Wednesday, April 17, 2013

Project-continued



Methods:
·         Study area description: demographics, target, current status.
·         Data sets: Apendix, tables
·         Model Overview: purpose, how it works, scenarios
·         Technical Overview

                It took a considerable amount of time to install ArcGIS onto the computer. Part of the problem was forgetting to connect to the VPN the other issue was establish the license through arc administrator. Once installed everything worked well; except the spatial analyst tools. I attempted to install the service pack to fix the issues but that didn’t seem to work. I put aside the spatial analyst component and began by adding layers and importing them through ArcCatalog. I then converted the raster layers to pyramids and put the vector layers into a file geodatabase. I created two geodatabases to store my data. One was created as a workspace, while the other was created to hold more refined layers. All of the data, downloads and manipulated files, are kept in one folder on my external hard drive. Organization is one of the biggest challenges and certainly essential to success.
                Next I put everything into Arc Map. I changed the projections all to VT State Plane NAD83. I also converted the county boundaries into raster format. This was fairly simple; it required opening the polygon to raster tool and placing the vector layer in. I added a new data frame to hold the initial layers. To do this open the insert menu and select new data frame; I labeled it as old data. In addition as I continued to process layers I attempted to keep only the newest and most relevant in the active data frame. The relative level of layers stayed the same in the active data frame. The other data frame accumulated quite a few layers due to all the intermediate steps and trial and error. When I left my computer for a couple hours ArcMap stopped working when the vpn disconnected, I lost a little bit of work. When I returned I enabled the spatial analyst using the “customize” menu then “extensions”.
                Then I used Euclidean distance to get the distance to roads. It allowed the conversion of a vector road layer into raster format. It seemed to work alright; however, I wasn’t able to clip it and the number of categories and classifications seemed really weird. After much trial and error I still wasn’t able to clip the layer correctly. I used the environments- processing extent (raster counties) and snapped to raster counties. The result was a rectangular shape. After researching I decided to use “extract by mask” to essentially clip the layer. Essentially this tool is like placing a mask on a piece of paper and only removing the area covered by the mask. By putting the layer into “in_raster” that you want to extract from, and placing the mask layer into “in_mask_data” the raster layer is restricted to the desired  area. I used the extract by mask tool for three layers: land use in 1992, land use in 2001, and the distance to roads (raster) layer. For all three of the processes I used the raster county layer as the input for “in_mask_data”. The land use layers were for the entire Champlain Valley. The “extract by mask” tool allowed the left border to be restricted to Vermont’s western border. Finally I classified the distance to roads with equal distance. I arbitrarily created 16 different intervals to show a greater and more specified range.
                I added a natural areas layer from ANR, into ArcCatalog. This layer consisted of a series of state designated areas that typically consist of some wild area. I used the conversion tool “polygon to raster”. The input layer was the natural areas polygon, set to the extent of the raster county boundaries. I sent the output to the VLT.gbd. On Tuesday April 9th I received a forwarded message from Chris at the Vermont Land Trust. The data I had been waiting for finally came. This data should easily fit into the current model. In ArcMap the data is in polygon format, I used the tool polygon to raster to convert the data. Then I converted it to tiff format and exported it to the new data frame.
                Next I began working on a model in Dinamica for transition matrices. Right away I was unsure how to move the data from ArcMap to Dinamica. After meeting with my professor, things began to make more sense. I found out that I needed to convert all the layers to tif. I created a new data frame to hold the tif layers and remain as organized as possible.  The new data frame was labeled as To_Dinamica. Each layer and variable was converted to tiff format in order to be used in Dinamica. Next I opened a transition matrix model. I altered the “save single step matrix” and “save multi step matrix” so they would save in a new folder in my hard drive. These were saved using the functor “save lookup table”. Once the file path was established to the folder “transition matrix” I was able to import the landuse.tif from 1992 and landuse.tif from 2001. These were imported using the functor “load categorical map”. I set the number of iterations to 9. This is the number of years between 2001 and 1992 and the number of repetitions the model requires. Next I ran the model and saved the data. The output from “save multi step matrix” identifies the significant transitions. It compounds the data from year to year and it will be this information that is used in the next step.
                 By altering the Weights of Evidence model from an example (Figure 1), I can incorporate the two land use matrices and static variables. In the “load categorical map” functors the land use maps from 1992 and 2001. One major way the model is different is that the static variables will not be stacked. This will result in more “load map” functors and additional “name static_var” functors. The “name static_var” will be held within the “Determine Weights of Evidence” container/functor. I will use slope, elevation, distance to roads, biomass, and Vermont Land Trust Properties as the static variables. I used 4 load map functors to add the first four of these static variables. The Vermont Land Trust layer required a load categorical map because the data is categorical. In addition 5 name maps were added under “determine weights of evidence ranges” and under “determine weights of evidence coefficients”.  Once all the functors were loaded, each  static variable was connected to a name map within each of the boxes. Each map should have two connections.  All of the name maps must be labeled the same in both boxes. Additionally they should be labeled corresponding to the data they are connected to. (ADD FIGURE AND SCREEN SHOT)
                An issue came up where Dinamica reported that the static variable was not working with the current format. Essentially the number of rows and columns in the static variables didn’t match the expected rows and columns based on the land use maps. After spending a considerable length of time with my professor assessing the potential problems and the meaning of the difference of columns and rows, an idea came to me. All of the static variables were processed to the boundaries of the state: however, the land use maps were restricted to the Lake Champlain basin on the Vermont side. Of course I forgot that I wrote this in my notes. I completely redid the model, and strayed from the example. After reconstructing the entire model functor by functor I recalled that the layers were not set to the same extent. I used extract by mask and restricted the layers to the extent of the land use maps. I had to manually adjust the rows and columns on the data acquired from VLT. This changed the raster cell size slightly; however, it didn’t seem to make a significant difference in the execution of the model.
Results/Discussion
2-1 (Agriculture to Urban)
2-8 (agriculture to urban-open)
3-1 (Brush to Urban)
3-2 (Brush to agriculture)
3-4 (Brush to forest)
4-3 (forest to brush)
Some that may be important but not significant: 2-4 (agriculture to forest), 4-1 (Forest to Urban), 4-2 (forest to agriculture)

Appendices
Table 1
BoundaryCounty_CNTYBNDS
Statewide
Polygon
VT State Plane Meters NAD83
VCGI
Basic outline of counties
LandLandcov_LCLU 
Statewide
Raster
NAD83
VCGI
Breakdown of land use in VT
TransRoad_RDSMAJ1
State
1500
NAD83
VTRANS
Vermont Roads by class
USGS National Elevation Dataset (NED) 30 M DEM
State
30 m raster
Vermont State Plane NAD83
VCGI
30M DEM, elevation
LandLandcov_LCLULCB01 
Champlain Basin
30 m raster
Vermont State Plane NaD83
UVM
Landuse in champlain basin (2001)
LandLandcov_LCLULCB92 
Champlain Basin
30 m raster
Vermont State Plane NAD83
UVM
Landuse in Champlain basin (1992)
EcologicHabitat_NATAREAS 
State
Poly
Vermont State Plane NAD83
ANR
Natural areas designated by the state (33 of em, various characteristics)
VLT_Property
State
Poly
NAD83
VLT
Vermont land trust property

Table 2
Figure 1: Biomass
Figure 2: elevation
Figure 3: Land Use 2001
Figure 4: Land Use 1992
Figure 5: Distance to Roads
Figure 6: Slope
Figure 7: Vermont Land Trust data
Figure 8: Dinamica, Lesson 6 Flow Chart
 

No comments:

Post a Comment