Previous PageNext Page

2.7. Analysis

For all image analysis executed by WinDisp3.5, you must supply maximum and minimum pixel values. By default the threshold values are 0-0; that is, they adapt automatically to the maximum and minimum values of the image. All pixel counts are included in the processing. To change the thresholds, use the command Process Thresholds. For details concerning this command, see Chapter 3.

2.7.1. File lists

Several functions of WinDisp3.5 use a list of one or more files for entry. Use of a file list avoids manual entry of names of files. An ASCII file, with values delimited by commas, contains a list of file names with one name per line with headings for rows and columns. Here is an example of a file list used to graph pixel values in a series of images:

The functions using a file list are the following:

Consult Chapter 3 for more details concerning these functions.
To create a file list directly inside a batch, use the command Batch Build List.

A special dialog box has been created to facilitate creating and editing these file lists.

By clicking on the browse button of any dialog box, the file list editor appears for each parameter requiring a file list.

 

A - File list editor Menu

File

   
 

New

Create a new empty list, with number of rows and columns defined

 

Open

Open an existing file list

 

Save

Save the active list in the active file list

 

Save As

Save the active list in a new file list

 

Exit

Close the window for creating lists

     

Rows

   
 

Edit names

Edit names of rows

 

Stop Edit Names

Exit the editor

 

Add

Add empty rows to the list

 

Remove

Remove the selected row from the list

     

Columns

   
 

Edit names

Edit names of columns

 

Stop Edit Names

Exit the editor

 

Add

Add empty columns to the list

 

Remove

Remove the selected column from the list

     

Filenames

   
 

Add

Select file names from the dialog box and insert them in the list

 

Remove

Remove selected filenames

     

Help

Open the Help file to the page treating file lists

B - File lists

This column is made up of a single file list which can be used to carry out statistics on the group of files listed.

C - Editing

The advantage of the file editor resides in the facility with which a list can be created or modified. The name of the list underlined in yellow can be directly changed, and the new list can be saved with the File Save or File Save As command.

2.7.2. Analysis of a series of images

This section, as well as the following two, presents the possibilities of using WinDisp3.5 in the domain of seasonal analysis. The examples are adapted for the agricultural region of the Sahel, covering the period of April to November. In particular, an extracted sample of data and the presentation in the form of a graph is presented for Burkina Faso (see sections 2.7.3. and 2.7.4.).

Statistical analysis for time series can be done by WinDisp3.5 for pixels in a series of images. The result will be presented in the form of an image in which each pixel is the result of applied analysis to the pixels in the same location in each of a series of images. The analysis can be performed for the following:

Max

Maximum value

Min

Minimum value

Avg

Average value

Range

Range between the maximum value and the minimum value

Sum

Sum of the values

Count

Number of valid pixels (situated between the maximum threshold and the minimum threshold in a polygon

Stddev

Standard deviation of values

Decloud

Temporal smoothing technique

Slope

Slope of the trend line of the values

MaxDate

Date the maximum value occurs

MinDate

Date the minimum value occurs

To use each of these commands in the Process Series menu, you must supply WinDisp3.5 with the following information: the name of the file list containig the list of images that you want analyzed, the location where you want to save the result of the analysis, and the name under which you want it saved. You will be asked to provide certain other parameters depending on the type of analysis done.

The commands Max, Min, Avg and Range are generally applied to NDVI (Normalized Difference Vegetation Index) images, in order to observe vegetation peaks, the lowest values, average conditions, or differences between maximum and minimum values during a season. See examples below. Min and Avg analysis exclude pixels having values corresponding to clouds or to an absence of data.

The command Sum is often used to create a cumulative image of CCD or ERF time series for the growing period. See example below. To use the command Sum you must supply a multiplication factor. The pixels of an image resulting from this analysis will have values equal to the sum of the corresponding pixels in the images analyzed; the multiplication factor will restore pixel values to the image, enabling it to be compared with the images analyzed. The resulting image will then be displayed using the same color table as the original images. For example, a Sum carried out on five images will have a multiplication factor of 0.2.

The command Count produces an image in which each pixel has a value equal to the number of valid pixels situated in the same position in the images of the series. For the analysis of a series of 24 images, for example, a pixel value of 22 signifies that 2 pixels in the series were considered invalid. A pixel is valid when it is not contaminated by water, clouds, or the absence of data. This applies to NDVI images. See example below.

The command Stddev (standard deviation) is also frequently used to analyze NDVI images. It produces an image representing the variation of each pixel in a series of images. In the following example, this analysis has been applied to a series of images representing the annual maximum vegetation in the Sahel. These images have been produced using Max analysis for a series of NDVI images from the first dekad of April to the third dekad of November 1982 to 1997, excluding 1991 and 1992 because the satellite images were unusable after the eruption of Pinatubo.

File list used is the following: Color table used: Image obtained:

The command Decloud is an interpolation procedure used to eliminate the interference caused by clouds in NDVI images. To apply this command you must supply WinDisp3.5 with the name of a filelist containing names of three NDVI images: the second image is the one to be corrected; the first and third are reference images that cannot contain clouds, at least in the area to be analyzed. You must also supply a percentage of negative deviation - 10% is used by default. This means that, if a pixel value in the second image deviates more than 10% from the average of the corresponding pixels of the first and third, then the pixel value of the second image will be replaced by this average.

In the following examples, the clouds of image DV97041 can be eliminated by interpolation of images DV97033 and DV97042:



The image resulting from interpolation:

The file list used:

The slope command is used to determine the slope of the values of each pixel in a series of images. The example below applies this to a series of NDVI images. For the Slope command you must enter a multiplication factor which increases the precision of the result. For example, if the multiplication factor is 1, a result equal to 100 for a given pixel indicates a null slope for the pixel; a result of 108 indicates a positive slope of 8%; and a result equal to 94 indicates a slope of negative 4%. If the multiplication factor is 10, the results for the same pixels would be 102, 184, and 38, increasing the precision of obtained results to +0.2%, +8.4%, and -3.8% respectively.

The commands DateMax and DateMin can be used on CCD images (Cold Cloud Duration) as well as NDVI, in order to determine the date of the peaks or the lowest values during a period. See the examples below. In the series of images in a file list, WinDisp3.5 picks out, for each pixel, the image that contains the highest or lowest value, and assigns to the pixel the number of this image in the file list. For example, in applying DateMax to the series of images in one of the file lists below, if a pixel is found to have the highest value in image 'dc97043.af', this pixel will have the value '12' in the resulting image. To make the resulting image readable, it will be necessary to open an adapted color table such as the DATE table reproduced below.

The following images illustrate the various types of analysis available, except Stddev and Decloud already illustrated above. The images are produced from analysis on either CCD satellite images from April to November 1997 or NDVI images covering the same period. Consult section 2.7.1 for details on file lists.

The color tables used to open images resulting from these analyses (consult section 2.4.5 for details on color tables):

Color Table C (clouds) Color Table V (vegetation) Color Table DATE

Color table PIXEL Color Table SLOPE
(adapted by a multiplication factor = 1)

The type of images and the color table used for these analyses are indicated under each image:

Maximum
Images NDVI - Color Table V
Minimum
Images NDVI - Color Table V
Average
Images NDVI - Color Table V
     

Slope
Images NDVI - Color Table V

Sum (multiplication factor = 0.042)
Images CCD - Color Table C

Count Pixels
Images NDVI - Color Table PIXEL
     

Slope (multiplication factor = 1)
Images NDVI -Color Table SLOPE

Date of Maximum
Images CCD - Color Table DATE

Date of Minimum

Images CCD - Color Table DATE

2.7.3. Process stats

With the Process Stats menu, you can extract statistics for points and polygons within the images. The results will be in the form of an ASCII table which can be used to present the data in the form of a graph. (Section 2.7.4)
The following commands are available:

Max

Maximum value

Min

Minimum value

Average

Average value

Stddev

Standard deviation of values

Range

Range of values

Count

Number of valid pixels (between the maximum and minimum thresholds) in a polygon

The example described in the next two paragraphs analyzes the average of NDVI images from the first dekad of April 1997 to the third dekad of November 1997 in the agricultural regions of the Sahel. The statistics are extracted for the provinces of Burkina Faso. In order to obtain a point of comparison for the analysis of data obtained for 1997, the same analysis is conducted on NDVI images (16 years) covering the same period.

All of the other analyses available in this menu can be done in the same way as the analysis for averaging.

A dialog box similar to the one shown at the right will appear when Process Stats is run:

The following parameters must be supplied:

A. File list of image names

You must provide the file list (.LST) containing the names of all images that you want to include in the analysis. All the images in the list must have the same header window and projection parameters, because the map masks (pixels from which statistics are extracted) are calculated only for the first image. For more information of file lists, see Section 2.7.1.
Example of Burkina Faso: the first analysis is done using the file list presented in Section 2.7.2, which includes the period from April to November 1997. The second analysis uses NDVI average images (16 years) from April to November.

B. Map file

The statistics are calculated relative to a prescribed polygon. It is necessary to provide WinDisp3.5 with the name of a correct geo-referenced map (.BNA) that can serve as the basis for statistical calculations.
Example of Burkina Faso: for each of the two analyses done, the boundary map of Burkina Faso is used as a feature reference. WinDisp3.5 will calculate the average pixel values within the borders of each province of Burkina Faso for each image on the list.

C. Stats File for results

Location and name (.STA) must be given for the table of results.
Example of Burkina Faso: the two data tables resulting from the analysis of average NDVI images of 1997 and the 16-year average NDVI are named 'ndvi97.sta' and 'ndviavg.sta' respectively.

D. Pixels around points (1,3,5,7,9)

This parameter is used to define the area around a pixel (for example a meteorological station) which can be included in the analysis of the pixel.
Example of Burkina Faso: in the case of analysis linked to some polygons this parameter is not used, so its value by default is equal to 1.

The results of the extraction are stored in an ASCII format table. The first line identifies the fields derived from the first column in each line of the list file. Each row of the table begins with the name of the cartographic feature, delimited by quotation marks, followed by the data from each image, separated by a comma and a space.
Example of Burkina Faso: statistics table 'ndvi97.sta' (the following is data represented in a graph in section 2.7.4.):

"stats", "1", "2", "3", "4", "5", "6", "7", "8", "9", "10", "11", "12", "13", "14", "15", "16", "17", "18", "19", "20", "21","22","23","24"
"Oudalan", 0.10, 0.10, 0.10, 0.09, 0.07, 0.07, 0.06, 0.07, 0.06, 0.07, 0.11, 0.11, 0.14, 0.18, 0.21, 0.20, 0.16, 0.14, 0.12, 0.11, 0.12, 0.13, 0.13, 0.14
"Soum", 0.11, 0.10, 0.10, 0.09, 0.07, 0.07, 0.07, 0.08, 0.08, 0.09, 0.13, 0.14, 0.18, 0.22, 0.24, 0.25, 0.21, 0.18, 0.16, 0.13, 0.14, 0.14, 0.14, 0.14
"Seno", 0.12, 0.11, 0.11, 0.07, 0.08, 0.09, 0.07, 0.09, 0.07, 0.08, 0.13, 0.13, 0.14, 0.17, 0.20, 0.22, 0.23, 0.20, 0.18, 0.14, 0.13, 0.14, 0.14, 0.14
"Yatenga", 0.13, 0.13, 0.12, 0.12, 0.09, 0.10, 0.10, 0.14, 0.13, 0.14, 0.16, 0.18, 0.22, 0.23, 0.27, 0.30, 0.28, 0.23, 0.22, 0.17, 0.18, 0.18, 0.18, 0.17

Up to 36 values can be stored in one row. By default, if more than half of the pixels in a polygon have a value outside the valid limits, -9999 will replace the extracted value from the polygon.

2.7.4. Viewing graphs

WinDisp3.5 allows the graphical representation of different information, for example the results of the command Process Stats (described in section 2.7.3.) or information relating to a displayed image.

The command View Graph Image Series is used to display a graph of the evolution, in a series of images, of the values of a designated pixel. This pixel is selected with the cursor, clicking on it in the reference image previously displayed.

To activate the command View Graph Image Series open any image (it does not matter which image; the aim is simply to activate the function). To facilitate the selection of the pixels you want, it is recommended that you superimpose a map on the displayed image.

The parameter that you must supply next is the name of the file list containing the names of the images with the data you wish to graph. Open a window in which the graphs can be displayed by selecting the pixel on the map.

If you want to view two graphs at the same time for comparison, you must use a file list as seen on the right. This file list permits the display on the same graph, of the evolution of the values of the pixel selected, in the NDVI images for the period from April to November 1997, and in the NDVI averages (16 years) from April to November.

To display a graph in the new window, select the designated pixel on the image marking on the map that you have superimposed (Afrad1.bna). As you can observe in the illustration below the legend of the graph is set up from a file list.

In this example, the red line (A) represents the evolution (April to November 1997) of the value of NDVI data for the selected pixel, while the green line (B) represents the 16 year NDVI averages for the same pixel.

The command View Graph Map Data is used to display a graph showing the evolution, for a series of images, for values of a selected polygon; in other words, graphing the relative temporal data presented in a table in which the first column contains the names corresponding to the names of cartographic features of the map displayed. The data originates from the table of results from the analysis done with Process Stats (see Section 2.7.3).

To activate the command View Graph Map Data, open any image. Superimpose the map you want to work with using Retrieve Map. You must supply the name of the file list containing the name of the statistics file where you want to put the graph data. Open this in the window by selecting the polygon on the map.

Example of Burkina Faso: the file list used is this list of two statistics files obtained from the analyses done on the NDVI images of 1997 and the NDVI averages, 'ndvi97.sta' and 'ndviavg.sta' (see the beginning of the example in Section 2.7.3.).

To display a graph of data in the new window, select the province on the reference map of Burkina Faso that you have superimposed on the image. As seen in the illustration below, a selected polygon appears in red; its name appears just above the graph, and the legend is set up through the file list.

In this example the red line (1) represents the evolution (April to November 1997) of the average NDVI data for the province of Gourma, while the green line (2) represents the 16-year NDVI average for the same province.

The command View Map Histogram is a function to display a histogram representing the number of pixels of an image, or of a difference image, having a value equal to each of the 256 possible pixel values.

The following graphs are obtained by this command for the satellite image of Africa for CCD for the third dekad of July 1997, and for the difference image between the first image and the one for the third dekad of November 1997:

In these graphs one can note that the images contain about 750 000 pixels; in the first graph these are divided mostly between very dry areas (pixel value=0) and ocean (pixel value=250). In the difference image the pixel values are mostly in the center of the color values (128) indicating that for the majority of the pixels, the two images used to calculate the difference image are similar.

2.7.5. Process SEDI

The objective of this command is to run Satellite Enhanced Data Interpolation (SEDI) routines. The SEDI interpolation method was developed for the Regional Remote Sensing Project based at Harare (Zimbabwe). This routine interpolates rainfall data measured at ground level stations with CCD images received from the FAO Artemis Project. The method has also been applied to other parameters such as potential evapotranspiration (PET) and altitude, and agricultural yields and NDVI.

Despite certain deviations from the basic idea, the concepts of this interpolation method have been published in many different places. A comparable method, called "co-kriging", has been applied in many geological studies concerning groundwater. SEDI is an easily used method for "assisted" interpolation. It can be applied to any parameter for which values are available for a certain number of geographic stations, as long as there is a background which must have a negative or positive relation with the parameter to interpolate.

Three elements are required to insure complete success with the SEDI method:

The SEDI method produces, in the form of a field, the parameter to interpolate. The calculation can be influenced by setting a number of input parameters.

An example using rainfall and CCD

Rainfall data are gathered on a dekad (10-day) basis in many countries of the world. The geostationary METEOSAT satellite produces infrared temperature images of the earth every half-hour. In tropical regions it can be assumed that areas with temperatures lower than minus 40 degrees Celsius are covered with clouds. The cumulated number of hours in a dekad with this low temperature is called "Cold Cloud Duration" (CCD), and is represented as an image. Each pixel of the image represents one data value, and can be assigned a color depending on the value. The relation between rainfall and CCD is positive. In other words, high rainfall values generally coincide with high CCD values.

The SEDI process is done in three steps:

Step 1: Extracting values from the image and calculating the ratio

For every point in the rainfall data, a value can be extracted from the CCD image. The SEDI method will find the pixel that coincides with a rainfall station and extract the pixel value. In some cases the value of one pixel does not give satisfactory results. Therefore SEDI allows the user to extract values of more than one pixel from the image, and take its average as the image value for the station.

The rainfall values are stored in an ASCII text file with the following format:

For each station we now have a rainfall value and a CCD value. The Spearman rank correlation coefficient yields a positive value. This means the relation between rainfall and CCD is positive. All stations for which no values could be extracted (either because they lay outside the image window or the extracted values are missing) are eliminated from the output file. Therefore the output file may contain fewer lines than the input file. The ratio between rainfall and CCD value is now calculated and stored in an ASCII table with the following format:

Step 2: Creating a regularly spaced grid from the ratios


The second step is to create a grid from the irregularly spaced ratios. The ratio grid is created with the inverse distance method with a weighting power of 2.

The software allows the user to set:

Step 3: Creating the SEDI image


The third step is to create the SEDI image. The process is simple. By multiplying the grid obtained in Step 2 by the background image, an estimate of the interpolated value to interpolate is obtained. For the rainfall and CCD data, a rainfall image is obtained by multiplying the values of the ratio grid by the values of the CCD image.

Some remarks concerning the image created:

Automated processing

The command Process SEDI Automatic can be used to perform all three steps at once. The command Process SEDI Assisted also performs the three steps automatically, and it also estimates values for most of the parameters. The estimated values are saved in a file (assist.ini) and can be used as a starting point for more refined calculations.

This file is automatically displayed in WinDisp3.5 after processing has been completed.

The SEDI software

The SEDI methods are incorporated in a DOS software package called DOS IGT (IDA GIS Tools). This package is freeware and can be downloaded from the FAO FTP site: FTP://FTP.FAO.ORG/SDRN/IGT/

The IGT manual describes the SEDI process in much more detail and is available at the same ftp site.

2.7.6. Modification of an image

There are many manipulations that can be applied to change the content or the presentation of an image.

The following is a list of the available functions. For explanations concerning the use of these functions, refer to Chapter 3.

Process Images Algebra

Process Images Compress

Process Images Window

Process Images Filter

Process Images Difference

Process Images New

Process Images Paste

Process Images Map

Process Images Mosaic

Process Header

Process Header Changer Value

Reproject

2.7.7. Importing and exporting

With WinDisp3.5 you can import images, maps and color tables, and convert them into a format which can be used in WinDisp3.5, e.g., IDA format for images, BNA for maps, and ASCII files for color tables. You can also export images and maps in different formats.
The following is a list of the available functions. For detailed explanations of the use of these functions, refer to Chapter 3.

Process Import Table Ida

Process Import Ascii Image

Process Import Binary Image

Process Import Erdas Image

Process Import Erdas Table

Process Import Idrisi Image

Process Import Idrisi Vector

Process Import Surfer Grid

Process Import Surfer Plot

Process Export Ascii Image

Process Export Binary Image

Process Export Erdas Image

Process Export Idrisi Image

Process Export Idrisi Vector

Process Export Surfer Grid

Process Export Surfer Blank


Previous PageTop Of PageNext Page
This manual was prepared by the Global Information and Early Warning System of the United Nations Food and Agriculture Organization
© FAO/GIEWS - 1999
E-mail: Giews1@fao.org