By Richard Tonkin, Statistician, ESCAP, and Setia Pramana, Director of Statistical Methodology and Data Science, BPS-Statistics Indonesia

Rice is a staple food across much of Asia and the Pacific, feeding over half the world’s population. Accurate estimates of rice production are therefore central to supporting food security, rural livelihoods, price stability, and sustainable agriculture. Yet producing timely, reliable, and sufficiently granular statistics on rice production remains a persistent challenge for many national statistical offices (NSOs).

Traditional survey-based approaches continue to play a critical role in agriculture statistics, but they are under increasing pressure. Survey fieldwork can be slow and expensive, and data coverage can be limited in remote or hard-to-reach areas. At the same time, policymakers are requesting more frequent updates and more detailed subnational information to reduce the risk of resource misallocation and mitigate food insecurity. Together, these pressures are prompting NSOs to explore new data sources, such as Earth Observation (EO) data, that can complement existing information.

The opportunity of Earth Observation data

Satellite imagery offers several clear advantages for producing official crop statistics. It provides wide geographic coverage, can be updated frequently, and allows consistent observation of crop conditions over time. For a crop such as rice, with distinctive phenological patterns, EO data can support crop mapping, monitoring of planting and harvesting cycles, and estimation of production yields. This opens the possibility of producing statistics even for small geographical areas, alongside national-level estimates.

However, EO data are not sufficient on their own. Machine learning models, based on satellite imagery, need to be trained with robust ground truth data that reflect different landscapes, climates, and growing environments. Equally important are the statistical standards and institutional processes needed to ensure estimates can be effectively validated, interpreted, and communicated to users. For these reasons, the way EO data are incorporated into official statistical production matters as much as the technology itself.

Integrating EO into official statistics: Indonesia’s approach

Indonesia provides an example of how these challenges can be addressed in practice. Through the One Data in Rice initiative, Statistics Indonesia (BPS) has led the development of a ‘Mixed Method approach’ that combines satellite imagery with ground-truth data from their Area Sampling Frame survey to improve rice production estimates in terms of accuracy, granularity, timeliness, and cost-effectiveness.

In developing the mixed-methods approach, BPS concluded that Indonesia’s diverse landscapes and cropping systems meant a single national model would not be sufficient. Instead, separate models are required for each region, reinforcing the need for high-quality training data and ongoing field validation. The initiative has also required close collaboration across government, bringing the NSO together with partners including the National Research and Innovation Agency (BRIN), the Ministry of National Development Planning (Bappenas), and the Ministry of Agriculture. This institutional coordination has been critical to ensuring that new methods align with policy needs and are embedded within existing statistical and data processes.

At the international level, the work has been supported through the UN Economic and Social Commission for Asia and the Pacific’s (ESCAP) project on Big Data for Official Statistics, funded by the 2030 Agenda sub-fund of the UN Peace and Development Trust Fund. A core focus of our work in this area is supporting countries in moving beyond one-off pilots or experiments, embedding new sources and methods into the regular production of official statistics. In Indonesia, we’ve worked with the Food and Agriculture Organization of the UN (FAO) to draw on expertise from academia, government, and the private sector, ensuring BPS have access to the latest methodological and technical developments. Through this support, the Indonesian government have successfully established an operational rice monitoring workflow using EO data, hosted by the National Computing Centre, supporting scalability and long-term sustainability.

Learning from practice

To support knowledge sharing, we have documented Indonesia’s experience in a video aimed at statisticians and policymakers interested in using EO for rice production statistics. The video goes through the approach, tools, and institutional arrangements in practical terms. It is intended to go beyond a promotional overview, serving instead as a real learning resource for national governments considering similar approaches.

As demand for more timely and detailed agricultural statistics continues to grow across the region, Indonesia’s experience highlights both the potential and the challenges of integrating innovative data sources into official statistics.