Lab 7

Lab 7

Vegetation Trend Analysis Using TerrSet

Jackson Radenz

Introduction:

As our climate and land use patters change, it is important to monitor how the vegetation on earth responds to the change.  The application of remote sensing allows us to study the temporal change of vegetation  across large amounts of area.   Remote sensing enables analysts, policy makers, and decision makers to see the big picture with a larger sense of ease.  It is increasingly important to study vegetation at large scales because it provides a large ecological service for us by creating oxygen and sequestering the large amounts of carbon produced by humans.  By sequestering carbon, smaller amounts of CO2 are released into the atmosphere, and global warming is reduced.  Not only does vegetation have large ecological benefits, by studying it at large scales it allows humans to make connections between ongoing processes and patterns on the earth.  For example, if there is a browning trend over a 5 year period, perhaps we are able to connect that to a drought over the area.  With that information we can predict what will occur to the surface of the earth if similar conditions exist in the future.  For this exercise, we used TerrSet to create OLS trend regression and significance outputs in order to study areas of browning and greening across the earth's surface.

OLS Trend Regression:

A common statistical tool that is used in studying temporal data across space is OLS Regression. Ols stands for Ordinary Least Squares.  OLS is a statistical regression that predicts unknown quantities of data.  It tries to find the line going through the sample data that minimizes the sum of the squared errors (Lehe & Powell, 2015).  Basically what OLS attempts to do is fit the data with a regression line that most accurately predicts the data (Figure 1).  By doing so, this allows analysts to visualize trends throughout data. 
Figure 1: OLS Regression

Data:

  • Imagery collected by: MODIS
  • Date Collected: 2000 - 2017
  • Type: 16 - day composite from 2000 - 2017 
  • Zone: h11v4 - Upper Midwest, Great Lake Region

Methodology:

  1. Create a new project in TerrSet
  2. Convert GEO/TIFF Stacked NDVI Raster file to Idrisi file
  3. Open the file in Earth Trends Modeler (Figure 2)
  4. Figure 2: Earth Trends Modeler
  5. Editing Earth Trends Modeler
  6. Change series  to annual and change the dates from 2000 - 2017
  7. The output will then consist of a time series visualization (Figure 3)

  8. Figure 3: Visualization Cube
  9.   Open Explore Trends
  10.  Create OLS Trend Regression and Significance Outputs 
  11. Convert the outputs to TIFF files that can be opened in ENVI
  12. Open both OLS Trend and Significance Outputs in Bath Math
  13.  Write (b1 lt 0.05)*b2
  14. Assign b1 to significance output file and OLS Trend to b2
  15. This will create a file that only keeps significant pixels
  16. Finally, our goal is to see the browning or greening trend over different types of landform
  17. Select ROI Tool
  18. Band threshold of ROI
  19. Type in the land class cover
  20. Subset ROI
  21. Create an output using selected ROIs (Figure 4)
  22. Figure 4: Subsetted ROIs
  23. Examine output's quick stats in order to determine areas of greening and browning over different types of land cover

Results:

Figure 4: Browning and greening across MODIS acquired imagery

Area with statistically significant browning trend:

16,322 km sq.

Areas with statistically significant greening trend:

20,489 km sq. 

Highest Magnitude of Browning:

-426.72

Highest Magnitude of Greening:

366.46

Figure 5: LULC areas with significant trend

Discussion & Conclusion:

Overall, the areas to the west of the collected imagery (North & South Dakota, Canada, Minnesota) saw the largest greening trend.  However, large parts of the imagery, especially areas in Michigan and Ohio received large areas of intense browning.  I found it interesting that closed evergreen forest experienced the largest amount of browning, by a large margin.  I hypothesize that is is mainly due to forest fires, commercial timber, invasive species, and urban encroachment.  I also found it interesting that water bodies experienced a large amount of greening.  I would contribute this to phosphorous and nitrogen run-off from farms into nearby water bodies, which causes large  amounts of growth of algae and aquatic plants.  Finally, herbaceous and closed deciduous forest experienced to largest amount of greening relative to the amount of browning.  I hypothesize that closed deciduous forest and herbaceous areas have received a renewed effort of preservation, which protects them from practices that would reduce their size.             

References:

Lehe, L., Powell, V. (2015) Ordinary Least Sqaures Regression. Explained Visuallyhttp://setosa.io/ev/ordinary-least-squares-regression/  Date accessed: 4/5/2015