# Calculating NDVI-Analyzing Results

I would like to calculate the NDVI for an area to identify vegetation cover over a 32 year period using. However, I am having challenges with the calculations I have attempted with Raster Calculator. I am using images acquired from the USGS Earth Explorer for the years 1988, 2000, 2017, 2018 and 2020. I have created the Multi Band Composites for each year and used the Bands from those Composite images to perform the calculation for NDVI in Raster Calculator.

For the first image 1988 (Landsat 4) I used Band 4 (NIR) and Band 3 (Red), using the equation

Float("1988\compo_1988 - Band_4" - "1988\compo_1988 - Band_3") / Float("1988\compo_1988 - Band_4" + "1988\compo_1988 - Band_3")

However, the results (seen in the screenshot below) were not what I was expecting, I.E values ranging from -1 - 1.

Is there a way I can get the values that are usually associated with NDVI? If not, how can I use these results to identify which areas are vegetation or not?

• What values do get when you actually query pixels? There is an identity tool in the toolbar that let's you click on the map and shows you the query value in a pop-up box. Commented Mar 16, 2021 at 14:34
• I checked the values for each of the classifications of interest and got the following figures: Water: -0.79, Bare Earth: 0.00, Non Mangrove Vegetation: 0.75-0.82, Urban: 0.13 - 0.22, Mangrove: 0.45 - 0.7. Commented Mar 16, 2021 at 14:47
• have you tried with the bands before you make the composite ? Mathematically this seems correct and should yield something between -1 and 1. Maybe the results are corrects but the displayed stats are not. In any case I would say that, roughly, you yellow = no vegetation and your orange = vegetation). Please go to reclassify to show the histogram of the data. Commented Mar 16, 2021 at 14:58
• That is showing you that the values are in the expected range. You may be getting strange bounding values due to collar values or ArcGIS just acting odd. You could use the raster calculator to bound the values by setting everything > 1 and < -1 to NODATA, or using the conditional tool Set Null. Commented Mar 16, 2021 at 17:51