Special Topics: Lab 15

In the final lab for the course, the focus was on Dasymetric mapping.  Dasymetric mapping uses additional information such as land type to improve determining where populations are allocated.  To help get a better idea of where the population is distributed, impervious data of roads to remove areas where people are unlikely to live.

To perform this analysis, I used the Zonal Statistics as Table tool to find the impervious areas of the census tracts.  Then I joined the new table to the census tract data.  Next I used the Intersect tool for the census tract and high school layers.  I added a field to calculate the area of the new layer.  I also used the impervious against the area to find a new area.  I used the new area multiplied by the population and divided by the before area to calculate the new population.

The reference population was 54,720 and the estimated was 54,661.  About 12% of the population is allocated incorrectly.

Special Topics: Lab 14

This week's lab covered the topic of spatial data aggregation.  The lab this week looked into how congressional districts are within the United States.  The first part of the lab looked at how compacted the districts are throughout the country.  Compactness is based on the shape of the polygon.  For example, oddly shaped polygons are not considered compact.

To determine the top 10 worst offenders of least compactness, I had to calculate the area and perimeter of each polygon.  The odder the shape, the more likely the longer the perimeter.  As the screenshot below will show, the district has a weird shape that clearly shows the district is covering across a lot of space to get certain groups in the district.

Compactness

The other aspect of gerrymandering is community.  It is ideal to have the least amount of districts in each county.  Having more than one district to cover a county show that certain areas of the county could be select to achieve certain results.  I made sure to exclude counties with large populations which would need multiple districts and then looked at how many districts fell in a county.  Below are the results of the analysis.

Community

Special Topics: Lab 13

Week 13 lab focused on how scale and resolution can affect the details of an image.  The first part of the lab looked at vector data while part be focused on raster data.  The first part of the lab compared polylines and polygons at different scales.  The second part of the lab compared LIDAR and SRTM with cell size at 90 m.  SRTM is a very high resolution images.

To compare the two, I had to change the projection of the DEM and resampled it to have the cell size of 90 m.  After that process was complete, I used the Slope tool to find the average slope.  I compared the slope to the LIDAR slope.  I also visually compared the two images.  The LIDAR image has a larger slope and the low elevations are more noticeable in the LIDAR image than SRTM.

Below are the results of the analysis I performed.  The SRTM images shows less change in the elevation than the LIDAR.  As seen in the average slopes, the SRTM has the lower slope.

LIDAR

SMRT

Special Topics: Lab 12

This week lab focused on comparing the OLS model to the GWR model. The Geographically Weighted Regression model.  The model looks at small sets of data.  The model uses an equation to incorporate the independent and dependent variables.  The shape of the results are determined by bandwidth and kernel type.

This lab looked at housing data to determine the relationship.  First the OLS model was used.  Next, I used the same variables to create the GWR model.  After running the models, I looked at the results to see which variable is the greatest interest.  By running the analysis with the GWR, it strengthens the relationship.  By taking the spatial relationship into consideration, it improves the model by using spatial aspect into play.  The OLS does not look at the spatial relationship.  The OLS just looks at the variables and the results don't take into account the kernel type or bandwidth.

Special Topics: Lab 11

Week 11 of Special Topics continued the topic of statistics and regression.  The topic this week taught how to find the best model performance.  Using the Ordinary Least Squares tool, The tool generate results using one or multiple variables.  The results state the coefficients, p values, VIF, and Jarque-Ber.

In order to determine if the selected variables are correct, more are needed, or some should be removed, The a few of the checks are if the independent variables are helping, what are the relationships, and are the variables redundant.  A few other checks look at if the model is biased, if all the needed variables are used, and how well the variables explain the dependent variable.

The lab this week required using the OLS tool along with the Exploratory Regression tool.  The Exploratory Regression tool produces results of whether models pass and goes through all the various options.  The results also use the Adjusted R Square and Akaike's Information.  These numbers can be used to determine fit and explains variation.

Special Topics: Lab 10

This week's lab was an introduction to statistics which involved correlations and bi-variate regression.  Part of the lab consisted of finding missing data for 20 years of rainfall for a rain station.  To find the missing data, I used the regression tool from the Data Analysis Toolpak.  Once I had the regression summary, I found the slope and intercept values.  I used the values in the formula y = m*x+b.  The x variable was the data I had for rain station B.

By using  y = m*x + b, it assumes that whenever x is 0 y equals the slope plus the intercept.  The intercept tells us how much change is in each variable while slope tell us how much the variable will go up or down.  It assumes if I have x, I can figure out y.  The regression analysis looks at where station A and B for the years there is data for both.  It finds the relationship between the stations.

Special Topics: Lab 9

This week's lab covered vertical accuracy of a DEM.  I used data points on top of the LIDAR layer and used the Extract Values tool.  Once the values were extracted, I had to convert the values from feet to meters.  The next thing was to calculate the difference between the LIDAR values and the field points.  Once the difference was calculated, I squared the difference and then summed it.  The next step was to find the average and then take the square root to find the Root Mean Square Average.

Once the RMSE was calculated, to determine the 95th percentile accuracy, I multiplied the RMSE by 1.96.  To find the 68th percentile accuracy, I multiplied the RMSE by 1.69.  The lower the number, the more accurate it is.  To determine if there was a bias, I had to find the Mean Error.  To find that, I took the sum of the difference and then take the average.  Below are the results.  The results show the most accurate and that the urban area had the most bias.

Accuracy Results

Special Topics: Lab 8

This week's lab covered interpolation.  The example used was for water quality of Tampa Bay.  Four different techniques were used to display the Biochemical Oxygen Demand (BOD).  One technique used non spatial analysis.  The other technique was Thiessen interpolation.  This analysis used the Create Thiessen polygon tool.  The input features are the BOD points and the output is all fields.  A mask needed to be applied to only show the Thiessen polygon for Tampa Bay.  This used the points in each polygon to determine the whole value.

Technique three used Inverse Distance Weighting interpolation.  The IDW tool is used to perform this and I had to adjust the radius and power.  The last technique was spline.  The Spline tool was used for both regularized spline and tension.  Since there were points very close to each other, it caused high concentrations.  The points needed to be removed or the average to be calculated.

The results of the analysis show fairly similar for each technique except the regularized spline.  Below is an example of the tension spline.

Tension Spline

Special Topics: Lab 7

The lab for this week compared TIN models with DEM models.  The TIN model shows the terrain and lake with fairly rigid lines. The DEM model shows the changes in elevation with far smoother contour lines.  The TIN didn't really show the terrain well, so I had to adjust the TIN by using the Edit TIN tool.

It is interesting that DEMs and TINs can be created from elevation points.  DEM models are represented by rasters while TINs are represented by vectors.  Since DEMs use pixels, the the changes in elevation seem smooth.  Creating a DEM with slope, aspect, and elevation was also different than using triangulated elevation points.

Modified TIN

Special Topics: Lab 6

This week's lab covered location allocation.  To perform this analysis, I added the distribution centers as the locations.  I added the customers as demand points.  I added the analysis settings and used an output of straight lines.  There was no impedance cutoff and it used all of the facilities.  After solving, not all of the customers seemed to be assigned the closest center.  Then I reassigned the market areas by performing a spatial join of market areas and the customers.  I performed table joins to figure out how many customers go to which facility.  The Summary Statistics tool which counted the customers in market areas.

Next I created a new feature class by joining the unassigned market area and new table.  Now the market areas are reassigned.

The weakness is that the allocation did not always have distribution centers going to the closest customers.  The strength is the ability to analyze all of the settings and inputs very quickly to show market areas.

Location Allocation

Special Topics: Lab 5

Lab 5 covers Vehicle Routing Problem and how adjusting setting and routes, it impacts how many stops and routes will be produced.  To start the analysis, I added the customer information for the orders.  I adjusted the parameters for pick up times and other parameters.  Next I added the distribution center for the depot.  I set the parameters for when the depot can be stopped at.

Next I added the routes I loaded the truck information.  For the route properties, I adjusted when the route can start, the cost per mile and cost per minute, assignment rules, and the maximum capacity.  I added route zones and adjusted the parameter to True so routes stay to the correct zone.  U-turns were also not allowed for the routes.

Finally, I ran the Solver and 6 orders were not reached.  To fix this, I adjusted the properties for Truck 15 and 16 to "include" the assignment rule.  Previously, they were set to exclude which meant forced the routes to only pick up orders assigned to that truck.

After adding the new routes, below is a screenshot of the new routes which services all orders.  Only one exceeds the time limit.  Customer service will increase since every order is taken care of and only one time violation.

Improved Routes

Special Topics: Lab 4

This week's lab looked at create new network datasets and adjusting the analysis settings.  I created a network dataset that used streets as the participating feature class.  I selected to model turns, but did not use restricted turns.  I used elevation fields, but did not select to use traffic modeling.  Once all the settings were adjusted, I built the network dataset.

In ArcMap, I added the recently created network dataset.  I enabled the network analysis and created a new route.  To create the route, I had to load the facilities for the stops.  The impedance was set to minutes and the stops were able to be reordered to find the fastest route.  However, the first and last stop had to remain the same.  The only restriction was one ways.

Next, I added the restricted turns to the Turn settings and rebuilt the network dataset.  In ArcMap, I added the streets and restricted turns layers.  I used the same network analysis setting as previously, and resolved the route.  By adding restricted turns, the route has to adjust slightly to find a new route.

Restricted Turns

The final route analysis required to build a new network analysis.  This time, I used traffic modeling.  I made sure to adjust all of the traffic settings.  The traffic data looked at speeds, and level of traffic.  It contains the free-flow speeds. The network analysis settings were the same was the other two routes.  With adding traffic information the route is adjusted to traffic speeds.  Previously, that information was not included in the travel time.

Traffic Data

Special Topics: Lab 3

This week's lab focused on analyzing the completeness of road networks.  I compared how complete the street centerlines and TIGER roads are for Jackson County.  Below is how I performed this analysis.

I needed to calculate the distance of the roads for both shapefiles.  I did this by adding a field and used Calculate Geometry.  This calculated the distance of each road segment in kilometers.  Next, I needed to find the roads that fall within the Grid shapefile.  To do this, I used the intersect tool to create a new shapefile that had all the road segments that intersect with the grids.  I used the tool on both the street centerlines and TIGER roads.

Once I had the new shapefiles of roads only within the grids, I had to recalculate the distance of the roads.  I used the calculate geometry tool again.  I then exported each attribute table of the new distance.  I found the difference between the length of road and also found the percentage.  Below is a table.

The map below shows the absolute difference between the roads completeness.  

Road Completeness Analysis

Special Topics: Lab 2

Lab 2 covered the National Standard for Spatial Data Accuracy.  The lab looked at junction points for two data sets.  To perform this lab, I needed to create two new Data set Networks for the street shapefiles.  I selected over 85 points of the junction shapefile for the city and over 85 for the USA street shapefile.

After selecting the points, I exported the points into new shapefiles.  I then added orthophotos to see where the true intersections are located.  Zooming into each pair a test points, I found the true location and added a reference point.  I did that for all points.  A unique identifier was applied to all three data sets.  I used the Add X Y Coordinates to the attribute tables.

I exported the attribute tables for the three shape files and inserted the data into the Horizontal Accuracy spreadsheet.  This spreadsheet found the difference between the X and Y coordinates.  The differences were squared and summed.  The Root Square Mean Error was also calculated and multiplied by 1.7308 for the standard error.  This was done for the city data points and the street data points.

City:
Positional Accuracy: Using the National Standard for Spatial Data Accuracy, tested 30.64 feet horizontal accuracy at 95% confidence level.
USA Street:
Positional Accuracy: Using the National Standard for Spatial Data Accuracy, test 62.67 feet horizontal accuracy at 95% confidence level.

Test Points

Special Topics Lab 1

This week's lab covers the precision and accuracy.  The first part looked at the precision and accuracy of a GPS unit.  A point was observed and mapped 50 times using the GPS devise.  To determine how accurate or precise the observations were, the average X and Y coordinates were found and a point added.  I performed a buffer around the average point for 1, 2, and 5 meters.

I did a spatial join with the average point and the 50 observation points to find the distance of each point to the average point.  I then found the distance for 50%, 68%, and 95% of the points.  I also then determined the average elevation and found the absolute difference of the average point to the observed points.  Below are the results, along with a map showing the average point and the observed points.

The reference point of the actual location was added to the map.  To see how accurate the average point was, I measured the distance from the reference point to the average point.  I did the same for elevation as well.  This was done to determine how accurate the average point is.  The average point was within 3 meters of the reference point and the elevation was within 6 meters.

The reference point and the average point differ by quite a bit.  The longitude and latitude is off by 3.8 meters from the reference point to the average.  The horizontal precision was 4.4 so it was greater than the true difference.  The elevation is off by 6 meters while the precision showed 3 meters.  Even though 4 meters is not a lot, depending on the need of knowing this location, it can be huge.  GPS units can only be so accurate and the unit puts the point fairly close to the true position.

The horizontal accuracy was 3.8 meters.  This is better than the horizontal precision.  The vertical accuracy is 6 meters which is worse than the vertical precision of 3 meters.  There was no evidence of bias in the results.

The second part of the lab covered calculating the Root Mean Square Error, mean, median, the percentiles, min and max values.  I then used the XY errors to plot a CDF chart.  I compared the chart to the metrics I calculated.  From looking at the chart, it is clear to see certain metrics like the percentiles or the min and maximum number.