In this lab, we worked with raw data to create Digtial Elevation Models (DEMs), comparing DEMs from topographic surveys and LiDAR to see differences between the two methods. Within a data type, we compared DEMs of different cell sizes to see how differences in resolutions affect the final product. In the final part of the lab we compared two DEMs from different time periods to analyze change.
Task 1: Ground Based Topography
For the first part of the lab, I took raw survey data collected in the Pats Cabin reach of Bridge Creek OR and converted it to a DEM. To create a DEM, raw point data is converted to Triangular Irregular Network (TIN) and from there, into a DEM. To prevent the TIN from interpolating across sharp changes in topography, breaklines were collected along with topographic points during the survey. Breaklines are collected in locations of rapid elevation change such as the top of banks in order to force the TIN to recognize those spots and prevent the processing from smoothing sharp topographic changes. The final vector data input to TIN calculation was a polygon showing survey extent. This forces the TIN to a fixed area, excluding points and areas that would negatively affect results. Once the TIN is calculated, A DEM is able to be computed. The DEM creation allows for a user-specified cell size. To illustrate the difference between DEMs of different resolutions at derived from the same TIN, I generated a range of DEMs for comparison.
After considering the different DEM resolutions, I decided to to use the DEM at a 1 cm resolution for my final map product. I chose this resoltion because it the smallest cell size of the DEMs that I derived and most detailed final product. The extent of the data set was small enough that my computer was able to derive and display it effectively. With larger data sets, the finer scale resolution might be require too much processing power to practical.
Task 2: Airborne LiDAR
For the second part of the lab, we downloaded raw LiDAR point data from opentopography.org and processed it into a DEM. The size of the data set meant that resolution needed to be carefully considered. I originally attempted to create a 1 cm DEM for the entire reach. After 1 hour, I cancelled the process and created a 10 cm DEM instead, which still took 30 minutes for ArcMap to create. The results the LiDAR DEM are shown below, along with a a comparison between a DEM derived from a collected topographic survey and one derived from raw LiDAR data. Both methods have their advantages and disadvantages. LiDAR can collect a large number of points (more than a human survey crew ever could) very rapidly and shows more topographic
variability because of the large number of points. However, it is not beneficial if you hoping to show in-channel features such a beaver dams. This is due to the fact that during processing, LiDAR point data is stripped of variation in order to show an even water surface. Topographic surveys have fewer points, but are more flexible since they can be easily used to show the features of interest.
Task 3: Change Detection Analysis
For task 3, we performed a change detection analysis to quantify the amount of change in Pat's Cabin Reach. For the analysis we compared the raw elevation differences between DEMs created from surveys taken in 2010 and 2011. From that initial DEM of difference (DoD), I excluded all values that showed an absolute change of less than 20 cm to create the final DoD, shown here on the right. The LiDAR derived DEM is used as a base layer. I ended up using different symbology for the two different maps, if I were to do this lab again, I think that I would make them more uniform between the two maps, with an identical north arrow, consistent use of the bounding box and the same transparency for the locator map. The maps serve their desired purpose and satisfy the requirements of the lab, but do not look great when compared to each other. One final thing that I did in this lab was to create a model in ArcGIS which should allow for the
the process of DoD calculation to be automated for future projects. The model, which is constructed visually, takes two rasters as inputs and subtracts one from the other to create an initial DoD. That initial DoD is then reclassified to into 1's and 0's, with 1's represented the values that fall outside the threshold of uncertainty and 0's as values with too much uncertainty to included in the final product. The reclassified raster is then multiplied by the initial raster in order to create the final product. ArcGIS uses the Python programming language and any model can be edited by the user after it is created. The model shown in the picture above can be downloaded here.
Joe Wheaton & Shannon Belmont Lab 07 - Building DEMs. 2015. gis.joewheaton.org. Web. 3 March 2015. http://gis.joewheaton.org/assignments/labs/lab-07---building-dems
Joe Wheaton & Shannon Belmont Lab 07 - Building DEMs - Task 1. 2015. gis.joewheaton.org. Web. 3 March 2015. http://gis.joewheaton.org/assignments/labs/lab-07---building-dems/tas
Joe Wheaton & Shannon Belmont Lab 07 - Building DEMs - Task 2. 2015. gis.joewheaton.org. Web. 3 March 2015. http://gis.joewheaton.org/assignments/labs/lab-07---building-dems/task-4
Joe Wheaton & Shannon Belmont Lab 07 - Building DEMs - Task 3. 2015. gis.joewheaton.org. Web. 3 March 2015. http://gis.joewheaton.org/assignments/labs/lab-07---building-dems/task-5