modisfast

Table of contents

• Overview
• Installation
• Get started
• Data collections available
• Manual testing of the functionality
• Foundational framework
• Comparison with similar R packages
• Citation
• Future developments
• Contributing
• Acknowledgments

Overview

modisfast is an R package designed for easy and fast downloads of MODIS Land products, VIIRS Land products, and GPM (Global Precipitation Measurement Mission) Earth Observation data.

modisfast uses the abilities offered by the OPeNDAP framework (Open-source Project for a Network Data Access Protocol) to download a subset of Earth Observation data cube, along spatial, temporal or any other data dimension (depth, …). This way, it reduces downloading time and disk usage to their minimum : no more 1° x 1° MODIS tiles with 10 bands when your region of interest is only 30 km x 30 km wide and you need 2 bands ! Moreover, modisfast enables parallel downloads of data.

This package is hence particularly suited for retrieving MODIS or VIIRS data over long time series and over areas, rather than short time series and points.

Importantly, the robust, sustainable, and cost-free foundational framework of modisfast, both for the data provider (NASA) and the software (R, OPeNDAP, the tidyverse and GDAL suite of packages and software), guarantees the long-term reliability and open-source nature of the package.

By enabling to download subsets of data cubes, modisfast facilites the access to Earth science data for R users in places where internet connection is slow or expensive and promotes digital sobriety for our research work.

Installation

You can install the released version of modisfast from CRAN with :

install.packages("modisfast")

or the development version (to get a bug fix or to use a feature from the development version) with :

if(!require(devtools)){install.packages("devtools")}
devtools::install_github("ptaconet/modisfast")

Get Started

Accessing and opening MODIS data with modisfast is a simple 3-steps workflow. This example shows how to download and import a one-year-long monthly time series of MODIS Normalized Difference Vegetation Index (NDVI) at 1 km spatial resolution over the whole country of Madagascar.

1/ First, define the variables of interest (ROI, time frame, collection, and bands) :

# Load the packages
library(modisfast)
library(sf)
library(terra)

# ROI and time range of interest
roi <- st_as_sf(data.frame(id = "madagascar", geom = "POLYGON((41.95 -11.37,51.26 -11.37,51.26 -26.17,41.95 -26.17,41.95 -11.37))"), wkt = "geom", crs = 4326) # a ROI of interest, format sf polygon
time_range <- as.Date(c("2023-01-01", "2023-12-31")) # a time range of interest

# MODIS collections and variables (bands) of interest
collection <- "MOD13A3.061" # run mf_list_collections() for an exhaustive list of collections available
variables <- c("_1_km_monthly_NDVI") # run mf_list_variables("MOD13A3.061") for an exhaustive list of variables available for the collection "MOD13A3.061"

2/ Then, get the URL of the data and download them :

## Login to Earthdata servers with your EOSDIS credentials.
# To create an account (free) go to : https://urs.earthdata.nasa.gov/.
log <- mf_login(credentials = c("username", "password")) # set your own EOSDIS username and password

## Get the URLs of the data
urls <- mf_get_url(
  collection = collection,
  variables = variables,
  roi = roi,
  time_range = time_range
)

## Download the data. By default the data is downloaded in a temporary directory, but you can specify a folder
res_dl <- mf_download_data(urls, parallel = TRUE)

3/ And finally, import the data in R as a terra::SpatRaster object using the function mf_import_data() ( :warning: see here why you should use this function, instead of the original terra::rast(), in the context of modisfast) :

r <- mf_import_data(
  path = dirname(res_dl$destfile[1]),
  collection = collection,
  proj_epsg = 4326
)

terra::plot(r, col = rev(terrain.colors(20)))

et voilà !

Want more examples ? modisfast provides three long-form documentations and examples to learn more about the package :

• a “Get started” article describing the core features of the package;
• a “Get data on several regions or periods of interest simultaneously” article detailing advanced functionalities of modisfast (for multi-time frame or multi-regions data access);
• a “Full use case” article showcasing an example of use of the package in a scientific context (here: landscape epidemiology).

Collections available in modisfast

Currently modisfast supports download of 77 data collections, extracted from the following meta-collections :

• MODIS land products made available by the NASA / USGS LP DAAC ( :arrow_right: source OPeNDAP server) ;
• VIIRS land products made available by the NASA / USGS LP DAAC ( :arrow_right: source OPeNDAP server)
• Global Precipitation Measurement (GPM) made available by the NASA / JAXA GES DISC ( :arrow_right: source OPeNDAP server).

Details of each product available for download are provided in the tables below or through the function mf_list_collections().

Manual testing of the functionality

Since most modisfast functions depend on EarthData credentials, automated tests are disabled. However, after installation, users can manually test the package’s functionality by running these lines of code :

# replace "username" and "password" with your own EOSDIS (Earthdata) credentials 
earthdata_un <- "username" 
earthdata_pw <- "password"

devtools::test("modisfast")

Foundational framework

Technically, modisfast is a programmatic interface (R wrapper) to several NASA OPeNDAP servers. OPeNDAP is the acronym for Open-source Project for a Network Data Access Protocol and designates both the software, the access protocol, and the corporation that develops them. The OPeNDAP is designed to simplify access to structured and high-volume data, such as satellite products, over the Web. It is a collaborative effort involving multiple institutions and companies, with open-source code, free software, and adherence to the Open Geospatial Consortium (OGC) standards. It is widely used by NASA, which partly finances it.

A key feature of OPeNDAP is its capability to apply filters at the data download process, ensuring that only the necessary data is retrieved. These filters, specified within a URL, can be spatial, temporal, or dimensional. Although powerful, OPeNDAP URLs are not trivial to build. modisfast facilitates this process by constructing the URL based on the spatial, temporal, and dimensional filters provided by the user in the function mf_get_url().

Let’s take an example to understand.

The following URL :arrow_down:

https://opendap.cr.usgs.gov/opendap/hyrax/MOD11A2.061/h17v08.ncml.nc4?MODIS_Grid_8Day_1km_LST_eos_cf_projection,LST_Day_1km[775:793][55:140][512:560],LST_Night_1km[775:793][55:140][512:560],QC_Day[775:793][55:140][512:560],QC_Night[775:793][55:140][512:560],time[775:793],YDim[55:140],XDim[512:560]

is a link to download the following subset of MOD11A2.061 data in netCDF :

• bands LST_Day_1km, LST_Night_1km, QC_Day, QC_Night ;
• each available date between the 2017-01-01 and the 2017-06-01 ;
• within the following bounding box (lon/lat): -5.41 8.84, -5.82 9.54.

The indices within the [] refer to values encoding for the spatial and temporal filters.

These OPeNDAP URLs are not trivial to build. modisfast converts the spatial, temporal and dimensional filters (R objects) provided by the user through the function mf_get_url() into the appropriate OPeNDAP URL(s). Subsequently, the function mf_download_data() allows for downloading the data using the httr and parallel packages.

Comparison with similar R packages

There are other R packages available for accessing MODIS data. Below is a comparison of modisfast with other packages available for downloading chunks of MODIS or VIIRS data :

* at the downloading phase

** Take a look at the article “Comparison of performance with other similar R packages” to get an overview of how modisfast compares to these packages in terms of data access time.

Citation

This package is licensed under a GNU General Public License v3.0 or later license.

We thank in advance people that use modisfast for citing it in their work / publication(s). For this, please use the following citation :

Taconet, P. & Moiroux N.(2024). modisfast: Fast and Efficient Access to MODIS Earth Observation Data. In CRAN: Contributed Packages. The R Foundation. https://doi.org/10.32614/cran.package.modisfast

Future developments

Future developments of the package may include access to additional data collections from other OPeNDAP servers, and support for a variety of data formats as they become available from data providers through their OPeNDAP servers. Furthermore, the creation of an RShiny application on top of the package is being considered, as a means of further simplifying data access for users with limited coding skills.

Contributing

All types of contributions are encouraged and valued. For more information, check out our Contributor Guidelines.

Please note that the modisfast project is released with a Contributor Code of Conduct. By contributing to this project, you agree to abide by its terms.

Acknowledgments

We thank NASA and its partners for making all their Earth science data freely available, and implementing open data access protocols such as OPeNDAP. modisfast heavily builds on top of the OPeNDAP, so we thank the non-profit OPeNDAP, Inc. for developing the eponym tool in an open and collaborative way.

We also thank the contributors that have tested the package, reviewed the documentation and brought valuable feedbacks to improve the package : Florian de Boissieu, Julien Taconet.

This work has been developed over the course of several research projects (REACT 1, REACT 2, ANORHYTHM and DIV-YOO) funded by Expertise France, the French National Research Agency (ANR), and the French National Research Institute for Sustainable Development (IRD).