Correlation between topographic attributes and NDVI: analysis and operational scales
Objectives:
-
Learn how to calculate topographic attributes from a DEM and NDVI from a Landsat 5 image.
-
Examine the correlation between topographic attributes and NDVI at multiple scales.
-
Understand the difference between analysis and operation scales.
Task 1 Data preparation
-
We first need to exclude all water, developed, barren land, hay/pasture, and wetlands from our study. We will use a landcover dataset to create a mask to do that ——栅格计算器SETNULL( ,,)
-
Add the Landcover raster into ArcMap. Check its attribute table and write down the coded values of water, developed, barren land, hay/pasture, woody wetlands, and emergent herbaceous wetlands, respectively.
Water:11;
Developed Open Space:21;Developed Low Intensity:22;Developed Medium Intensity:23;Developed High Intensity:24;
Barren land:31;
Hay/pasture:81;
Woody wetlands:90;
Emergent herbaceous wetlands:95.
-
Use the Reclassify tool to reclassify all the values you identified in step 2 to NoData and name your output raster LC_Mask. All other values to 1. Please note that once the pixels in a raster are reclassified as NoData, they will not be considered in analysis anymore. ——Resample
-
We also need to exclude the flat area as the authors did in the article. They used a threshold value of 10% of the slopes that are calculated from the 240 m DEM to exclude the flat area. ——先slope计算坡度——栅格计算器 calculator
-
Add the DEM_30 raster into ArcMap. Export the data but change the pixel size to 240 m and name the output DEM_240.
-
Calculate the slope from DEM_240 and name the output Slope_240. Make sure you use the percent as the unit of the output slope. ——先slope计算坡度
-
Reclassify all pixels with a slope value less than 10% to NoData and name the output as Slope_Mask. All other values to 1.——栅格计算器 calculator
-
Use the Extract by Mask tool to combine the LC_Mask and Slope_Mask together and name the output Final_Mask. Read the help information if this is the first time you use this tool.
各位同学, 用extract by mask做掩模提取DEM或者NDVI的时候,请点一下Extract by Mask 对话框上的Environments键
然后在Environment Settings里面点Raster Analysis。
Cell size选上你要做掩模的栅格(比如说DEM_30).
这样你的输出栅格的格子大小就是由你输入的栅格的格子大小决定,而不是由掩模(final_mask的格子大小)决定。
Task 2 Topographic attributes and NDVI calculation
-
Export the DEM with a new cell size of 90, 150, 210, and 270 meters respectively. Name the outputs as DEM_90, DEM_150, DEM_210, and DEM_270, respectively.
-
Calculate the slopes from DEM_30, DEM_90, DEM_150, DEM_210, and DEM_270 and name the outputs as Slope_30, Slope _90, Slope _150, Slope _210, and Slope _270, respectively.
-
Use the Raster Calculator to calculate NDVI from the NIR (band 4) and R (band 3) bands of the Landsat 5 image and name the output NDVI_30.
-
Export the NDVI_30 with a new cell size of 90, 150, 210, and 270 meters respectively. Name the output as NDVI _90, NDVI _150, NDVI _210, and NDVI _270, respectively.
-
Use the Extract by Mask tool and the Final_Mask to mask out the excluded area from the NDVI_30, NDVI _90, NDVI _150, NDVI _210, and NDVI _270, respectively. Name the final outputs as NDVI_30_ex, NDVI _90_ex, NDVI _150_ex, NDVI _210_ex, and NDVI _270_ex, respectively.
注意环境里设成不同分辨率条件下尺度。该步骤目的去除一些人为干扰
Task 3 Correlation analysis
-
Use the tool Band Collection Statistics to calculate the correlation coefficients between the elevation and NDVI values at multiple scales. Read the help information if this is the first time you use this tool. Finish the following table.
Table 1 Correlation coefficients between the elevation and NDVI values at different scales
Scales | 30 | 90 | 150 | 210 | 270 |
r |
-
Use the tool Band Collection Statistics to calculate the correlation coefficients between the slope and NDVI values at multiple scales. Finish the following table.
Table 2 Correlation coefficients between the slope and NDVI values at different scales
Scales | 30 | 90 | 150 | 210 | 270 |
r |
-
Challenge question: Thought we have calculated the correlation coefficients, we don’t know whether the correlation is statistically significant or not. We would need to export the values and run the correlation analysis in a software package like SPSS to test whether the correlation is statistically significant or not. You are not required to do this in this project, but you are encouraged to figure out how to do it.
-
Let’s assume that all correlations in steps 13 and 14 are statistically significant. Use the information in Tables 1 and 2 to answer the following questions.
-
How do the correlation coefficients change with respect to the spatial scale?
-
How does this project help you understand the difference between analysis scale and operational scales in geographic research?
-