0

i am fairly new to Moran's I and am currently trying to find out if it's even applicable to my data.

I have a collection of binary rasters showing the spread of Bark Beetles between 2018 and 2024. All pixels with value = 1 show infested/disturbed tree whereas the 0s show healthy trees.

I used global Moran's I to test out if my infestation areas are clustered and get I values are 0.8 with p <0.001, which i assumed indicate a high concentration of spatial clustering.

I have calculated LISA clusters for each year, which look like this: enter image description here

My objective is to find out if it is possible to extract a statistical pattern throughout my time series, i.e., that infestations in year t cluster around those infestations in year t-1. I tried bivariate Moran from the ESDA python library and my results looks like this:

2025-08-07 → 2025-08-07 : I = 0.109, p = 0.001, z = 201.11

2025-08-07 → 2025-08-07 : I = 0.032, p = 0.001, z = 57.69

2025-08-07 → 2025-08-07 : I = 0.145, p = 0.001, z = 255.65

2025-08-07 → 2025-08-07 : I = 0.091, p = 0.001, z = 168.29

2025-08-07 → 2025-08-07 : I = 0.119, p = 0.001, z = 221.69

2025-08-07 → 2025-08-07 : I = 0.083, p = 0.001, z = 160.46

Is Moran's I a good approach for my problem or is there a better-suited option?

The results don't seem to bad, however the z values seem to be very large...

Improve this question
4
  • What do you mean by "cluster around"? Are you trying to test for some spatial spread of infections from time t to time t+1?
    – Spacedman
    Commented Jul 23 at 14:19
  • yes exactly, i want to find out if the infections in t+1 happen around those from time t, or if they appear elsewhere
    – Sebi
    Commented Jul 24 at 7:29
  • This might be best treated like an SIR disease process. Cells are either Susceptible, Infected (ie can spread to nearby cells), or Removed (doesn't spread the disease and can't be infected again). You are then interested in if the S->I transition probability is dependent on number of I cells in the neighbourhood. There's more details needed first, like I don't know if cells are removed (eg do the beetles kill all the trees?) or if "infected" just means presence of at least one infected tree in a cell... There's too much to discuss in a comment though...
    – Spacedman
    Commented Jul 24 at 8:18
  • that approach makes sense, yes my original data was based on sentinel 2 and collected with a ML approach, so it would mean that the spectral signal of that pixel changed, i cannot detect individual trees my rasters only contain new infestations, infected pixels from previous years are not carried over
    – Sebi
    Commented Jul 24 at 9:02

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.