Statistics from space: a new approach to agricultural production data for Mozambique

By Rowena Valmonte-Santos

Crop production statistics are essential for any government, but especially so in Mozambique where agriculture accounts for a quarter of economic production and the livelihoods of 80 percent of the population. The Statistics from Space project, funded by the Government of the Republic of Korea, aims to develop a highly accurate and efficient system to estimate and forecast production of different crops across the country by 2025.

The project team worked closely with stakeholders in the Mozambique government to identify their pain points and what would be needed from a new system. They identified challenges such as the diverse cropping systems in Mozambique, the high cost of its current system and timeliness, reliability, and usability. Since the project aligns closely with CGIAR Initiative on Digital Innovation, which focuses on harnessing digital technologies for better decision-making in food, land, and water systems, the Initiative team actively participates in the project to promote synergies in supporting the crop production information system.

The solution

For satellite remote sensing-based data analysis to work, it is still necessary to go to the field to capture quality groundtruth data to develop the model and validate the result. The innovation in Statistics from Space is to use time-series satellite images to identify where, when, and how many samples need to be taken to get an accurate picture of yields across the diverse and dynamic farming systems around the country.

Unlike other methods which choose sampling sites based on where the population lives, Statistics from Space identifies different kinds of crop production areas identified in satellite images. The images show how greenness on the land varies during an average year (over the last 20 years), and this is combined with land-use and soil data to identify strata of land with similar characteristics.

The next step is to identify sampling locations across each of the land strata so that each is well covered.

Combining layers of data to identify land strata and sampling sites. Source: Statistics from Space with maps courtesy of CENACARTA

As a pilot, during the main cropping season in 2023-2024, field agents visited each of the 92 identified sampling sites in Gaza province and recorded the area of different crops being grown across the 500 x 500 square meter section. Some sites were also mapped using drones to test the utility of fine resolution data.

In the sampling phase, the study team trained technical staff from national institutes in Mozambique to participate in the sampling, collaborated with local extension systems and consulted with location communities. Local universities provided supervisors and field agents, while the University of Twente developed a customized data collection mobile app for them to use.

Field sampling using the smartphone app

Making some assumptions about the varieties and practices used to grow crops, the team used crop modelling to estimate the yield of crops based on weather, soil, and satellite-estimated leaf area data. The simulations were also run backwards in time to compare with historical records of crop yields, showing that the uncalibrated modelling technique had an error rate between 20-30%. The prediction accuracy is expected to improve by including province-level data on fertilizer use, irrigation, crop varieties and farm size, as well as on-farm yield data to calibrate the model.

Engaging stakeholders

Statistics from Space started as a research project in 2022, funded by the Government of the Republic of Korea (Ministry of Agriculture, Food and Rural Affairs; MAFRA) with in-kind contributions from CGIAR Initiative on Digital Innovation. It brings together partners such as Research Centre for Agricultural Food Policy and Programs (CEPPAG) at Eduardo Mondlane University (UEM) in collaboration with the International Food Policy Research Institute (IFPRI) and in partnership with the Faculty of Geo-Information Science and Earth Observation (ITC) of the University of Twente, Netherlands and the Seoul National University (SNU), South Korea.

A workshop was held among all these stakeholders on June 18, 2024. It provided opportunities to solicit feedback, strengthen institutional collaboration, and identify different avenues of working together on innovation development not only in agriculture but in other economic sectors as well. The project partners and key stakeholders also agreed that the remaining period of the project will be a transition to knowledge sharing, capacity strengthening, and transfer of the technologies developed to the local institutions. These next steps are aligned with discussions to invest in agriculture, infrastructure, communications, and technology that took place during the visit of the Mozambique Government delegation, led by President Felipe Nyusi, in South Korea on June 3, 2024.

Discussions during the Stakeholder Workshop and Engagement, VIP Hotel, Maputo, Mozambique, June 18, 2024.

The Stakeholder Workshop and Engagement brought together 35 participants representing the Directorate of Planning and Policies, MADER; Mozambique Central Bank; Mozambique National Cartography and Remote Sensing Center (CENACARTA); National Directorate of Internal Trade, Ministry of Industry and Commerce; National Directorate of Economic and Development Policies, Ministry of Economy and Finance; National Geospatial Development Agency (ADE); National Institute of Statistics (INE); Faculty of Agronomy, UEM; Faculty of Veterinary, UEM; School of Business, UEM; Scientific Directorate, UEM; United Nations-World Institute for Development Economics Research; World Food Program; and the main project institutes namely, CEPPAG, IFPRI, ITC, University of Twente, and SNU.

Public link

Please use the above public link if you want to share this noodl on another website.