Version 2.1 (February 2010)
Some users wrote to me, regretting that they were unable to use the
information in StatusMaps (water/land pixels, ice/snow pixels, cloudy
pixels) in their current bitwise form. I thought therefore to add an option
to extract any desired combination of masks from any number of StatusMap
After extracting it, a mask can be displayed using a graphic utility
(DISPLAY) excerpted from ADDAPIX for Win32 (WINADX), an application that I
am preparing, and will be released soon.
DISPLAY is also used to visualise the location of the Region to be
In general, use masks with prudence: they are generated
automatically from the values of the four radiometric bands, and
sometimes this appears to lead to unreliable and bizarre results, like ice/snow
pixels found around some great lakes in Central Africa(!). See the help on
line for some pictures (but I hope to have done something wrong and that things
are actually not like that...).
Two utilities have been added to derive from the .TXT documentation files
associated to the cropped images the text files necessary to load BIL images
in ArcView/ArcGis or ENVI. Two more utilities enable the conversion of BIL
images to the WINDISP format, or vice versa. Their purpose
is to change the format of the images and the text files that describe them
without needing to re-crop the Regiion of Interest from the original zipped
VGT distribution files.
This version also
fixes a problem with 2010 VGT distribution files, refused by the
Actually, when I wrote CropVgt I did not think it would be used for so long,
so the correctness of input filenames was checked only up to year 2009!
Version 2.0 (November 2008)
While working on a new version of Addapix for Win32 (Principal Components
Analysis and Clustering on a set of satellite images), it occurred to me
that some functions conceived for that application could also be useful to CROPVGT
I decided to spend some time adapting them, and here they are.
1. Any number of NDVI images forming a time series, cropped
from VGT continental images, can be improved by suitably correcting the time series
of each pixel. Unreliably low values caused by mist or partial cloudiness are
corrected, missing values are simulated, and the new corrected
images thus obtained are saved.
2. From these corrected images 36 dekadal long-term means
images (historical averages) and 36 dekadal standard deviations images (measuring
the inter-annual variability) are computed.
The distribution zipped file includes a sample time series of 405
dekadal images (April 1, 1998 through June 3, 2009) of a Region around
Lake Tana, Ethiopia, cropped (in one go!) from the NDVI VGT continental
images of Africa. They are used for an exercise on correction/computation of long-term
statistics. A PDF file of instructions for that exercise is also included.
- in case of multiple extraction only one file .TXT, that describes the
is saved to file. Before there was one TXT file for each image
which needlessly multiplied
the number of the TXT files.
- when processing NDVI images, if the requested output
WINDISP format the code indicated
for clouded pixels is now correctly saved in the header of the
added support for processing global
NDVI images, available on the free VGT site starting from August 2006.
Thanks to dr. Sander Zwart for bringing the use of these images to my
same as version 0.8, with the only
difference that the number of decimals declared in the header of output
images saved in WINDISP format has been changed from 0 to 3. This
allows the user to compute correctly some areal statistics using WINDISP.
At the time
being, S10 and D10 products (ten-day MVC and BiDirectional Reflectance
syntheses respectively) can be freely downloaded from the SPOT4/VEGETATION
available files are zipped, and each of them covers a whole continent, or
part thereof: they are therefore quite large. Besides, the planes included
in each zipped file are in HDF format (Hierarchical Data Format) and it is
not immediate to interpret their headers or point to the beginning of the
bitmap to extract it, or to crop a sub-region.
After spending quite a lot of time creating batch commands to do the job (so
easy to overlook something!), I decided to write a Win32 program to do it in
a fully automatic way. Here is my
contribution to the VEGETATION users' community...
can be used to extract a user-defined Region of
Interest from any number of files downloaded from the free VGT site.
A time series of cropped
NDVI images (no limit on their number) can then be improved by interpolating
through the maxima of the series of each pixel, and simulating reasonable
values for missing ones. Using the corrected images thus obtained, long-term
means and standard deviations can be computed by pixel and dekad, taking
into account all the available years.
The user selects:
any number of zipped VGT files
to be processed; they can reside in different folders, but must
all refer to the same type of product (S10 or D10) and the same continent;
the output folder where
the images being cropped are to be saved;
the bands to be
the Region of Interest to be extracted from all the
files and all the bands indicated. This is
done by entering the longitude/latitude of the RoI's upper-left and
to have the clouds
automatically masked in NDVI images, deriving the necessary information
from the Status Map.
scans one by one all the user-chosen VGT zipped files. From each input
file, the bands selected by the user are extracted. For each band, the HDF
header and the accompanying LOG file are read, then the Region of Interest
is cropped and saved to a file named <BBBYYMMDREG.EXT>,
BBB' is a group of two or three
letters indicating the band;
' YY '
are the year's last two digits;
stands for the month ('01' to '12');
stands for the dekad ('1' to '3');
REG ' stands
for up to three letters entered by the user to label the RoI.
' EXT' is an extension
that depends on the output image format chosen by the user. It is:
if they are saved as pure binary bitmaps, accompanied by
suitable documentation files to enable their
loading in ArcView/ArcGis, or ENVI, or other applications;
if they are saved in WINDISP format: a binary bitmap, preceded by a
512-byte binary header.
Optionally, in NDVI images cloudy pixels can be masked by assigning them a code
indicated by the user.
The output images
accompanied by a documentation text file, having the same filename and
How to install
the new version 2.1.
It comes as a zipped file: unzip it anywhere on your hard disk (please,
not on the desktop!). It creates a folder <cropvgt>
with all the necessary software and data. It includes a service
(do not tamper with it!) and a sub-folder <Tana>
(see below). To run the application, double-click on the executable
The Windows Registry is not modified, so no
installation is actually necessary. You can optionally edit the file
CROPVGT.INI to change some parameters loaded by the programme at startup.
programme includes a help that offers some info derived from the
VEGETATION FAQ. The paper by Duchemin and others, presented at the
2000 Belgirate Conference, on which the D10 BDR products are based, is
programme's directory <cropvgt> contains a folder < tana >, with
405 NDVI images cropped around
Lake Tana, Ethiopia. Each image is 225x225 pixels (2°x2°). The file instructions.pdf provides the
instructions for an exercise on image correction, and computation of
For an example of how the
long-term means can be used to construct an eco-climatic classification or
to monitor the evolution of the cropping season,
download a worked example on Sudan 2005
(PPT presentation, data and software: 170 MB all together).
Problems with online
At the time being
CROPVGT offers only an online help in WINHELP
(*.HLP) format, which is no longer supported by Windows Vista.
Until a .CHM help file is ready, users of Vista can
dowload and install WINHELP pushing the button
Thanks to the Info-ZIP
group for making Unzip32.dll
is derived from AICON (the Artemis Image CONverter), a program that
I wrote for FAO/ARTEMIS, available from
a dynamic link library developed during my very slow (still ongoing)
attempt to convert ADDAPIX into a Win32 package.
ADDAPIX (Data Analysis
pixel-by-pixel), also partially funded by FAO/ARTEMIS, is a DOS package, quite old and no longer supported.
However, a Windows 32 version (Addapix for
Windows, or WINADX) is being prepared
and will soon be ready.
Feel free to link this
page, or copy the installation file to your site, as long as the author's
name goes along. Please, do not omit to offer your suggestions and to
inform me of the bugs you may find.
(page maintained by S. Griguolo -
email: email@example.com - last update: February 2010)