Trend Analysis Through Time Series of Raster Data

12634
31
Jump to solution
07-27-2017 11:14 AM
ShouvikJha
Occasional Contributor III

Hi All, 

I have total 13 raster files for different years (2001 - 2013). All data are in tiff format and same spatial extent (row and column). I am trying to perform regression line slope/trend analysis between each grid points for 13 raster data sets. I have tried in ArcGIS raster calculator but I won't run the task due to complexity. 

The outcome would measure the net change between pixels through my time series data. 

 

So my question is, can it be performed using raster calculator or ArcPy Script?  For such type of statistical analysis using ArcGIS, really I am facing one of the big limitation with ArcGIS. 

 

Using excel, it can be calculated using SLOPE command where X (Time) and Y (value) has been used. But for creating a Spatial map of Slope/trend analysis, we need to perform using raster data.   

 

The download link is given below for Sample data sets (Due to size limit, I could not attach the data with here )

Data Link: Dropbox - Water_Use_Efficiency.zip 

Please share your experience. 

Thanks in advance. 

 

0 Kudos
31 Replies
耀军李
New Contributor

Dear Dan Patterson, can you explain how to calculate the linear regression in detail? I can got the mean but failed to got the trend (slope of linear regression)

0 Kudos
TilmannSteinmetz1
New Contributor II

That looks like a nice way of scripting it. Have you seen the "Space Time Pattern Mining" Toolbox though? toolboxes\system toolboxes\space time pattern mining tools.pyt\utilities

It contains tools to create a space time cube by converting your rasters into a multidimensional NetCDF file with a time dimension, and visualise that in time and space - I imagine you should be able to use that at least for visualisation of your time series - though I'm not sure that you can actually get a tabular (or other) output of the deltas between the timesteps.  There is an "Emerging Hot Spot Analysis" tool in the toolbox which can be used to visualise statistically significant trend/change areas over time.