Introduction

This post illustrates how to use Karttur’s GeoImagine Framework for calculating cross correlations between image time series data and a global index. Cross correlation between two image time series is covered in another post (to be completed).

Prerequisites

You must have setup the Karttur’s GeoImagine Framework as described in earlier posts. You must also have added the monthly TRMM rainfall dataset. Alternatively you can use any other time series dataset with regular (e.g. monthly) intervals. And you must have imported the at least one of global clmate indexes as described in this post.

Crosscorrelation

The crosscorr uses a mdofieid version of The autocorrelation of a time series analysis the correlation between the the signal at a given time, and how that correlates with the signal at different lags (or later times). The results reveal cyclic (e.g. seasonal) patterns and the temporal dependencies. This inforamtion can be used both for adjusting the time series data and for improving forecast models.

Framework process

The process for analysing autocorrelation in Karttur’s GeoImagine Framework is autocorrelate. The autocorrelation can be set to either a full autocorrelation function (acf) or a partial acf (pacf), as defined in the statsmodel package.

XML commands

The example below shows how to run acf (default) for a time series of 20 years on monthly TRMM rainfall data. The number of lags (nlags) to compute must be given. If mirror is set, the integer value will be used for adjusting seasonality to an annual cycle, and the nlags must be set to the total number of season.