Rwanda - Rwanda Seasonal Agricultural Survey 2024

To provide the basis for conducting probability surveys that comprehensively cover farm-level data and to enhance the precision of survey estimates, SAS uses a Multiple Frame Sampling (MFS) methodology. This approach involves constructing an area frame from which the survey sample is drawn. In addition, a list frame of Large-Scale Farmers (LSF), with at least 10 hectares of agricultural land, is done to complement the area frame. This ensures coverage of crops predominantly cultivated by large-scale farmers, which may not be adequately represented in the area frame alone. The construction of an area frame involves several steps, including land cover classification, land stratification and sampling of segments.

Land classification is the first step in the designing of the sampling frame of the Seasonal Agriculture Survey. This process involves categorizing the total available land in the country into different land use or land cover types with the purpose of enhancing sampling precision by targeting the adequate land. With a combination of different spatial layers available in the country, plus a photo interpretation of a series (2010 to 2023) of high-resolution (50 to 30 cm) satellite images the total land of the country was divided into 14 land cover classes Among 14 land cover classes, only 6 are related to agricultural activities include Agricultural land on hillside, non-rice agricultural Wetland, mixed rangeland, Low-density built-up area, wetlands designated for Paddy rice and Tea plantation. The subsequent step involves constructing the area frame which includes grouping the land cover classes linked to agricultural activities into strata to identify agricultural strata to be considered in the sampling frame

The stratification is a result of a combination of sampling units (clusters) and land use/land cover. The stratification assigns each cluster a stratum based on the predominant land class type. Among the fourteen land cover classes, four are included in the agricultural survey frame, while the others are excluded.
The included land cover classes comprise hillside agricultural land, non-rice agricultural land, mixed rangeland, and Low-density built-up area (with potential for agricultural production, including kitchen gardens, fruit trees, and livestock). Certain agricultural land classes are excluded from the sampling frame. For instance, tea plantations are omitted due to regular monitoring by the National Agricultural Export Development Board (NAEB), and wetlands designated for paddy rice cultivation are typically considered in Large-Scale Farmers, making them another component of the survey frame. Moreover, Since the 2024 SAS, a new land cover class called Exclusive Rangeland has been introduced specifically for areas used for pastoral activities. This class is also excluded from the sampling frame.

Out of Five defined strata, only dominant hill crop land stratum, dominant wetland crops stratum, dominant rangeland stratum and mixed stratum are considered as land potential for agriculture. The remaining stratum is the non-agricultural land. Note that clusters covered by tea plantations and wetlands designated for paddy rice cultivation are not considered in the area sample frame due to reasons stated above. Thus, SAS is conducted on 4 above mentioned strata. At first stage,1200 segments are selected and allocated at district level based on the power allocation approach (Bankier, 19881). Sampled segments inside each district are distributed among strata with a proportional-to-area criterion. Specifically, for Season C, a shorter season that does not cover the entire land potential for agriculture, the sample was selected only from three sub-strata: Dominant Wetland, Season A and B frames in the volcanic agro-ecological zone with consistent rainfall, and special sites with irrigation infrastructure. For Season C, 946 segments were selected and allocated at the district level.

At the second stage, 25 sample points are systematically selected, following a special distance of 60 meters between points. Sample points serve as reporting units within each segment. Enumerators visit each point, identify and delineate the plots in which the sample point falls, and collect records of land use and related information.
The recorded information represents the characteristics of the whole segment which are extrapolated to the stratum level and hence the combination of strata within each district provides district area related statistics.


The stratified two-stage sample design used with the new area frame, the first stage sampling probability for the sample segments in each stratum was calculated.

The second stage probability was calculated at the plot level based on the assumption that the plots within each sample segment were implicitly selected with PPS using the area of the plot as the measure of size.