top of page
  • Stellamaris Nakacwa - Everywhere She Maps Program Director

YouthMappers Crop Identification Pilot Project- Tanzania

YouthMappers is collaborating with the U.S. Department of Agriculture (USDA) Foreign Agriculture Service, and the U.S. Agency for International Development (USAID) to provide YouthMappers’ chapters in Tanzania with the opportunity to receive technical training and experience implementing an agricultural field data collection survey. The YouthMappers network in Tanzania is spread across 12 public universities with over 90 students overall.

The dynamic advancement of technology and availability of remote sensing data demand field validation of crop data types to improve confidence in crop forecasting models and agricultural assessments. However, given the diversity in local conditions and crops across a country, region, or continent, the process of obtaining sufficient ground truth spatial reference data, remains a challenge. Working with YouthMappers to implement a crowd-sourcing approach has the potential to supplement the much needed data for improving the agricultural assessments and provide foundational information for many countries to apply modern agricultural techniques that help farmers and the entire agricultural industry increase yields and production, adapt to the effects of climate change, and fully participate in the modern globalized agricultural economy.

Target crops

Within Tanzania, this pilot project focuses on providing crop identification data that supports the estimations and market analysis conducted by the USDA/FAS’ International Production Assessment Division (IPAD). The target crops are maize/corn, rice, cotton, peanuts, and sorghum.


6 out of the 12 chapters are directly undertaking the processes and exploring open source methods to collect a minimum of 300 data points per geographical field zone location. Students will capitalise on their open mapping skills, and a free mobile data collection app to create a scalable, standardized approach to data collection. Working together with OpenMap Tanzania, they are undertaking activities including; planning, training on data quality enabling factors including the size of farms, and community engagement in three regions: Mwanza, Dodoma and Arusha.

Crop Coverage in 3 areas of Arusha, Dodoma & Mwanza.

The sample areas for the project are informed by the geospatial data requirements for commodity crop monitoring and reporting conducted by USDA/FAS’ International Production Assessment Division (IPAD). Important data features include grains (Corn, Rice, Sorghum, Wheat, Barley, Millet, Oats, Rye), oilseeds (Groundnuts/peanuts, Sunflower, Soybeans, Rapeseed, Copra, Flaxseed), and Cotton. In Dodoma, 50% of the farms are involved in oilseed production according to World Bank’s senior economic analyst Douglas Zhihua Zeng. Grains such as rice and corn are predominant in Mwanza and Arusha respectively. Cotton is found predominantly in the Mwanza area.

Another significant factor for selecting areas of interest for agricultural field data collection includes the crop growth cycle and stage of development for most of the target crops. It is more difficult to accurately identify younger/recently planted crops than mature/fully developed crops. It is also incredibly challenging to identify crops after they have been harvested, as sometimes the plant is cut back to the bare soil. Typically, crops develop during a rainy period.

Growth Cycle per Crop.

Another factor influencing the distance covered for collecting ground-truth data is the expected quantity and type of remotely sensed data that would then get analyzed with the ground-truth data. The areas of interest (Arusha, Mwanza, and Dodoma) for the project are informed by USDA/FAS’ International Production Assessment Division (IPAD). This agency leverages freely available satellite imagery for its machine learning efforts. From the freely and globally available imagery of MODIS, Landsat, and Sentinel-2, it is the Sentinel-2 imagery at 10-20 meters spatial resolution that is most appropriate for Tanzania due to the relatively small field sizes commonly found in the country. One objective of this project is to conduct crop type mapping with multiple scenes of Sentinel-2 imagery of Tanzania, or at least the northwestern part of Tanzania. Sentinel-2 imagery supports the area of interest for the machine learning analysis and thus informs the quantity and distribution of the ground-truth data sampling.

Other machine learning projects with remotely sensed data might make use of aerial or drone imagery, covering a smaller geographic area. While the extent of the imagery might have a smaller geographic footprint, the data volume is higher due to the higher spatial resolution of the data and therefore requires more ground-truth data sampling in a smaller geographic extent.

While using KoboCollect, students will collect at a minimum, 100 data points per target crop (maize, rice, cotton, sorghum, and peanuts). Collectively across the three regions, 300 data points per target crop should be collected (with maize at 400 data points, as it is the most broadly available). Additionally, 400 data points will be collected for other land cover types to help distinguish confusing land cover from the target crops, across the three regions. Together, this is at least 2,000 crop/non-crop identification data points (or field outlines). Collecting additional data is incredibly helpful, providing it is spatially distributed.

Data quality is controlled by following the specific goals per district. These data will need to be spatially distributed to better inform machine learning models interpreting in-season satellite imagery. Fields that are at least 30 m x 30 m will be surveyed at regular intervals along the roads. The following metrics provide a guide regarding the expectations for finding the target crops and support logistical planning to organize the fieldwork.




Identifying at least 300 target crop fields/plots. Arusha-Moshi is a bimodal area, with mostly maize and rice to find, plus some sorghum and millet.

  • 100 maize fields

  • 100 rice fields

  • 20-50 sorghum

  • 30% of the data collection will cover: Approximately 50 fields/plots (whether they are vegetables, orchards, other crops, wetlands, large stands of trees, or other unique or identifiable features in the landscape. )

Identifying at least 1,000 fields/plots.

Mwanza is a bimodal crop growing area.

  • 200 Maize fields

  • 200 Cotton fields

  • 200 Rice fields

  • 50 Peanuts or groundnut fields

  • 50 Sorghum fields

  • 30 to 40% of the data collection should also cover: (whether they are vegetables, orchards, other crops, wetlands, large stands of trees, other unique or identifiable features in the landscape.)


Only about 1,429 (4%) hectares are under irrigation

Identifying at least 800 fields/plots.

Dodoma is a unimodal area, where harvest could begin in May.

  • 250 Sorghum fields

  • 100 Maize fields

  • 100 Millet fields (not an original target, but very prevalent here)

  • 50 Sunflower fields (not an original target, but a valuable crop of interest)

  • 50-100 Peanuts or groundnut fields

  • 20-50 Cotton fields

  • 20 or so Rice fields

  • 30% of the data collection should also cover: (whether they are vegetables, orchards, other crops, wetlands, large stands of trees, or other unique or identifiable features in the landscape.)


Sorghum is likely to be found south of Dodoma city and more maize is likely to be found east of Dodoma city. Most of the maize is north and south of Mpwapwa City.

In the first week of the project implementation, 14 students were trained in a five-day workshop designed to review the technical literacy on agricultural crop mapping, field pilot design and also accumulate knowledge on their potential contribution to the agricultural industry through professional engagement. The trained participants and supervisors will return to their respective universities and train more participants.

Following our social media handles @youthmappers for more project engagement and updates.


bottom of page