying Sand and Gravel Limestone Trap rock Salt Iron and Titanium Talc Lead and zinc Gypsum Other Wells Oil- natural flow (for production) Oil- forced flow (for production) Gas (for production) Junk Water used for oil production Gas or oil storage wells Industrial product pipelines

189 189 189 189 188 189 142 189 189 187 189 190 190 190 100 189

692 693 694 695 700 710 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 740

141 100 121 186c 130 130 750 753 750 750 750 750 750 750 750 750 130 130 130 130 130 130 130 130

74

741 742 743 744 749 800 810 820 821 822 823 826 827 830 831 832 833

Gas Water Brine Petroleum products Other PUBLIC SERVICES Electric and gas Water Flood control Water supply Water treatment facilities Water transmission - improvements Water transmission - outside plant Communication Telephone Telegraph Radio Television other than community antenna television Community antenna television Telecommunications Cellular telephone towers Transportation Motor Vehicle Ceiling railroad Nonceiling railroad Air Water Bridges, tunnels, and Subways Pipelines Waste disposal Solid wastes Landfills and dumps Sewage treatment and water pollution control Air pollution control Special franchise property Electric and gas Water Telephone Miscellaneous Pipelines Television Electric and gas

130 130 130 130 130 150 150 150 150 400 151 150 150 140 140 140 140

834 835 836 837 840 841 842 843 844 845 846 847 850 851 852 853 854 860 861 862 866 867 868 869 870

140 145 140 145 140 100 143 143 150 150 140 150 150 150 150 150 150 100 150 150 140 100 150 140 150

75

871 872 873 874 875 876 877 880 882 883 884 885

Electric and gas facilities Electric substation Gas measuring and regulation station Electric power generation facility - Hydro Electric power generation facility - Fossil fuel Electric power generation facility Nuclear Electric power generation facility - Other fuel Electric and gas transmission facilities Electric transmission Improvement Gas transmission improvement Electric distribution - outside plant property Gas distribution - outside plant property WILD, FORESTED, CONSERVATION LANDS AND PUBLIC PARKS Private wild and forest lands except for private hunting and fishing Forest land under section 480 of the real property tax law Forest land under section 480a of the real property tax law Private hunting and fishing clubs State owned forest lands

150 153 150 134h 134 134n 134 150 150 150 150 150

900

400

910

400

911

400

912 920 930

400 400 400

931

State owned land (forest preserve) in the Adirondack or Catskill Parks Taxable under section 532a of the real property tax law

400

932

State owned land other than forest preserve covered under section 532 b, a, d, e, f, or g of the real property tax law

400

76

940

Reforested land and other related conservation purposes State owned reforestation land taxable under sections 534 and 536 of the real property tax law County owned reforested land Hudson river and black river regulating district land Public parks State owned public parks, recreation areas, and other multiple uses County owned public parks and recreation areas City/town/village public parks and recreation areas Other wild or conservation lands Wetlands, either privately or governmentally owned, subject to specific restrictions as to use Land under water, either privately or governmentally owned Taxable state owned conservation easements Other taxable state land assessments Adirondack park aggregate additional assessments

400

941 942

400 400

950 960

400 400

961

400

962

400

963 970

400 400

971

600

972 980 990

500 400 400

991

400

992

Hudson river-black river regulating district aggregate additional assessments Transitional assessments for taxable state owned land Transitional assessments for exempt state owned land

400

993

400

994

400

77

Appendix D: QC Procedure for Land Use Digitizing Product Quality Control


Setup: Operate at a scale of 1:4800, except when verifying any Urban or Built Up Land, which should be verified at a scale of 1:2400. Approaches: Database Driven and Map Driven QC. It is necessary to use two approaches to accomplish a complete QC Check. Those approaches are a Database Driven and Map Driven QC. Database Driven QC The Database Driven QC is performed by checking, verifying, and editing information in the database. Such information includes polygon area and polygon land use classification. The advantages to Database Driven QC are that one can correct small polygons that may have been missed in a Map Driven approach, verify large polygons, which often constitute a large area of the tile, find and verify unusual land use classifications, specifically classify present level 1 codes, and first time classify 999 codes. The disadvantages to Database Driven QC are that one may miss errors in which polygon size and land use classification seem normal, where the polygon is classified incorrectly. For example, if many normal sized (.5- 500 acre) commercial polygons are labeled residential, this will be missed in the Database Driven approach. Map Driven QC The Map Driven QC is performed by checking, verifying, and editing information on the map. Such information includes land use classification based on relationship of spatial features to each other and land use classification based on additional map data (such as other GIS layers). The advantages to Map Driven QC are that one obtains an accurate picture of the spatial relationships between a given polygon and adjacent areas and one forms an accurate picture of a polygons true land use classification by referring to additional data sources, such as aerial photography, building point data, parcel data, etc. The disadvantages of Map Driven QC are that you will not easily be able to locate small or sliver polygons for correction. Also, sometimes one does not realize the extent of large polygons (500 acres +) because the polygon is too large to completely fit on the screen at a given scale. Also, it is often difficult to locate the unusual land use classifications for verification.

78

I. Procedure for Performing Database Driven QC 1. Recalculate Area 2. Recalculate Perimeter 3. Recalculate Acres 4. Correct Small Polygons. Identify all polygons less than .5 acres. Correct all of these polygons that are non urban/ built up land. This is often done by performing a union with appropriate adjacent polygon. 5. Verify Large Polygons. Verify any polygons greater than 500 acres. 6. Classify 999 Codes. Look at all 999 codes and attempt to classify if possible. 7. Checking Urban & Built Up Land A. Look at any 100 codes and classify more specifically if possible B. Give R_Density to any 111 codes missing this. C. Verify any R_Density greater than 15. D. Scroll through and verify any uncommon codes- codes that are not 111, 120, 130, or 141. 8. Checking Agricultural Land A. Look at all 200 codes and classify more specifically if possible. B. Verify any uncommon codes- codes that are not 210, 211, 212 or 240. 9. Checking Rangeland A. Look at all 300 codes and classify more specifically if possible. B. Scroll through this section of database. If a polygon appears unusual, verify that polygon. 10. Checking Forestland A. Look at all 400 codes and classify more specifically if possible. B. Verify any uncommon codes- codes that are not 411, 421, or 431. C. Scroll through this section of database. If a polygon appears unusual, verify that polygon. 11. Checking Water A. Look at all 500 codes and classify more specifically if possible. B. There are usually so few of these, for QC it is best to look at each polygon.

79

12. Checking Wetlands A. Look at all 600 codes and classify more specifically if possible. B. There are usually so few of these, for QC it is best to look at each polygon. 13. Checking Barren Lands C. Check all Barren Lands codes in database. A. Look at all 700 codes and classify more specifically if possible. B. There are usually so few of these, for QC it is best to look at each polygon. II. Procedure for Performing Map Driven QC A. Record the total amount of time it has taken you to do Database Driven QC. Divide that time by six and spend at least that much time doing a Map Driven QC. This means that you will spend at least 10 minutes doing a Map Driven QC for every hour spent doing QC through the Database Driven approach. B. Scroll through the map getting an overall picture of this area and the spatial relationships between features on the map. Correct or comment on any polygons that seem out of place.

80

Appendix E: Procedure for creating confusion matrix to compare photo interpretation land use and satellite derived land use.
The general procedure used to perform this comparison is as follows: Step 1. It was necessary to convert the planimetric land use data into a raster format. A. Combine all tiles within a county using Arc toolbox tool Append: Create a new folder under the file folder specified of the county being compared. This folder was titled Level_1_compare. This folder was then used for all files created in this process. Create a copy of one of the tiles so that it can be used to append all other tiles to it. Save it in the Level_1_Compare folder. Ex: Make copy of Barbours_040217_5e and rename it Sullivan_all Using the Append tool, enter the following: Input Features: add all tiles excluding original tile that you renamed (i.e. not Barbours) Output Features: Sullivan_all Schema Type: Test Note: all input features (tiles) must be in same folder, or they must be appended separately. If files are large or there are many, you may have to append only a few tiles at a time. B. Simplify classification system to Anderson level 1: In ArcView, start editing Sullivan_all. Open attribute table and use the find and replace function using the following codes: Replace all 1** level codes, except 140s, with 100. Replace all 14* codes with 140. Replace all 240 codes with 100. Replace all 2** codes, except 240, with 200. Replace all 3** codes with 300. Replace all 4** codes with 400. Replace all 5** codes with 500. Replace all 6** codes with 600. Replace all 7** codes with 700. Leave 9** codes as 999. Save changes to file when complete.

81

C. Reproject file to match coordinate system of satellite dataset using the Arc Toolbox Project tool (under Data Management Tools\Projections and Transformations\Feature). Enter the following: Input Dataset or Feature Class: ..\Level_1_compare\Sullivan_all.shp Output Dataset or Feature Class: ..compare\ Sullivan_all_UMT.shp Output Coordinate System: Select Coordinate System button to open the Spatial Reference Properties window. Select the Import button and browse to find the satellite dataset that you will be comparing your data to. Geographic Transformation (optional): Chose the first geographic transformation option available. Even though ESRI suggests that this is optional, it is usually a requirement to choose one of the options. D. Convert feature class to raster dataset. Open ArcView with county feature class as well as satellite dataset. Set the spatial analyst Options to: Working directory: \Level_1_Compare Analysis Mask: <None> Analysis extent: Same as Layer Sullivan_all_UTM Snap extent to: umd_lulc_v105 (or current satellite dataset you are using) Analysis cell size: As Specified Below Cell Size: 10m Convert features to raster: Input features: Sullivan_all_utm Field: LU_Code Output cell size: 10 Output raster: Level_1_compare\LAL_Sull Open attribute table and determine which Values correspond with each LU_Code. Reclassify file to change values to actual land use code: Input raster: lal_sull-lal_sull Reclass field: Value Old Value New value Ex. 1 200 only: 2 300 3 100 4 141 5 500 6 600 7 700 8 400 9 999 Output raster: Level_1_compare\LAL_Sull_lv1 Step 2. Cutting satellite dataset to county A. Reclassify satellite dataset to Anderson Level 1 classification system:

82

Set Options to: Working directory: \Level_1_compare\ Analysis mask: LAL_sull_lv1 Snap extent to: umd_lulc_v105 (or other satellite dataset) Cell size: 10 m Using spatial analyst Reclassify tool, change old satellite land use codes to Anderson level 1 codes as stated below. Rename the file: Ex: UMD_Sull_lv1
Anderson Level 1 LU/LC Codes 500 100 100 100 140 100 100 100 100 700 700 400 400 400 200 200 200 600 600 600 600

U of M LU Codes

U of M Description

Anderson Level 1 Description

1 3 4 5 8 10 11 12 15 17 18 20 21 22 25 26 30 35 36 37 38

Open water Low intensity development Med intensity development High intensity development Transportation Urban/residential deciduous tree Urban/residential evergreen tree Urban/residential mixed trees Urban/residential/recreational grass Extractive Barren Deciduous forests Evergreen forest Mixed (deciduous-evergreen) forest Pasture/hay Croplands "Natural" grass Deciduous wooded wetlands Evergreen wooded wetland Emergent (sedge-herb) wetland Mixed wetland

Water Urban or Built-Up Land Urban or Built-Up Land Urban or Built-Up Land Transportation/Communication Urban or Built-Up Land Urban or Built-Up Land Urban or Built-Up Land Urban or Built-Up Land Barren Lands Barren Lands Forestland Forestland Forestland Agriculture Agriculture Agriculture Wetlands Wetlands Wetlands Wetlands

83

PSU 2000 LU Codes

PSU 2000 Description

Anderson Level 1 LU/LC Codes 500 100 100 200 200 400 400 400 600 600 700 700 700

Anderson Level 1 Description

1 2 3 4 5 7 8 9 10 11 12 13 15

Water Low Density Urban High Density Urban Hay Pasture Row Crops Coniferous Forest Mixed Forest Deciduous Forest Woody Wetland Emergent Wetland Quarries Coal Mines Transitional

Water Urban or Built-Up Land Urban or Built-Up Land Agriculture Agriculture Forestland Forestland Forestland Wetlands Wetlands Barren Lands Barren Lands Barren Lands

NLCD 2001 LU Codes 11 21 22 23 24 31 41 42 43 81 82 90 95

NLCD 2001 Description

Anderson Level 2 LU/LC Codes 500 100 100 100 100 700 400 400 400 200 200 600 600

Anderson Level 1 Description

Open Water Developed, Open Space Developed, Low Intensity Developed, Medium Intensity Developed, High Intensity Barren Land (Rock/Sand/Clay) Deciduous Forest Evergreen Forest Mixed Forest Pasture/Hay Cultivated Crops Woody Wetlands Emergent Herbaceous Wetlands

Water Urban or Built-Up Urban or Built-Up Urban or Built-Up Urban or Built-Up Barren Lands Forestland Forestland Forestland Agriculture Agriculture Wetland Wetland

Ex: UMD_Sull_lv1 Combine photo interpretation and satellite derived datasets B. Using the Combine tool (Spatial Analyst Tools\Local\), combine the photo interpretation and satellite derived datasets. Export the attribute table as a .dbf. Open .dbf in excel and use to create confusion matrix. Ex: New Grid File: Com_UMD_LAL_1 Dbf file: Sullivan Combine_UMD_LAL_lv1.dbf Excel File: Sullivan Combine UMD_LAL_Lv1.xls

84

S-ar putea să vă placă și