2.1 On Windows

1. Install and load the gWidgetsRGtk2 package:

   install.packages("gWidgetsRGtk2")
   library(gWidgetsRGtk2)

   Upon loading the package, an error window will probably appear. Don’t worry! This is just signaling that libatk-1.0-0.dll is missing from your system. This is due to the fact that library “GTK+” is not yet installed on your system and needs to be installed. To do so, press “OK”. A new window dialog window will appear, asking if you want to install “GTK+”. Select “Install GTK+” and then “OK”. Windows will download and install the GTK+ library. When it finishes, the RSession will be restarted and you should be ready to go!¹

2. Install MODIStsp package from GitHub. (You’ll need to have the “devtools” package installed and loaded)

   install.packages("devtools")
   library(devtools)
   install_github("lbusett/MODIStsp")

2.2 On Linux systems

1. Install the following required dependencies:
   • Cairo \(\ge\) 1.0.0, ATK \(\ge\) 1.10.0, Pango \(\ge\) 1.10.0, GTK+ \(\ge\) 2.8.0, GLib \(\ge\) 2.8.0 (required by package RGtk2)
   • Curl (required by package curl)
   • GDAL \(\ge\) 1.6.3, PROJ.4 \(\ge\) 4.4.9 (required by package rgdal)

   On Debian and Ubuntu-based systems, to install packages open a terminal and type

   sudo apt-get install r-cran-cairodevice r-cran-rgtk2 libcurl4-openssl-dev libgdal-dev libproj-dev

2. From R install the libraries gWidgetsRGtk2 and devtools:

   install.packages(c("devtools","gWidgetsRGtk2"))

3. Install MODIStsp package from GitHub (you’ll need to have the “devtools” package loaded):

   library(devtools)
   install_github("lbusett/MODIStsp")

2.3 Dependencies

MODIStsp exploits functionalities of several other “R” packages. In particular, the following packages are imported:

bitops (\(\ge\) 1.9.6), data.table (\(\ge\) 1.9.6), gdalUtils (\(\ge\) 2.0.1.7), gWidgetsRGtk2 (\(\ge\) 0.0-54), hash (\(\ge\) 2.2.6), plyr (\(\ge\) 1.8.3), raster (\(\ge\) 2.5-2), RCurl (\(\ge\) 1.95-4.8), rgdal (\(\ge\) 1.1-8), rgeos (\(\ge\) 0.3-8), xts (\(\ge\) 1.0-10), sp (\(\ge\) 1.2-2), stringr (\(\ge\) 1.0.0), XML (\(\ge\) 3.98-1.1),

while the following are suggested:

knitr, rmarkdown, png, grid.

Additionaly, MODIStsp requires availability of the GDAL - Geospatial Data Abstraction Library v. >1.11.1, with support for HDF4 raster format.

3.1 Interactive mode

To run the tool in interactive mode, load the package and launch the MODIS_tsp function, with no parameters:

library(MODIStsp)
MODIStsp()

This opens a GUI from which processing options can be specified and eventually saved (or loaded).² A description of the different processing parameters to be selected is reported in the “Selection of Processing parameters” subsection.

3.2 Non-Interactive mode (Using a previously saved options file)

MODIStsp can be also launched in non-interactive mode by setting the optional “GUI” parameter to FALSE, and the “Options_File” parameter to the path of a previously saved Options file. This allows to exploit MODIStsp functionalities within generic “R” processing scripts

library(MODIStsp) 
# --> Specify the path to a valid options file saved in advance
options_file = "X:/yourpath/youroptions.RData"   from the GUI
MODIStsp(gui = FALSE, options_File = options_file)

Specifying also the “spatial_file_path” parameter overrides the output extent of the selected Options File. This allows to perform the same preprocessing on different extents using a single Options File, by looping on an array of spatial files representing the desired output extents.

For example:

# Create a character array containing a list of shapefiles (or other spatial files)
extent_list = list.files("X:/path/containing/some/shapefiles/", "\\.shp$")  

# loop on the list of spatial files and run MODIStsp using each of them to automatically 
# define the output extent (A separate output folder is created for each input spatial file).

for (single_shape in extent_list) {
  MODIStsp(gui = FALSE, options_File = "X:/yourpath/youroptions.RData", 
          spatial_file_path = single_shape )
}

4.1 MODIS Product, Satellites and Layers

Allows to select the MODIS product of interest from a drop-down menu. The user can also select which MODIS sensors should be considered for download and creation of the time series (“Terra”, “Aqua” or both). After selecting the product, pushing the “Select Layers” button opens the “Select Processing Layers” GUI panel (Figure 2), from which the user must select which MODIS original and/or derived QI and SI layers should be processed:

The “Select Processing Layers” GUI

1. The left-hand frame allows to select which original MODIS layers should be processed
2. The central frame allows to select which Quality Indicators should be extracted from the original MODIS Quality Assurance layers.
3. For MODIS products containing surface reflectance data, the right-hand frame allows to select which additional Spectral Indexes should be computed.³

Some of the most commonly used Spectral Indexes are available for computation by default (Table II).

Users can however specify other SIs to be computed without modifying MODIStsp source code by clicking on the “Add Custom Index” button, which allow to provide info related to the new desired SI using a simple GUI (Figure 3) interface.

The GUI for insertion of additional Spectral Indexes

Provided information (e.g., correct bandnames, computable formula, etc…) is automatically checked, and the new index added in the list of available ones for all products allowing its computation at the next MODIStsp execution (Figure 4).

4.2 Processing Period

Allows to specify the starting and ending dates (dd/mm/yyyy) to be considered for the creation of the time series.

4.3 Spatial Extent

Allows to define the area of interest for the processing. Two main options are possible: .

1. Full Tiles Extent: the user must specify which MODIS tiles he would like to process using the “Start” and “End” horizontal and vertical sliders in the “Required MODIS Tiles” frame. During processing, data from the different tiles is mosaiced, and a single file covering the total area is produced for each acquisition date (Note: pressing the “show map” button, a representation of the MODIS tiles grid is shown to facilitate the selection).

2. Resized: the user can specify the spatial extent of the desired outputs either;

1. Manually inserting the coordinates of the Upper Left and Lower Right corners of the area of interest in the “Bounding Box” frame. Coordinates of the corners must be provided in the coordinate system of the selected output projection.

2. pressing the “Load Extent from a Spatial File” and selecting a raster or vector spatial file. In this case, the bounding box of the selected file is retrieved, converted in the selected output projection, and shown in the “Bounding Box” frame.Required input MODIS tiles are also automatically retrieved from the output extent, and the tiles selection sliders modified accordingly.

4.4 Reprojection and Resize

Allows to specify the options to be used for reprojecting and resizing the MODIS images. In particular:

1. The “Output Projection” menu allows to either select one of the pre-defined output projections or specify a user-defined one by selecting “User Defined” and then inserting a valid “Proj4” string in the pop-up window. Validity of the Proj4 string is automatically checked, and error messages issued if the check fails.

2. The “Output Resolution”, “Pixel Size” and “Reprojection Method” menus allow to specify whether output images should inherit their spatial resolution from the original MODIS files, or be resampled to a user-defined resolution. In the latter case, output spatial resolution must be specified in the measure units of the selected output projection. At the moment, the resampling method can instead be chosen among “Nearest Neighbour” and “Mode” (Useful for downsampling purposes). Other resampling methods (e.g., bilinear, cubic) are not currently supported since i) they cannot be used for resampling of categorical variables such as the QA and QI layers, and ii) using them on continuous variable (e.g., reflectance, VI values) without performing an a-priori data cleanind would risk to contaminate the values of high-quality observations with those of low-quality ones.

4.5 Processing Options

Allows first of all to specify the format desired for the output images. Two of the most commonly formats used in remote sensing applications are available at the moment: ENVI binary and GeoTiff. If GeoTiff is selected, the type of file compression can be also specified among “None”, “PACKBITS”, “LZW” and “DEFLATE”.

The user can then specify if virtual multitemporal files should be created. These virtual files allow access to the entire time series of images as a single file without the need of creating large multitemporal raster images. Available virtual files formats are ENVI metafiles and GDAL “vrt” files. Additionally, the user can select if he desires to save the time series also as “R” rasterStack objects (with temporal information added through the “setZ” method of the raster package). This may be useful in order to easily access the preprocessed MODIS data within “R” scripts.

Finally, users can select if the NoData values of MODIS layers should be kept at their original values, or changed to those specified within the “MODIStsp_Products_Opts” XML file. By selecting “Yes” in the “Change Original NODATA values” checkbox, NoData of outputs are set to the largest integer value possible for the data type of the processed layer (e.g., for 8-bit unsigned integer layers, NoData is set always to 255, for 16-bit signed integer layers to 32767, and for 16-bit unsigned integer layers to 65535). Information about the new NoData values is stored both in the output rasters, and in the XML files associated with them.

4.6 Main Output Folder for Time Series Storage

Allows to specify the main folder where the pre-processed time series data will be saved stored. The “Reprocess Existing Data” checkbox allows to specify if images already available should be reprocessed if a new run of MODIStsp is launched with the same output folder. If set to “No”, MODIStsp skips dates for which output files following the MODIStsp naming conventions are already present in the output folder. This allows to incrementally extend MODIS time series without reprocessing already available dates.

4.7 Output Folder for Original HDF Storage

Allows to specify the folder where downloaded original MODIS HDF files are stored. The “delete original HDF files” checkbox allows to specify if the downloaded images should be deleted from the file system at the end of the processing. To avoid accidental file deletion, this is always set to “No” by default, and a warning is issued before execution whenever the selection is changed to “Yes”.

8.1 Adding links to desktop/Start menu for standalone execution

Links to the MODIStsp.bat or MODIStsp.sh standalone launchers can be created automatically launching from R the function MODIStsp_install_launcher()

• In Linux: this creates a desktop entry (accessible from the menu in the sections “Science” and “Geography”), and a symbolic link in a known path (default: /usr/bin/MODIStsp). If the path of the symbolic link is included in the user PATH variable, the standalone execution can be done simply calling MODIStsp -g -s "/yourpath/youroptions.RData".

• In Windows:: this creates a link in the Start Menu and optionally a desktop shortcut.

See ?install_MODIStsp_launcher for details and path customisations.

9.1 Installing R

sudo apt-get install r-base

9.2 Installing GDAL \(\ge\) 1.11.1