Respected fellows greetings
I need to find a co-relation between NDVI and LST values of my study area but i am unable to get pixel value of my raster data in a table form PLEASE suggest a suitable way to get all pixel values in a table format or suggest any direct way to find co-relation between NDVI and LST in raster form.
Hey Muhammad. Have you tried using R? Try to check this link if it helps: http://neondataskills.org/R/Extract-NDVI-From-Rasters-In-R/
Cheers!
Muhammad, there are many ways to do this. If I were you, I would definitely try Eric's suggestion in R and try other open source tools first like in Python. Alternatively, I have suggested a way here in ERDAS Imagine (ver. 2014). There is a tool called 'Pixels to ASCII' under the 'Manage Data' > 'Conversion'- Menu tab. This tool lets you export the pixel values along with the coordinates in a raster (NDVI or LST) into an ASCII format which you can then open in Microsoft Excel e.g..The older versions of ERDAS Imagine may have this tool under the 'Utilities' tab. Anyway look at the attached screenshot for directions. This tool might be helpful for you while doing statistical analysis between NDVI and LST.
Excel file size become too heavy and fails to open completely, what can i do besides decreasing area size?
Alright, I thought you were working with subsets of LST and NDVI data. Obviously, while using 'Pixels to ASCII' tool for an entire scene of a satellite Image such as Landsat, we are talking about exporting at least a few millions of pixels! So even if you are able to open the ascii file in some software package, it may not be a good idea to run statistical analysis on such huge datasets. As you may have figured out by now, it is quite a memory intensive task for the computer. In ArcGIS Desktop, most of the inbuilt tools available for statistical analysis work with vector data. Converting your raster to point data is also not a feasible solution due to the considerable size of the data. However, you may want to check out the 'Spatial Analyst Tools > Multivariate > Band Collection Statistics'. It just produces a text file with 'Correlation and Covariance' Matrices. There are similar tools in 'R' which can give you the Pearson's Correlation coefficient etc. I ran a similar NDVI, LST analysis in R and found some negative correlation, which kind of made sense but had a hard time plotting the results spatially due to inexperience in programming. I will let you know if I succeed. Alternatively, experts from Esri have suggested a way using 'Create Random Points' and 'Extract Multi-Values to Points' in ArcGIS. In IDRISI there is an inbuilt tool for regression analysis. But, if you are fine with scripting in Python and R, that would be great.
Anyway, this is the summary of what I found so far. Comparing two rasters to find possible correlation seems at first like a simple problem, but in reality, one has to take into consideration a number of factors such as the kind of data (quantitative, qualitative, nominal, ordinal, interval, ratio, parametric, non-parametric etc.) one is dealing with. Once this is determined, a possible tool can be sought to implement a method like Correlation, Linear Regression, Ordinary Least Squares, Geographically Weighted Regression, Spatial Autocorrelation etc. Please take some time to go through the links in order to avoid reinventing the wheel.
Sorry for such a long and imprecise answer.
https://www.researchgate.net/post/How_to_statistically_compare_two_maps
https://geonet.esri.com/message/705356-re-trend-analysis-through-time-series-of-raster-data
https://gis.stackexchange.com/questions/179126/spearman-correlation-between-two-rasters-in-r
https://geonet.esri.com/thread/184249-re-perform-raster-calculation-from-multiple-sub-folder-using-arcpy#comment-640193
https://geonet.esri.com/thread/193050-linear-regression-raster-data
https://www.youtube.com/watch?v=zITvBMrqM-E
James your answer is really very detailed and helpful
thanks for your time and effort