I have a collection of images with NDVI and time bands that I want to compute a pixel-wise linear regression on with a corresponding color map that represents NDVI trends in the time series.
Page 47 of this article http://repositori.uji.es/xmlui/bitstream/handle/10234/154568/TFM_2014_ghazaryanG.pdf?sequence=1&isAllowed=y
is an exact indication of what I'm hoping to do, however, when I apply the Google Earth Engine methods that were provided at the bottom of this article to my collection, I don't know how to gain access to the pixel-wise slopes of the computed linearFit reduction. The visual slope values do not seem to be an accurate representation of the possible variability in NDVI trends and I hope to verify the visual trends with actual numbers.
Hi Ian,
I think you can you the following codes to get the pixel-wise linear fit for your NDVI datasets in Google Earth Engine:
https://developers.google.com/earth-engine/reducers_regression
https://gis.stackexchange.com/questions/262434/linear-regression-google-earth-engine
Also, a very recent article on this can be useful:
Article The Grass Is Not Always Greener on the Other Side: Seasonal ...
Cheers,
Have a look at this website. It could be of help:
https://developers.google.com/earth-engine/reducers_regression
Thank you to Nikul Kumari and Ambrose A Zemba for your help. I discovered that the 'scale' property of the linear fit exceeded the min and max values that I was allowing for its visualization. Using the 'Inspector' tab in the Google Earth Engine Code Editor, I was able to verify that the range of the slope was between [-300000, 300000]. Adjusting visualization parameters to this range allowed for much more accurate visualization.
Thanks again!
In the Google Earth Engine (GEE), pixel-wise linear regression is often carried out by using the ee.Reducer.linearFit() method. This method calculates a simple linear regression for an ImageCollection, a series of Images that represent a time series. Each pixel in the output image represents the linear regression slope for that particular pixel over time.
Here's a simple example using NDVI data:
javascriptCopy codevar collection = ee.ImageCollection('YOUR_IMAGE_COLLECTION') .select('NDVI') // Select the NDVI band .sort('system:time_start'); // Sort by time to ensure correct ordering // Compute a pixel-wise linear regression var linearFit = collection.reduce(ee.Reducer.linearFit()); // The scale is the pixel-wise slope of the linear regression var slope = linearFit.select('scale');
In this code:
Once you have your slopes, you can visualize them using GEE's built-in functions:
javascriptCopy codeMap.addLayer(slope, {min: -0.01, max: 0.01, palette: ['blue', 'white', 'green']}, 'Slope');
This would add the slope layer to the map with a color palette ranging from blue (for negative slope values) to white (for near-zero slope values) to green (for positive slope values).
If you'd like to export the slope values to verify them numerically, you can export the data as a GeoTIFF using the Export.image.toDrive() function:
javascriptCopy codeExport.image.toDrive({ image: slope, description: 'NDVI_slope', scale: 30, region: YOUR_REGION });
In this code, replace YOUR_REGION with the region you're interested in.
After exporting, you can load the GeoTIFF file into GIS software like QGIS or ArcGIS, or a data analysis tool like Python or R, to inspect the slope values numerically.
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology
- Pharmaceuticals
- Drugs