Analytical functions: growing season analyses

Tutorial (En) Advanced useAdd WMS layer

Topic 7: Display and examination of images

M 2. MAPPING THE START, END AND LENGTH OF THE GROWING SEASON USING THE NDVI

1. Learning objectives and requirementes

Objectives

  • Understanding of the agro-meteorological definitions of the start and the end  of the growing season used in literature.
  • Understanding  of the vegetation index (NDVI) based definitions of start and end of the growing season.
  • To be able to create images displaying the start, end and length of the growing season based on simple vegetation index criteria.

The requirements for this topic are:

  • basic knowledge of agro-meteorology and vegetation dynamics.
  • understanding of the vegetation index and its behavior in time (topic I4,W9).
  • to be able to display images and create difference images (topic W3)
  • understanding of the principles of batch processing in RAMANI Cloud Service (topic W6).
  • to know how to create color tables  in RAMANI Cloud Service (topic W8).

2. Introduction

2.1 Importance of the growing season assessment

Monitoring the length of the growing season is an important objective of agrometeorological measurements. The length of the growing season is  related  to crop production levels (Sivakumar, 1988, Kassam and Andrews,1975, Jones and Stockinger , 1972), and to perennial vegetation / rangeland production (Townshend & Justice, 1986) . An estimated start , end or length of the growing season is used in  food security models (FEWS, 1993, Coulibaly, 1995) and can be used in climatic models (Nijhuis and Groten, 1997).

Instead of measuring the phenology of vegetation directly, agrometeorologists always have based their assessment on measurement and modeling of rainfall, evaporation and soil water balances (Sivakumar, 1987,1988,1992, Diepen, van et al., 1992 ,etc). FAO (1978) defined the growing period as the period during a year, when precipitation exceeds half of the potential evapotranspiration, plus a period required to evapotranspire

an assumed 100 mm of water from excess moisture from previous decades (or less, if little or no residual moisture is  available).

Q 1 Do you know the meaning of evapotranspiration and difference between

potential and actual evapotranspiration (ETp / ETa)  ?

F If not, try to find some basic agrometeorological literature

Measuring rainfall has the advantage that long data time series are available, making probability assessments possible. If station data are available, modelling can be done through interpolation for any spatial resolution and for any time resolution using a dynamic modeling approach (e.g. Driessen,1992). It also avoids to take into account spatial heterogeneity of vegetation cover, so that modeling can be done for any real or hypothetical cropping system, and for all climatic conditions. Also the occurrence of excessive rainfall can best be measured with ground station data. Therefore agrometeorological crop water models remain  an important tool in land evaluation and agricultural research.

Q 2 Which  agrometeorological models have been used in your country ?

F Try to find applications in literature and check with the agro-

meteorological office and agronomic research institutions.

A rainfall-based agrometeorological assessment of the growing season however only represents a model of assumed vegetation cover and phenology. In absence of  information on real water availability to vegetation, monitoring is done on a basis of a  rainfall -evapotranspiration ratio (the moisture index),  often by using an assumed  water holding capacity  of the soil (AWC).  In some semi arid zones, however rainfall shows such a high spatial variability,  that rainfall values extracted from stations of a  distance of more than 10 km are not better than long term average rainfall values (Dugdale and Milford , 1986 for sahelian areas). The factor most difficult to estimate is  the run-off coefficient, which highly varies in space and time, especially on loamy soils of semi arid zones with a tendency of surface crusting (Karnieli,1994, Lamachère , 1990 ).

Q 3 To how many rainfall stations do you have easy access in your country?

Which is the average size of an area they have to represent ?

F Check topic W 4, paragr.5.1: your answers to the questions Q 9-11

Therefore it is useful to develop methods to monitor vegetation development directly by remote sensing, using a vegetation index. It can be predicted, that direct vegetation monitoring will replace some functions of classical rainfall based meteorology because of the following trends:

  • increasing length of time series (NOAA data are available since 1982)
  • increasing resolution of satellite systems
  • complete and large rainfall data sets of particular areas are rare
  • more powerful computer hardware becomes available to small user communities
  • software for processing image time series becomes more user-friendly (e.g. RAMANI Cloud Service)

2.2 Remote sensing based vegetation monitoring methods

A satelllite vegetation model can be based on METEOSAT rainfall or evapotranspiration estimates (Seguin et al., 1986, Rosema,1990). In a data scarce environment it has a  better spatial resolution than models based on interpolation of  meteorological station data (5 km). The accuracy of rainfall estimates however still seems to be more appropriate for climatic models than for vegetation and crop monitoring (Milford & Dugdale, 1989).  Rainfall estimates have however shown to be least accurate during the most crucial period of the year, namely during  the start of the growing season (Diallo & Turpeinen, 1988). Rainfall Estimate (RFE) images , which are satellite estimations  merged with ground observations may be more reliable. However the error of +/- 60 mm/decade[2] is too high to detect the start of the  the growing season, since we know, that a rainfall event of 15 to 25 mm during an appropriate period can already be enough for farmers to start sowing.

Sivakumar (1988,1992), using interpolated rainfall data, and Martini (1993), using METEOSAT rainfall estimates, developed maps of the length of the growing season. They apply it to calculate recommended sowing dates for  farmers , which is a useful contribution to agricultural research and  land use planning. This valuable application should however not be confused with  monitoring of the actual  crop and vegetation development as basis for food security early warning.

In the following paragraph we will map the estimated start, end and length of the growing season using using a relatively simple method based on  vegetation index data.

3. Defining the start and the end of the growing season

3.1 Phenological cycle

Different vegetation types and crops  have different phenological cycles, largly depending on the climate of the country, but also of terrain and soil types and human management. The figure on the next page  is an example of the phenological cycles of cereals, grasses and trees and shrubs in the Sudanian zone, which is a climatological zone with 650 to 1000 mm stretching  through the center and south of the Sahelian countries.

Q 4 Check for the main climatic zones of your own country :

What is the normal length of the growing season in your country/area ?

When is the first and the latest possible date for sowing rainfed crops ?

When is the first and last harvesting date for important rainfed crops ?

When do trees normally start to develop leaves ?

When is there a generalized development of natural vegetation ?

In which micro-areas is there a very early vegetation development ?

T 1 Improve upon the crop calendar available on the internet [3] (topic W 7)

Adapting it for different climatic zones and adding information similar to

the calendar above. The calendars will be essential for interpretation

of the monitoring data.

For those from Sahelian countries:

check this figure for the Sudanian zone, and prepare another

graph for the (northern and/or southern Sahelian climatic zone)

3.2 Agrometeorological definitions for the start and end of the growing

season

The table below summarizes the different criteria used in literature.

Most criteria are based on precipitation and evapotranspiration.

Q 5  Why is the ratio between precipitation and evapotranspiration often used

as indicator for the start/end  of the  growing season?

AUTHOR & COUNTRY

START

END

Troll, 1965

India

P>ETp

P<ETp

Robertson, 1976;

Virmani et al, 1982

India

P/ETp>0.33

P/ETp<0.33

Hagenzieker, 1985;

Burkina Faso

decade in which p,(P=>1/3ETp)>0.7 (or 0.6 for northern part of the country) and if in the succeeding decade p,(P=>1/3ETp)>0.5 (or 0.4 for north)

decade in which p,(P=1/3ETp)<0.2 and if in the succeding decade p,(P=1/3ETp)< 0.1

Sivakumar, 1990

Burkina Faso

Date after May 1, when P accumulated over 3 consecutive days is at least 20mm and when no dry spell within the next 30 days exceeds 7 days

Date after Sept. 1, following which no rain occured over a period of 20 days (approx. 2 dekads)

Coe and Stern, 1986

Nigeria

The first decade after April 1, with more than 20mm in a rain spell of 1 or two days.

First decade after Sept. 1, during which the water balance decreased to zero (derived from a simple water balance equation assuming water at field capacity of 100mm and evaporation rate of 5mm per day)

FAO, 1978;

Henricksen and Durkin ,1985

Ethiopia

Kassam et al, 1991

Kenya

First decade when P=> 0.5 ETp

decade after the end of the humid period when (P100/ETp)<0.5 plus time required to evapotranspire stored soil moisture.

P=Precipitation, ETp=Potential evapotranspiration and p=probability

Criteria defined by Troll (1965) were made for climatic zone definitions (P>ETp), while others use a lower/ earlier precipitation threshold for defining the start of the crop growing season based on observation of farmer’s sowing dates.

Q 6 Which specific agro-meteorological criteria are used for the growing

season in your country ?

F If you don’t know , apply criteria from general literature and try to obtain

more information from the meteorological office and agronomic research

institutions

3.3 NDVI definitions of the growing season

Several researchers developed methods for estimating the start and the end of the growing season using NDVI parameters. The underlying thought is that as soon as the vegetation starts growing, the NDVI will respond by positive NDVI increments between consecutive decade images.  The principle of all these methods is that they measure a minimum increment or decrement over a certain period of time  (delta NDVI or D NDVI)

NDVI

Max. NDVI

Max. Min.

NDVI NDVI

Dry season level

1                                       dekads                               36

Q 6 Where on this curve could you define the start and the end of the

season ?

Hendricksen and Durkin (1986) developed a method to estimate the start and the end of the growing season in Ethiopia. They defined the start as that moment in time when the NDVI increments between weekly NDVI images become distinctly and consistently positive and that these increments will lead to maximum NDVI increments (D NDVI) for a particular growing season. This is indicated in the graph above. They defined the end of the growing season as that point in time at which the most negative rate of change in NDVI occurs during the declining phase of the curve, provided that it forms part of a consistent trend of several negative values and does not return to a positive in less than three consecutive weeks. The disadvantage of this method is that the start can only be assessed after the maximum has been reached and can therefore not be used in near real time for early warning during the growing season.

Groten (1993) defined the start of the growing season in Burkina Faso as the decade with NDVI an increment of 1 digital value followed by another decade with an increment of_³_ 3 digital values. This equation was based on  P/0.5*ETp ratios measured from 21 rainfall stations.  Ouedraogo (1995) recalibrated the method of Groten (1993) per climatological zone in Burkina Faso. He defined the start of the growing season as the point in time with at least two consecutive positive NDVI increments, and in addition an earliest possible starting date [4]and a minimum increment [5]per climatic zone.

Ocatre (1997) adopted the following decision rules to define the end of the growing season:

· The NDVI end of the growing season should exclude NDVI decrements resulting from interruptions within the growing season. The first possible end was considered the average end of the humid period.

· There should be two consecutive NDVI decrements after the first possible end of the growing season.[6]

Influence of land cover and land use type

The start and end of the growing season based on agrometeorological criteria is usually well defined for crops. The NDVI will measure crop development for those pixels which are dominated by agricultural fields. The main crop areas are in most cases associated with settlements or can be derived from a land use (intensity) map[7]. The decision rules implemented on the basis of agro-meteorological criteria are therefore an indication of sowing dates.

In areas largely dominated by grass fallows or annual natural vegetation the NDVI will measure the cover of  grasses. The NDVI curve and reaction of annual grasses to rainfall is similar to the one of crops, although crops sown late can show a delayed start of the growing season. In an area dominated by young fallows the NDVI curve is usually indicative of fallow and crop development.

Forested areas or areas dominated by dense bush or woodland can have quite a different phenological behaviour: an early greening, even before the first rains come, and a very gradual decrease of the NDVI after the rainy season.

In the following we assume, that the main crop areas are known , and that areas dominated by woody vegetation (like forest reserves) are excluded from the interpretation and calculation.

4. Mapping the start of the growing season in RAMANI Cloud Service

4.1 Decision rules

We will use  rather simple criteria to define the start of the growing season in this course. It is up to users with a specific interest to develop more accurate county specific methods and calibrations. The basic idea presented here is however generally valid. We will apply the following criteria  to calculate the start of the growing season:

· The growing season can only start after the first possible date, based on country specific climatic characteristics.

· There should be two consecutive positive NDVI increments.

There will be no bioclimatic stratification used. A stratified approach can be realized in a more general GIS (e.g. Idrisi, ILWIS).

We will use examples from Burkina Faso and Zimbabwe in this chapter. The first possible start of the growing season in Burkina Faso is considered the 12th decade  (last decade of April) and in Zimbabwe the 28th decade (first decade of October). These are based on the crop calendars presented in Topic W 9 and expert knowledge. The calculation is stopped 12 dekads after the first possible start , assuming that the growing season must have started by then.

.

Q 8 What do you think about the date of the earliest possible real start ?

When  do farmers start to  sow in your country ? Optionally: are the

dekads chosen for Burkina Faso and Zimbabwe realistic ?

4.2 Implementing the decision rules with  batch processing in RAMANI Cloud Service

The decision rules presented above can be implemented using batch processing. The principles of batch processing were described in topic W 6. We will first describe the implementation of the decision rules in general terms and then we step-by-step create a batch file which calculates the start of the growing season.

· We decided  on the condition that there should be two consecutive NDVI increments. This can be realized by using the the @IF(A,B,C) function. In Process/Images/Algebra function. This function has three arguments: A,B and C (condition, then,else)

Argument A is the condition. In our case the condition is (img2 >img1)

(img3 >img2), where img1 to 3 are consecutive decadal NDVI images.

Argument B is the value which is assigned to the pixel of the output

image when the condition is met and

Argument C is the value which is assigned to the pixel when the

condition is not met.

· By applying the @IF function to each of the 12 decade images from the first possible decade onwards it is calculated for each pixel whether the start could have taken place or not. If the start has taken place according to the definition in a particular decade, the  value corresponding to the number of the decade (decade nr.) is assigned to the pixel, if not (else), a value of 255 is given.

· Now we have 12 images which indicate for a pixel whether the start could have taken place in that decade or not. Now will use the Process/Series/Min function with the 12 images as input in a list file to create an image which displays the start of the growing season as the lowest decade number.

· a color table is produced or used to display the start of the growing season image (Topic W8). Keep in mind that RAMANI Cloud Service uses digital counts (DNs) to assign ranges of image values to colours.

Q 9   Why is the Process/Series/Min function used to calculate the final start

of the growing season image?

Q10 What does the condition ((img2 > img1) & (img3 >img2)) imply?

We illustrate this procedure using an example. The normal start of the growing season in Zimbabwe will be calculated using the historical  reference images (averages ‘82-’95) on the tutorial CD ROM:

ΠWe select the decade images relevant for the start of the growing season and copy them to a work directory where we can rename them in a sequential order. When the growing season starts within the calendar year this step is not necessary (e.g. Burkina Faso). In the Zimbabwe case the files \datazb\ndvi\avg27.img to \datazb\ndvi\avg4.img were copied to another directory. Then the value of 36 is added to the decade number of the images of the beginning of the year, so that we have a series renamed as gs27.img to gs40.img.

 We then start to record the batch file[8] which  calculates the start of the growing season images. This can be done with  the Batch/Record function .

Ž We will create  12 images (see above) by using a For/Next loop. So in the Batch menu we will now select Batch/Begin.

A variable name (e.g. decade) needs to be specified and we will make a loop from 27 to 38 in steps of 1. The images 39 and 40 are  used  , when the last image 38 is entered and compared to the image with dekad+1 and +2.

 To implement the if-condition,  we will need to compare three consecutive dekadal images . Specifying  variable names for these dekads can be done by Batch/Variable Set. Three variables have to be specified:

-Variable A which is given the value %decade%

-Variable B which is given the value %dekad%+1

-Variable C which is given the value %dekad%+2

 No we will stop recording the batch file by selecting the Batch/Stop function and  further edit it manually. Now open the batch file using Batch/Edit. When the steps 1 to 4 are completed  the batch file should look like:

Batch For Begin, "decade, 27, 38, 1"

Batch Variable Set, "A, %decade%"

Batch Variable Set, "B, %dekad%+1"

Batch Variable Set, "C, %dekad%+2"

Batch For End, ""

Insert the following line before the Batch For End line (fill in the appropriate pathnames):

Process Images Algebra, "@IF((img2>img1)&(img3>img2),%B%,255) ,C:\….\stnor%B%.IMG,img2,C:\…\GS%B%.IMG,img1,C:\….\GS%A%.IMG,img3,C:\…\GS%C%.IMG"

This line checks IF the condition described above is true and THEN assigns a value of %B% to the output image and ELSE a value of 255. The first parameter for this function is the name of the output image (c:\…\stnor%B%.img  for  the normal start decade image , and  the other three parameters are the names of the variables for the input files (img2, img1 and img3).  Then we have to save the batch file and play it (Batch/Play).

‘ Now we can create the final start of the growing season image by using the Process/Series/Min function. As you may recall from topic W7, the Process/Series function requires a name of the output image  and a list file with filenames in the first column. In this case the names of the 12 files generated by the batch file should be included in the file list.

F specify a logic name and directory for the output image

’ We then  create a color table (see also topic W8) for the image. Remember that  the values are area specific, and that for the Zimbabwe case a value of 28 means that the growing season has started in the first decade of October and 39 in the third decade of January (because of renumbering the january images).  A value of 255 means that  no start of the growing season has been measured.  Create a color table which assigns blue and green colors to an early start, yellowish colors for a +/- normal starting date and  red colors for a late start of the growing season (risk of crop failure).

The final result of this example is displayed below (wrong color table)

Q11 If you are familiar with Zimbabwe  you will know that the growing season in average starts in end October/ beginning of November[9]. This is visible in the largest part of  the image. What could be reasons for an area to show a very early normal start (beginning of October) ?  What could be  a reason for a late normal start (December ?)

T 1 Create an image which displays the normal start of the growing season.

If you choose to work on your own country, carefully define decision

rules before: Earliest possible starting decade,  based on your agricultural/

ecological knowledge. For Burkina  Faso apply the following instructions:

F Assume that the start will take place between the 2nd decade of April and the 3rd decade of July, so for the calculation you will need images from the 1st decade of April until the 1st decade of August (\databf\ndvi\avg10.img to \databf\ndvi\avg22.img)

F Use the steps described above for the Zimbabwe example, but with the following modifications:

F Step1: Skip step 1

F Step 3: make a loop from 10 to 20 in steps of 1.

F Step 7: Create a color table which is appropriate for the final output image. A value of 11 means the growing season has started in the 2nd decade of April and a value of 21 means the growing season has started the third decade of July.

F Display the image with color table created .

Q 12 Can you describe the difference between the start of the growing

season in the north of the country and in the south of the country?

5. Mapping an early or late start  of the growing season

(Differences from normal)

A useful application for food security monitoring is the creation of a map which displays the difference from a normal start. So far we have calculated the start of the growing season based on historical average images. In a similar way you can create images which display the start of the growing season for a growing season of interest. You should exclude clouded pixel values[10] or use cloud corrected images[11] in this case. Cloud correction methods are covered in topic W 11.

With the images algebra function you can subtract the two images[12]. The difference image created now displays the delay or the advance in dekads of the start of the growing season.

Q13 Why should you use cloud corrected images for  estimate the start of

the growing season of a current year ?

T 2 Create an image which displays the difference from the normal start of

the growing season in the following way for a country of your choice:

F cope with cloud effects by choosing method a), b) or c)

a) use available cloud corrected  images for Burkina Faso from on the tutorial CDROM (\decloud\cvi9610.img to cvi9622.img)

b) exclude clouds in computation by setting valid values to 2-255 in Process theshold (when using FEWS NDVI data)

c) apply a cloud correction (see topic W11) to the images series (for Burkina Faso: \databf\ndvi\vi9610.img to \databf\nd vi\vi9622.img and save the images to a logic directory and with a logic name.

F Create an image which display the start of the growing season in 1996

by the  same procedure as  in task 1.  For in Burkina Faso use data

from the tutorial CD ROM,. Save the output images  to a logical directory.

F Create the difference image with the Process/Images/Algebra function

typing in the difference equation:

127+ (A-B)[13]

The image created in task 1 is image B and the image created in the second step of this task is image A (save the output image with a logical name in a logic directory). You may also use another constant as 127 [14]. It only assures, that no negative values occur in the output image. In this case, don’t forget to adapt the color table below:

F Create a color table with the color table editor, like the one displayed below and display the created image using this color table:

Q14 Why is a value of 123 equal to a delay of the start of the growing

season of 5 dekads In this image ?

Q 15 Check the colour table: The basic idea is that a delay should beindicated by red colors and an early start in green colors. A normal start has been given neutral (grey or white) colors. Could you improve upon the color table that the degree of earlyness or delay is indicated more clearly ?

The image computed for Burkina Faso should look like:

Q16 What can you conclude about the start of the growing season in 1996?

Q17 Could you think about a  legend , which indicates maybe more clearly

whether  there is a delay or an early start of the growing season ?

6. Mapping the end of the growing season in RAMANI Cloud Service

6.1 Decision rules

The end of the growing season can be estimated using a similar method as the calculation of the start of the growing season. We will use the same decision rules as Ocatre (1997), which can be easily implemented.

· The NDVI end of the growing season should exclude NDVI decrements resulting from interruptions within the growing season. A first possible end of the growing season for Burkina is decade 25 (1st decade of September) and for Zimbabwe decade 5 (2nd decade of February).

· There should be two consecutive NDVI decrements after the first possible end of the growing season

Q18 Do you agree with these decision rules? If not, provide better ones.

6.2 Implementing the decision rules using batch processing in

Windups

The same procedure as discussed in 4.2 can be applied to the end of the growing season using some slight modifications. We will illustrate this procedure using an example; the calculation of the normal end of the growing season in Zimbabwe.

In short the sequence of steps is the following (check for details the procedure for the start of the growing season, and the example batch file in W6 batches).

(1) We will need to select the images where the growing season might end and copy them to a work directory where we can rename them sequentially, if required (only if the year boundary is crossed). In the Zimbabwe case we

Use the files of decade 4 to 15[15] (so that we can calculate a possible end in decade 4 to 13). However since we renamed the January images for the start of the growing season we will need to do the same for the end of the growing season: We rename all images to decade number + 36. Now the first decade 4 is 40 and decade 15 is 51.

(2) Then we prepare a batch file to calculate the end of the growing season image for the decades 40 to 49 in steps of 1. The images 50 and 51 are used, when the last image 49 is entered into the loop and compared to the image with dekad+1 and +2.

Batch for Begin, "decade, 40, 49, 1"

Batch Variable Set, "A, %decade%"

Batch Variable Set, "B, %dekad%+1"

Batch Variable Set, "C, %dekad%+2"

Process Images Algebra, "@IF((img2<img1)&(img3<img2),%B%,255) ,C:\….\endnor%B%.IMG,img2,C:\…\GS%B%.IMG,img1,C:\….\GS%A%.IMG,img3,C:\…\GS%C%.IMG"Batch For End, ""

Don’t forget to enter the complete path name.

The batch program checks for the dekads 40 to 49 in steps of 1, whether the condition for the end of the growing season is fulfilled:

IF the condition (two consecutive NDVI decrements) is true, THEN

the value of %B% (for the end decade) is assigned to the output image. Else (when the condition is false) the value of 255 is given. The first parameter for this function is the name of the output image (c:\…\endnor%B%.img) for the decade of the end of the growing season and the other three parameters are the names of the input files (img2, img1 and img3).

After the batch file is created and played (Batch/Play), 10 output images are generated c:\…\endnor41.img todnor50.img, because of the loop we created. (The end is considered to be the second decrease image (variable B).

(3)  Now we can create the final start of the growing season image by using the Process/Series/Min function. The list file with filenames in the first column should contain the names of the 10 files generated by the batch file.

(4) An appropriate color table is created, giving red colors to an early end

and green-blue for a late end. The colors should be in the opposite

sequence as those for the start of the growing season!!!

The result of this example is displayed below

Q 19 What is the earliest and latest normal end of the growing season in

Zimbabwe ? What could explain the differences ?

T 3 Create an image displaying the normal end of the growing season in Burkina Faso. Use the same procedure as we did for Zimbabwe, but apply the following modifications:

Step 1: There is no need to rename the images in this case since both the start and the end are in the middle of the growing season. So use the images \databf\ndvi\avg24.img to \databf\ndvi\avg34.img in the analysis.

Step 3: Make a loop from image 24 to 32.

Step 7: Create a color table (with the color table editor) which pertains to Burkina Faso. A value of 25 means the growing season has ended in the 1st decade of September and a value of 33 means the growing season has ended the last decade of November.

Q 20 In which decade does the growing season approximately end in the

north, middle and south of Burkina Faso respectively? Can you

explain the reasons ?

7. Mapping the length of the growing season in RAMANI Cloud Service

Finally we can determine the length of the growing season by subtracting the start from the end of the growing season. The length of the growing season can be an important parameter in for example crop models. The length of the growing season image in decades can now be calculated very easily using the Process/Images/Algebra function. The length of the growing season can be also be displayed in days by setting the image type in Process header to user defined and the slope to 10  (image values are multiplied by 10 for 10 days per decade). You will of course also have to create a color table for the length of the growing season image. The image below displays the normal length of the growing season in Zimbabwe.

This normal image does not show  many differences .

T 4 Create a color table and legend which displays more differences

In the range of 90 to 180 days.

Q 21 Are you familiar with crop varieties ? If yes, compare the normal

length of the growing season with the length of crop cycles.

Also compare the values to the crop calendar. If you don’t agree

With the results of computation, improve upon decision rules used.

T 5 If you want to better understand the significance and reliability of the

product, compare the result with images downloaded from the ADDS

internet site

· go to the ADDS server , data themes – digital map data

· download *.gif files of relevant thematic maps : Land use intensity (see next page),  agro-climatic zones etc. You can do this after

viewing them in Netscape by saving them using the right mouse

button. You can also import these *.gif files into a Word document

using insert picture.

T 6 Create an image displaying the normal length of the growing season of

your country. For  Burkina Faso use the images created in T 1 (start)

and T 3 (end). Verify the results with crop calendars, climatic data etc.