The estimation of agricultural production in Argentina is of fundamental importance due to its key role among the country’s main macroeconomic variables. Although the agricultural sector contributes approximately 6% of GDP, agro-industrial exports account for about 35% of Argentina’s total exports, with a significant impact on national accounts. Moreover, the information generated is required not only by government offices but also by other economic and social stakeholders, such as producers’ organizations, exporters, input suppliers, insurance agencies, and NGOs, among others.
This document describes the methodology implemented by the Dirección Nacional de Estimaciones Agrícolas of Argentina for estimating the sown area of crops.
Estimation methodologies have evolved over the years, shifting from highly subjective approaches to more objective and reliable methods. By the mid-2000s, the difficulty of obtaining satellite images with adequate spatial and temporal resolution and free of cloud cover, combined with limited data processing capacity, represented a significant obstacle, often resulting in unsatisfactory outcomes. In this context, the Random Segments Method (RSM) was developed as a way to obtain more precise, robust, and independent estimates, regardless of the availability of suitable satellite imagery. The method combines statistical tools with elements of remote sensing and GIS.
The RSM consists of a probabilistic, stratified area sampling design. Strata are defined as zones within each district (Partido or Departamento) that exhibit homogeneous agricultural land use. Four types of strata are established: “A,” predominantly agricultural zones; “B,” zones with moderate agricultural coverage (mixed areas); “C,” zones with low agricultural coverage (livestock areas); and “D,” zones with low or null probability of agricultural use, such as cities, lagoons, mountain ranges, etc.
Based on this stratification, random points are selected and relocated to accessible roads, then converted into 4-kilometer lines, which are expanded 500 meters on each side, generating segments of approximately 400 hectares. Each segment constitutes the sampling unit and is fully surveyed, identifying the different Land Use Units (LUUs), both agricultural (crops, pastures, fallows, stubble, natural grasslands, temporary fields) and non-agricultural (urban areas, infrastructure, water bodies, among others).
Field operations include the delimitation and characterization of LUUs through direct observation and georeferenced recording via a mobile application, followed by data download and systematization. Surveys are conducted twice a year: one for summer crops and another for winter crops. From these data, estimates are generated and then expanded to the population level using statistical procedures. The methodology allows the calculation of sampling errors, such as standard error and coefficient of variation, ensuring the assessment of result accuracy.
The method leverages the presence of delegations across the productive territory, and by keeping segments fixed, it enables the monitoring of each sample point over time. Its random distribution has allowed, particularly in its early stages, the inclusion of regions in national statistics that would otherwise have remained underestimated. Currently, approximately 4,800 segments are surveyed across an agricultural area estimated in 2024 at 50 million hectares.
In addition, field data serve as an optimal input for the production of crop maps, a product that complements and simultaneously supports the RSM when stratification updates are required due to land use changes.
