Skip to content

Herds Activity Mapping and Analytical Classification routine ‒ an HMM-based GPS data processing routine

License

Notifications You must be signed in to change notification settings

ScriBanana/HAMAC

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

HAMAC Routine

The Herds Activity Mapping and Analytical Classification (HAMAC) routine prepares, classifies into activities and produces variouss representations of GPS and accelerometer data gathered from tracking livestock to train behaviour models and exploit the resulting data in diverse ways. It makes use of parallel computing and leverages the Hidden Markov Model framework to classify activities. The later part relies heavily on the moveHMM R package

Citation

To cite this routine, please use the following:

Scriban, A., Nabeneza, S., Cornelis, D., Salgado, P., 2024. Herds Activity Mapping and Analytical Classification. Cirad, Montpellier, France. https://gitlab.cirad.fr/selmet/hamac.

Usage

The routine is divided into functional blocks, each containing individual script pages, each corresponding to a unitary operation. The order of the groups and script pages is sequential, yet all elements are optional. Intermediate data saves within each group allow modular execution, as required. Here is a breakdown of what each group and individual script does:

A. Data cleaning and formatting

A1. Scans a directory to load all GPS and Accelerometer data files from tracking

A2. Cleans and formats GPS data

A3. Adds additional environmental variables to the GPS data

A4. Cleans and formats accelero data

A5. Graphical representation of the temporal data coverage

B. Data association and preparation

B1. Associates GPS data to the animals

B2. Associates GPS and accelerometer data

B3. Computes trajectory metrics to fit HMMs on

C. Model training

C1. Preliminary script to load the fit function and the data and prime the log

C2. Runs a single fitting process on a given parameter set

C3. Parameter space exploration and sampling of fits to ensure the optimal likelihood

D. Exploiting model outputs

D1. Quality of fit and trajectory metrics

D2. Plots of the density of probability for trajectory metrics

D3. Simulating virtual mobility

E. Exploiting activity data

E1. Preliminary script to merge GPS data and activity from a model

E2. Graphical representations of the temporal repartition of activities

E3. Diverse behavioural metrics

E4. Associates position and activity data to land-use data

E5. Repartition of activities over land-use and seasons

License

This code is made open-source under the CeCILL-B licence.

Relevant publication and datasets

Scriban, A., Nabeneza, S., Cornélis, D., Delay, É., Vayssières, J., Cesaro, J.-D., Salgado, P., 2024. GPS-based hidden Markov models to document pastoral mobility in the Sahel. Sensors. Submitted.

Scriban, Arthur; Nabeneza, Serge; Salgado, Paulo, 2024, "Mobility, behaviour and land-use data from GPS tracking cattle herds in Sahel agropastoral systems", https://doi.org/10.18167/DVN1/GHJKQO, CIRAD Dataverse

About

Herds Activity Mapping and Analytical Classification routine ‒ an HMM-based GPS data processing routine

Topics

Resources

License

Stars

Watchers

Forks

Languages