Focal (linear) regression between every 5x5 pixels group of cells of two rasters

As part of Extreme Rainfall model development activities (When the rainfall will trigger a flood?), and based on two rasters with same resolution and dimension, we would like to create a new raster of SLOPE (a) and INTERCEPT (b) by performing linear regression between every 5×5 pixels of the two rasters, such that each pixel of the SLOPE and INTERCEPT will hold the regression slope and intercept value obtained from linear regression of the corresponding 5×5 pixels that surround that pixel.

Application

The result will be use as input to generate flood probability when the daily rainfall forecast is available.

Data

Indonesia data are provided for example analysis, both data came from below.

• Daily rainfall estimate. 39 years (1981-2019) daily rainfall data used in the analysis are downloaded from Climate Hazards Center - UC Santa Barbara (https://chc.ucsb.edu/data-sets/chirps), and
• Monthly Water History. This Monthly History collection holds the entire history of water detection on a month-by-month basis. The collection contains 430 images, one for each month between March 1984 and December 2019. Each pixel was individually classified into water / non-water using an expert system and the results were collated into a monthly history for the entire time period and two epochs (1984-1999, 2000-2019) for change detection. Downloaded from Google Earth Engine: https://developers.google.com/earth-engine/datasets/catalog/JRC_GSW1_2_MonthlyHistory

WDay1_03.tif is the maximum 1-day rainfall value in March for the period 1981 - 2019, and

gi03.tif is result from ln(x/(1-x)), where x is the percentage of historical flood (pixel with water) occurrence in March for the period 1984-2019, calculated based on Monthly Water History data. How to derive the percentage of historical flood occurrence, can use the following GEE script: https://github.com/wfpidn/GEE/blob/master/script/frequency-pixel-with-water.js

Contents

• Example of raster file (GeoTIFF).
  • input
  • output
• R script for the focal-regression.
• Map output in PNG format
  • img

Example output

• Slope map.
• Intercept map.
• Pvalue for slope.
• Pvalue for intercept.

References

• https://gis.stackexchange.com/questions/278979/linear-regression-between-every-3×3-pixels-between-two-rasters-using-r/278985#278985

Contact

Using above reference, the code re-write by Anggita Annisa - former VAM intern. If you have any question related to this tool and application, contact Benny Istanto