Data Collection Process

Data collection entailed a couple of activities that are important in order to allow for accurate representation of the dataset. Moreover, the project also envisaged that not only vector gesopatial datasets are collected but also raster datasets.

  1. Identification of the targeted crops in each of the countries of interest.

  2. Identification of specific areas that the targeted crops are grown in each of the country.

  3. Identification season that each individual crop is grown (Long Rains & Short Rains).

  4. Stratification of the areas and scientific sampling procedures for accurate data capture

  1. Planning field visits in tandem with crop and season.

  1. Planning of Drone Image acquisition based on crop season and weather information and other flight Planning processes.

  2. Actual Data collection.

Targeted Crops Identification

The crops were targeted based on the impact envisaged for this program, this is because the program is supposed to offer crop analytics at scale. Therefore the rationale for identification of crops was majorly on crops that are widely grown by majorly small holder farmers in the countries of interest.

The program also sought to provide data for these crops that are widely grown with no existing publicly available datasets and thus bringing about predictability since, data would be available, for simulation of future seasons and yield prediction. Moreover, the program brings about an oportunity to understand spatial and temporal distribuition of crops, information on growing period and various challenges that farmers go through. This would later in turn inform product development at PULA, patners and authorities.

The crops were identified through various means;

Research was done on the major crops small holder farmers grow through reading of publicly available reports. It was also complemented with resources available at PULA where there is a dedicated call center that was used as a reference to get views directly from calls made to farmers. Enumerators were also used in finding out distribuition and crop types grown accross different areas.

Identification of Areas where Crops are Grown

Having the crop types identified, it is important to focus specific areas that are under-represented when it comes to availability of data, but with a good number of farmers.

The process entailed an elaborate desktop analysis prior zoning to the areas that were deemed suitable, this was done using remote sensing approach of crop suitability mapping. It is achieved by integration of a number of data products from various sensors mainly;

  • Rainfall.

  • Soil Moisture

  • Topography

  • Temprature

  • Soil

To efficiently manage the process, Agro-ecological zones were used as the geographical units in doing the analysis. Agro-ecologocal zones are units of land that exhibit similarilities or homogeneority in terms of soil and climate conditions. Thus, if crops were to be grown in an Agro-ecological zone they may have similarilities in growing patterns and length.

Vegetation indices were developed and integrated with the aforementioned datasets to form basis of crop suitability mapping, this entailed analysis such as land and water productivity.

Seasons that Crops are grown

Rainfall Data acquired together with other ancillary data was used to determine the period with which the crops were planted. On the other, remotely sensed data was used to study crop phenology and give an indication on when each crop was planted and have estimates on other activities that precedes planting.

Stratification and Sampling

After Identification of areas that the crops are grown, sampling technique was applied to come up with areas that would be visited. This was done by subdivision of the Agro-ecological zones into grids and generation random points that are equal for each grid, the grids would later be used to assign Enumerators the areas they need to cover for data collection exercise. The same was also used to determine which areas drones would be flown to aid remote sensing data acquisition.

Planning Field Visits

With the covid-19 situation field visits were somewhat a challenge, however field visits were planned based, on crop calender and protocals developed by PULA for various crops. The protocal;

  1. Box placement and farm boundary data collection(GPS location collected and images)

  2. In season visit for crop development reporting

  3. Wet harvest crop cut experiment

  4. Dry harvest crop cut experiment

Planning Drone Imagery acquisition

Drone imagery acquisition was greatly influenced by areas that crops are grown, type of farming practises; (inter and/or monocropping), the distribution of farmers geographically. Atmospheric conditions and permissions to fly were also part of the items that were consdered during planning and execution phase.

Actual Data Collection

This is a very crucial phase in the program since, it influeces the quality of data. How the process is undertaken affects the efficiency of the data collection activity(s). Therefore certain activities have to occur during this phase in order to get things right these includes;

  1. Define protocals for data collection for respective crops.

  2. Recruitment of Enumerators.

  3. Training of the Enumerators.

  4. Calling the farmrers to prepare them for the oncoming exercise.

  5. Calling the farmer to book appointment on when to carry out the exercise of Box placement.

  6. Box Placement

  7. Harvest

    1. Wet harvest

    2. Dry harvest

  8. Quality Check

Box placement is a process whereby, Enumerators set up a box using logs and ropes measuring 5 by 8 meters. Enumerators use a built a mobile application by PULA to collect the Geographical co-ordinates of the boxes and the farm, this information is then saved and awaits a number of processes, for data quality and further analysis. This can be done during growing phase of the plant or two to four weeks before harvest.

Harvesting was done inside the Placed box, whereby the Enumerator(s) and the farmer together havest and weigh the harvest, this is then placed into an separate guny bag that is then put to dry and weighed and recoreded. This information is Crucial as it will help to generate coefficient(s) of moisture for a crop. The information together with photographs is Keyed into the mobile application and submiited to the servers.