Urban growth modelling and prediction using markov chain and cellular automata

Sir, I need to create a kernel of size 3X3 to run it over a lulc classified image of time 't1' treating the lulc as a matrix of nXm size. After which it is expected to generate a new matrix of same size as that of out base matrix, the new matrix will be our lulc image of time 'T2' in future. The kernel should contain properties of Markov Chain and Cellular Automata and should give weightage to any of the parameters( distance from roads, population density, distance from CBD or DEM) or all of them as per our convenience. The base matrix which I have got values ranging from 1 to 4. It shows different classes like 1-vegetation, 2-water body, 3-built-up and 4-other classes. We are interested only on the built-up.

The kernel should be intelligent enough to logically change the values of classes, mainly class 3. The kernel should work on each cell one by one considering it's neighboring cells. The no. of cells of different classes falling in the 3x3 zone of any particular cell should be taken care of. Like how many pixels of class built-up are there around a cell or the class vegetation or others.

The issue names no repository files, tests, or implementation entry point. Start by clarifying the expected Markov-chain and cellular-automata behavior, the input and output matrix format, and how built-up cells and optional parameters should be weighted. Done would require an agreed implementation scope and validation against expected future land-use results.

Written by the indexing model from the issue text.