DGVMTools-package: An overview of the DGVMTools package

In MagicForrest/FireMIPTools: FireMIP extensions for DGVMTools

An overview of the DGVMTools package

Description

This package is designed for reading, processing and plotting output from Dynamic Vegetation Models (DGVMs), land surface models from climate models and other spatial representions of the terrestrial biosphere or land surface. There are many such models and each have their own output format. This package gives a framework for reading the different outputs from these models and putting them into a common internal representation. Once this is done, it provides many tools for analysing the model results and comparing them to data (and each other). These include common tasks such as:

• Subsetting and averaging certain time periods or spatial extents,

• Calculating total values for certain seasons, growth forms (for example all trees), leaf phenologies (for example all evergreen vegetation), climate zones (for example all tropical vegetation), and others,

• Making maps and time series plots of the model output,

• Reading in observed data (satellite data, flux tower data etc.) to which model the model results can be compared,

• Calculating statistical metrics to quantify the agreement between models and data, and producing plots (scatter plots, histograms of residuals) to visually compare models and data,

• Making a classification biomes or vegetation types from the model output and plotting them,

• Making differences maps to evaluate the effects of a model change or spatially comparing models to data,

• Storing averaged model data on disk for fast re-reading,

• Saving model data in standard netCDF format.

Details

There is obviously some overlap in functionality with the raster and sp R packages. Indeed this package builds on raster and sp functionality. The advantages of DGVMTools over only raster and sp for analysing DGVM output are two-fold. Firstly, it is tailored for a typical DGVM analysis workflow. Concepts like 'model runs' and 'plant functional types' are explicit objects with their own meta-data (if you don't know what these concepts are then this package probably isn't for you). Once these objects are correctly defined (which is not difficult), analysis is very convenient. Many common tasks are already coded efficiently into functions, and because of the meta-data attached to the objects, these functions can do a lot of 'sensible and standard' stuff without too much direction from the user. Secondly, the data is stored internally as a data.table (as opposed to a data.frame). The advantage of this is that data.tables are very, very much faster than data.frames for many operations (check out the data.table package documentation and webpage for more info). This is obviously a great advantage when working with very large spatial-temporal datasets.

Furthermore, the objects defined in DGVMTools can be very easily converted into rasters or data.frames, and so can fit directly back into any existing R code, packages or other machinery. This the transition to using DGVMTools is very smooth. DGVMTools can be used to read the data and perform standard operations, but then data can be converted to a raster or data.frame for more specific or idosyncratic R scripts or functions. However, DGVMTools is intended to be a reasonably complete analysis environment in itself. It should be possible to go from model output all the way to results and publication quality plots using only DGVMTools and some base R functionality for other tasks.

MagicForrest/FireMIPTools documentation built on April 22, 2023, 4:24 a.m.