G-tech advancing agriculture in Sudan

G-tech advancing agriculture in Sudan


Agriculture sector in Sudan gains more importance not only because it is essential for feeding the people of the country, but also because more than 70% of total oil fields of the pre-divided Sudan, became a part of South Sudan, approximately 2 years ago.

The social and agro economics of the country is directly related to the success of the agriculture seasons in both irrigated and rain-fed sectors. The statistical forecasting for crops acreage and production are vital information for the economic sector’s decision makers as it is correlated to most trade policies. Moreover, the productivity geo-information is used for evaluation and enhancing of land production. Agricultural tracts of Sudan, falling in the Sahel region of Africa, face extensive threat due to their dependence on rain water. The amount of rain and it’s progression in different parts of Sudan varies from year to year and hence the resultant crop acreage and productivity. It has, therefore, become very important to know before-hand the acreage and production of important staple food grains like sorghum, millet and cash crops like cotton, groundnut and sesame.

Satellite remote-sensing based classification is the statistical technique to group population of pixels in broad classes as per their inherent characteristics. The different crops being grown during agricultural season can be estimated using different sets of multi-temporal satellite images, aided with high level of image interpretation and processing techniques. The crop yield values calculated using satellite images, along with the field data on crop yield, are used to calculate the crop production.

Agricultural Bank of Sudan and Ministry of Agriculture in Sudan have been funding the monitoring of summer crops in the middle & eastern parts of Sudan using remote-sensing and GIS for the past two years to ensure that they are pre-informed about the expected crop acreage and production and take necessary decisions.

With the availability of a constellation of 5 satellites of RapidEye having resolution of 5m, it became possible to capture nearly cloud free data even during the rainy season and also to map individual crops at field level. Statistics, generated at field level were further compiled to create separate statistics for the irrigated and rain-fed areas in the different states, thus giving a clearer picture of expected crop acreage and production in the study area.

The study area covered approximately 210,000 sq km representing both irrigated and rain-fed sectors in middle and eastern parts of Sudan. The large irrigated schemes of Al Gazira, Al Rahad, New Halfa, and Suki, as well as the rain-fed schemes in parts of Gedaref, Sennar, Kasala, Gazira, White Nile and Blue Nile states, formed a part of the target area.

Figure Showing the Study Area in Yellow

Figure Showing the Study Area in Yellow

Large Irrigated Schemes within the Study Area

Large Irrigated Schemes within the Study Area

Following datasets were used for the preparation of different spatial maps required for this project:


Analysis was done for the requirement of software and hardware to import, process, create and analyze the different datasets required for agricultural mapping, acreage estimation, yield estimation, and production projection.

Three rounds of field surveys were conducted. To make the field survey relevant and also to understand the nature of crops, it was planned to carry out the first field survey primarily in the areas, which could not be visited during last year’s study. The objective was also to visit different parts of the scheme to get first hand idea about the crops being grown in the study area during the current summer cropping season.

Understanding of Cropping Pattern in the Study Area:

The visit to different parts of the irrigated schemes, and also the visit to the mechanized and rain-fed cultivation regions of the project area, helped in knowing the different crops which were sown during the summer season.

The following table shows the general timing for the growth of different crops in the study area

Table showing general life span of different crops in the Study Area

Table showing general life span of different crops in the Study Area

Dotted lines with different colours in each row show the period of early and late cropping for most of the crops, except for Millet and Sesame. In the study area, the time of sowing of Millet and Sesame was almost fixed for the entire study area.

Methodology for Crop Acreage Estimation

Detailed methodology used for crop acreage estimation has been given below-

Spectral Classification

Satellite remote-sensing based spectral classification is the statistical technique to group population of pixels in broad classes as per their inherent characteristics.

Time Series Based Classification

In the areas where a difference exists in the crop cycle, multi-temporal satellite images are used to map the crop based on ‘time series classification’. In time series classification, images are acquired keeping in mind the transplantation and harvesting dates of different crops.

Figure showing variation of spectral signatures with time, used for image identification

Figure showing variation of spectral signatures with time, used for image identification

NDVI Slicing Technique

In the areas where no spectral signature is available for the classification of crops, NDVI slicing technique is found to be very useful for differentiating between different crops. In this technique, NDVI is run for satellite images. The NDVI values are then processed to remove all the negative values to retain data for the vegetation only. Further, based on the general crop vigour ranges and the acreage of crop as reported by different government agencies for the respective years, slicing of NDVI value is done to map the different crops.

In this project, hybrid classification, time series method and NDVI slicing techniques were primarily adopted to map different crops. The decision about the appropriate method to be adopted was based on the crop type and also the existing cropping pattern in the study area.
Based on different methods, the crop fields were classified primarily into summer crops and non-crop areas/vegetation.

The following figure shows the general methodology adopted for classification of different crops in the study area

Figure showing flow chart of the Crop Acreage Estimation Process

Figure showing flow chart of the Crop Acreage Estimation Process

Once the crop distribution map was created, the crop acreage was estimated based on the following formula-

Crop Acreage = No. of Pixels under Each Crop x Area of One Pixel

Methodology for Yield and Production Estimation

Remote-sensing has the potential of not only identifying the crop classes but also to estimate the acreage and predict the produce based on the yield information collected either through CCE (crop cutting experiments) data collected from the field or modelled thorough various statistical techniques correlating that with the NDVI values at crop heading stage.

Crop yield varies from region to region depending on several factors such as, crop variety, geographical location and soil type, temperature, timely availability of irrigated water, local agricultural practices, crop transplanting, harvesting dates etc. Therefore, crop variety-wise average yield values are collected from the field or other sources.

The following paragraphs show the different parameters and sources used for final estimation of crop yield and to project crop production.

Statistical Data on Crop Yield

Historical statistical data on crop yield and projected production values for different crops of the study area were collected from the Ministry of Agriculture’s Statistical Department.

Outputs and Deliverable

The following outputs and deliverables were obtained:

•    Map showing spatial distribution of crop acreage in the study area
•    Map showing spatial distribution of crop performance in the study area
•    Map showing water stressed areas
•    Data showing acreage, yield estimation and projected production at different levels (Study Area – State – Schemes – Farm).
•    Data showing results of analysis of correlation between crop performance and other controlling (affecting) factors including rainfall rate and rainfall distribution, for further enhancement of production
•    Data showing geospatial analysis of seasonal ‘Crop Rotation’ adopted in the study area. The resultant data provided information on how far the farmers adopted the ‘Standard Crop Rotation’ in their cultivation process
•    Data showing correlation of the crop performance and production with farmer’s GIS database related to loan and insurance system

Results and Discussions

During the third field visit, more than 200 distributed points were visited and investigated, for crops classification, while 100 farms were investigated for the projected yield categories based on NDVI, historical statistical data on crop yield and estimates from farmers and field experts.
The result showed that:

–    91% accuracy for crops classification has been achieved
–    87% accuracy for yield estimation has been achieved

Out of the various crop classifications validated in the field, remarkable achievement was made in the classification of the main staple food of the country. Out of the total 92 points validated in the field, which were declared as Sorghum, based on the laboratory classification, 90 points were found to be correct. This equals to an accuracy of 97.8% in the classification of Sorghum.
The lowest accuracy was achieved in the case of Millet, in the White Nile State, 4 out of 12 fields declared as Millet in the laboratory, were found to be dry grass during the field validation. However, this misclassification was limited to a particular region of White Nile state only.

The study was found to be very useful in providing data for the policy and decision making process in the country. The final report and data were used as a reference guide for government’s ‘economic sector’ to develop trade policies and fix crops prices, while the ‘agriculture sector’ used it for evaluation of the season and also for formulation of guidelines for enhancing the production in the coming seasons.

The Agricultural Bank of Sudan, which provided loans and funds to farmers, used the results of this study and the related GIS database to follow the loan recovery process. They also used the data for evaluation of production in relation to the existing insurance system.