09-19-2017 09:02 AM
I have just watched two videos on NDVI from Hexagon Geospatial and one of those videos is yours.
But I have a few basic questions"
1. is there a fundamental difference between simply calcualting NDVI as shown in this video:
and NDVI mean as shown in the video you uploaded? http://community.hexagongeospatial.com/t5/ERDAS-IMAGINE-Q-A/Mean-NDVI-value-calculation/ta-p/6352
I know that the mean is literally the average pixel values. My question is would there be a fundamental difference in outputs if these two methods were used?
I did try your method by loading the vector which is the shape file of my AOI and got an error message that says , units are not the same.
I wish to subset my AOI from the Landsat image and then calculate my NDVI value and if i do a subset, i do not think i need to use your method of adding a vector shapefile. Or what do you think?
Should I use a subsetted image , will i still come out with an output that can be termed an average NDVI value?
Lastly, what is an ndvi map? is it distinctly different from the ndvi resulting image?
With best regards as i hope for a helpful clue from you as an expert in this field.
09-19-2017 09:30 AM
Yes, outputs will be different.
Calculating the NDVI index as explained in the video (the first link) will give you a grayscale image, where each pixel will have the value of NDVI between -1 and 1. Then via standard tools or Spatial Modeler you can create a thematic image with the color ramp added (for each pixel or group of pixels). So the result here is a raster image.
Calculating the NDVI mean using steps I described in the article (second link) will give you vector data as a result with an additional column added to attribute table that will show MEAN NDVI value for each polygon in your vector file. This could be useful if you would have a shapefile with polygons (where each polygon represents a field with some crops growing on it) and you need to create graphs for monitoring crops growth dynamic. So you would take images for different dates, calculate mean NDVI for each field, create a graph to have some general idea of how well or not some particular crop is growing within a field.
So here you should decide what you need - a colored raster image (NDVI map) or a vector file with a mean NDVI value for each field.
As I got you need to calculate NDVI for your specific AOI, so you don't need to follow the process from my article.
Creating an NDVI map is basically following the process from the first video and additionally attaching a color ramp (with some standard colors from red through yellow to green OR grey-brown-green). Also you can merge pixels to groups based on some criteria and assign specific color to each group (can be done within Spatial Modeler).
Also you don't need to subset your image. You can apply AOI directly in the Indices window by clicking on AOI... button:
If you have a shapefile, here is an article on how to convert shapefile to AOI Converting shapefile features to AOI in the IMAGINE ribbon
Hope I have answered your questions.
10-04-2017 04:07 AM
Many thanks for your detailed answers.
But as you would know, sometimes , videos are best remote support tools
So I would be glad if you could eitehr do a video on ho wto add the colour rams or maybe a step by step method on how to add colour ramps to teh grey scale NDVI image
I will greatly appreciate this please
10-04-2017 05:22 AM - edited 10-04-2017 05:27 AM
Similar question was already answered in the following Community post http://community.hexagongeospatial.com/t5/Support-ERDAS-IMAGINE/Question-NDVI/m-p/15643#M1033
And also a Spatial Model that attaching attributes to raster is in our Spatial Recipes http://community.hexagongeospatial.com/t5/Spatial-Recipes/Attach-Attributes-to-Raster/ta-p/772
10-10-2017 02:12 AM
Please what are your views on calculating NDVI with TOA reflectance units ?
I am working with multitemporal images from one sensor and want to monitor changes over my AOI in the last 20 years.
I am using the Landsat Reflectance Conversion in ERDAS IMAGINE 2016.
Would I be breaching any scientific gap if I calculate my NDVI with TOA values? since atmospheric correction has not be taken into account in the Landsat Conversion tool.
Is it absolutely neccessary to convert to Surface Reflectance values .
With great thanks
P.S : MVardovs and Dr Ian
10-10-2017 07:18 AM
I think Landsat uses something called GAIN. Whihc means adjusting data DN values depending on imaging aquisition conditions to get sensor capabilities in use in optimal way. Due this there can be variation in data pixel values over the times - which might have effect on NDVI values.
So if yo just trust pixel values in data you need to be aware on following things
1) If you have just one sensor over the whole period then wavelenghts etc. will not change. If you use multiple sensors their different band combinations/wavelenghts might have some affect on result
2) If single sensor - how this Gain value has been altered over the times. I guess it should be readable from image data itself.
3) Obviously weather and temporal variation affects so you get images on different moments of growing season in different lightning conditions and weather history.
In perfect world you would always go on ground reflectance and you would have all weather information in hands. As we are not in perfect world you must handle these in some scientific method as well as you can and evaluate their effect to your conclutions. NDVI tries to normalize things in very easy way that different datasets are comparable. It is not perfect like any other method is not perfect and due this you need to consider these aspects too. The more accurate information you can input to your process the better results you should get.