Collecting and managing data

General principles

Standardizing ontologies and schema

To enable the harmonisation of feed data from different data sources and across various production systems and geographies, a shared ontology and a standardized data schema are ideals. An ontology defines common concepts and relationships (e.g., feed types, nutrient composition, emission factors, and their interlinkages), ensuring consistent meaning across datasets. A schema specifies how these data are structured, including field names, units, formats, and database relationships, enabling interoperability and aggregation. Together, they ensure statistical robustness, comparability, and compatibility with national monitoring, reporting, and verification and climate finance reporting systems. Given the existence and use of several ontologies and data schemas, a single globally shared ontology and standardized data schema might be an unrealistic expectation. In reality, recognising and reusing existing ontologies and schema where possible moves us closer to harmonization. Feed data resources collected and handled using several consistent ontologies and schema are more easily harmonised, as opposed to those collected and handled using an infinite number of approaches. 

Development projects, public initiatives, and private investment generate huge amounts of data. Collecting, using, generating and storing data is costly and takes time. When data isn’t FAIR (Findable, Accessible, Interoperable, and Reusable), the effort of collecting data is often duplicated, or data is lost or overlooked, meaning its potential value is not realised (CABI 2021). Collecting, managing, and storing data in a FAIR way has several advantages:

• Findable: Data is assigned unique, persistent identifiers (e.g., DOI) and described using rich metadata, making it easily indexed in searchable resources. 
• Accessible: Once found, users can access data through clear, standardized, and secure protocols. 
• Interoperable: Data uses common, machine-readable formats and controlled vocabularies, allowing it to be integrated with other systems and data sets. 
• Reusable:Data has clear usage licenses and accurate documentation, allowing it to be properly cited and repurposed for future research. 

This Three-point FAIRification Framework from GO FAIR provides practical “how to” guidance to stakeholders seeking to go FAIR. By following this framework, stakeholders can ensure their efforts toward FAIR data principles will be optimally coordinated with the efforts of other stakeholders in the GO FAIR community. The three-point framework maximizes reuse of existing resources, maximizes interoperability, and accelerates convergence on standards and technologies supporting FAIR data and services. GO FAIR is a bottom-up, stakeholder-driven and self-governed initiative that aims to implement the FAIR data principles, making data Findable, Accessible, Interoperable and Reusable.

And additional resource is the FAIR Process Framework which was developed by the Gates Foundation specifically for agricultural investment projects. This six-step framework was designed in collaboration with over 32 global partners. It offers a comprehensive suite of tools to guide users in implementing FAIR principles across data-rich agricultural development investments. Among the tools is a directory of repositories and other data platforms and services.

Deciding where to store data is an important step to ensuring that data is reliably accessible to users. This resource from The Turing Way provides guidance on choosing ideal repositories to store data. The Turing Way is an open science, open collaboration, and community-driven handbook to reproducible, ethical and collaborative data science. 

The table below summarizes the suitability of methods for collecting data on parameters. Green indicates high suitability, yellow indicates moderate suitability, and white indicates low suitability.

Overview 

A sample survey is a way of obtaining data about a population (e.g. pastoral cattle herders in Ethiopia, or fish-farming households in Chittagong) by collecting data from a sample of this population. To draw statistically-based conclusions about the wider population from the sample, probability-based sampling is required. A census is a type of survey that targets the entire population, rather than a sub-sample. Because of their smaller sample size, surveys are often capable of collecting more comprehensive data than agricultural or livestock censuses(FAO 2018a). In general, household surveys have households as the population (whether or not they are engaged in agriculture), while agricultural surveys have farm holdings as the population. In practice, integrated household and agricultural surveys are often used, such as the World Bank's Living Standards Measurement Study. 

Collecting Data 

Surveys can be undertaken in a range of ways, such as written questionnaires, online forms, and face-to-face interviews. Data can also be collected via mobile devices such as tablets and smartphones. Open Data Kit (ODK), is a system for facilitating data collection using mobile devices. Appendix B of this resource guide provides more information on ODK.

Key Issues 

Sample surveys can involve considerable costs. These costs can be reduced by coordinating with other surveys to avoid the same data being gathered multiple times. 

Further, a clearly defined sampling frame is essential to ensure that feed basket data are statistically representative and comparable over time. The sampling frame provides a consistent structure for selecting farms and herds across defined geographic areas and production system strata. This enables robust aggregation from individual observations to higher administrative levels and broader geographies. According to the FAO, “using census sampling frames can ensure that the indicators derived from a survey are consistent with census estimates of other parameters”(FAO & GRA 2020). 

Additionally, according to the FAO, face-to-face interviews “are beset by a range of limitations that may reduce the quality of the data being collected, including heavy reliance on farmer memory and/or record-keeping, which may vary between farmers in terms of accuracy and reliability…To overcome the limitations of direct interview, respondent answers may be supported by direct measurement from the enumerator, focus groups or technical coefficients and data from administrative reports and records”(FAO 2018a). Enumerators are the single most important determinant of feed basket data quality as they are in direct contact with livestock keepers and therefore shape the accuracy, completeness, and credibility of the data collected. For guidance on enumerator training, see Appendix A.

The way in which a survey is administered can affect the results, for example a study in India (Anderson et al. 2024) found that data gathered via phones calls were less precise than those collected in-person. They further noted that “time-series population statistics may be disrupted as survey units update procedures to take advantage of more pervasive information and communication technologies”. 

Resources

1. The World Bank’s Enhancing Climate-Smart Outcomes for Livestock Systems: A How-To Guide Following the Project Cycle(Ng’ang’a et al. 2024) provides the questionnaire used to collect feed data for a dairy cow GHG mitigation project in Bangladesh.
2. FAO’s Greenhouse gas emissions from aquaculture: a life cycle assessment of three Asian systems(Robb et al. 2017) used multiple surveys to collect the feed data required to undertake life-cycle analyses of three Asian aquaculture systems. 
3. Statistics for Sustainable Development (Stats4SD) provide videos on research designs and methodology, including videos on using ODK.
4. FAO’s Guidelines on methods for estimating livestock production and productivity(2018a) provides guidance on gathering livestock data via surveys.

Overview 

Rapid analysis refers to a suite of methods designed to gather feed data in a timely and cost-effective manner. While these methods generally do not employ probability-based sampling—meaning results cannot be used to draw statistically-based conclusions about a wider population—they serve as a vital complement to formal surveys by providing localized or expert-driven insights. Common techniques include expert interviews, stakeholder workshops, and focus groups. These are particularly useful for establishing initial GHG inventories or defining parameters that are relatively consistent across a group, such as regional grass yields. 

Collecting Data 

A primary example of a systematic rapid analysis is the Feed Assessment Tool (FEAST), developed by the International Livestock Research Institute (ILRI). FEAST offers a structured yet rapid approach to assessing local feed resource availability and use. The tool also offers a gendered version, called G-FEAST; this can explore gender influence on feeding practices and the uptake of feed interventions, as well as gender-based variety in opportunities and constraints in livestock feeding. 

• Data Collection Logic: The methodology relies primarily on participatory group discussions with livestock keepers to estimate diet composition. 
• Nutritional Estimation: Rather than laboratory analysis, the digestibility of the resulting diet is estimated using a pre-defined dataset integrated within the tool. 
• Cost-Effectiveness: By utilizing participatory discussions instead of extensive individual questionnaires, the cost of implementing a feed assessment is significantly reduced. 

Key Issues 

While efficient, rapid analysis involves trade-offs in precision: 

• Quantification Challenges: Participatory methods can result in errors that are difficult to quantify due to a heavy reliance on farmer recall. 
• Unit Conversion: There is inherent uncertainty when converting local mass and volume units into standardized kilograms of dry matter. 
• Scope: Results are best suited for local intervention design rather than national-level statistical reporting. 

Resources 

• FEAST Global Site: Comprehensive manuals and the software tool for conducting feed analysis.
  • The Gendered feed assessment tool (G-FEAST) allows users to systematically and rapidly assess feed resources at site level using a gendered lens to develop a site-specific strategy for improving feed supply and utilization through technical or organizational interventions.

Overview 

Direct measurement involves the physical quantification of feed resources, feed allocation, and diet composition at farm level. Direct measurement can also be used to improve the validity of parameters used to estimate feed intake. In livestock systems, this approach enables precise estimation of the actual feed basket consumed by specific animal groups by measuring feeds as offered and, where possible, as consumed. Unlike recall-based surveys, direct measurement reduces reporting bias and improves the accuracy of intake, nutrient supply, and GHG estimation models, particularly in Tier 2 GHG gas inventories (FAO & GRA 2020). This method is especially relevant for validation studies and ration formulation research where higher precision is required. 

Collecting data 

• Feed availability: Feed availability can be measured through direct weighing of harvested fodder, standing biomass sampling, or quantification of purchased feed stocks. For grazing systems, pasture biomass cuts and quadrat sampling can estimate available forage mass (FAO 2018a). 
• Feeding practices: Feeding frequency, allocation, and timing can be directly observed and recorded. In stall-fed systems, enumerators can measure feed offered per feeding session and document refusals to estimate actual intake (FAO & GRA 2020). 
• Feed nutritional content: Representative feed samples can be collected and analyzed for dry matter, crude protein, fibre fractions, and energy content using wet chemistry and/or Near Infrared Spectroscopy (NIRS). NIRS provides rapid and reliable assessment of feed quality (Ndung’u et al. 2019). 
• Feed consumption (percent of different feed items in feed basket): Direct feed basket measurement requires calibrated scales, standardized sampling protocols, and accurate dry matter determination. Proper documentation and quality control are essential to ensure that intake data are consistent and analytically robust (FAO 2018a). While this is the most accurate way to estimate animal diets, it is also very labour intensive (FAO & GRA 2020). Other ways of quantifying diets include various methods for analysing animal faeces, such as micro-histology, near infrared spectroscopy or n-alkane analysis (FAO & GRA 2020). 

Key Issues 

Direct measurement is resource-intensive in terms of time, labour, and analytical costs. Direct measurement is often best applied to a subsample rather than the entire survey population (FAO 2018a). In livestock systems, this approach is commonly used to calibrate or validate recall-based survey data. Challenges include farmer compliance, seasonal variability, and the need for technical expertise in feed sampling and laboratory analysis (Ndung’u et al. 2019).

Resources 

• A new approach for improving emission factors for enteric methane emissions of cattle in smallholder systems of East Africa – Results for Nyando, Western Kenya, (Goopy et al. 2018), used direct measurements of livestock parameters (e.g. cattle weights and milk yield) and pasture yield and digestibility to improve enteric methane emission factors.
• A study titled Improved region-specific emission factors for enteric methane emissions from cattle in smallholder mixed crop: livestock systems of Nandi County, Kenya(Ndungu et al. 2019) employed a mixture of direct measurement and farmer recall used to determine yields. The study combined direct feed measurement with farmer recall to refine methane emission factors in smallholder mixed crop–livestock systems in Kenya. Their study demonstrates how integrating direct measurement improves the reliability of emission estimates and diet characterization under real farm conditions. 

Overview

Earth observation (EO) uses remote sensing technologies to estimate feed availability at scale. This approach enables feed balance assessments across large geographic areas without requiring extensive ground-based data collection. EO methods integrate satellite-derived biomass estimates with land use classification, crop type data, and phenology information to quantify biomass available for ruminant livestock consumption. The approach is particularly valuable for data-constrained regions where traditional survey methods are costly or logistically challenging. 

Collecting Data 

Earth observation (EO) based feed availability is estimated using gridded products including land use, above-ground dry matter productivity, crop type, phenology, biomass burning patterns, and protected area. Dry matter availability for ruminant consumption is calculated by attributing dry matter productivity to land use categories (e.g., cropland, grassland, shrubland, forest) based on proportional coverage. Feedable crop residues are estimated using harvest indices, while grass and browse availability are adjusted for utilisation rates. 

Ground truthing through surveys is essential for validating EO-based estimates. Validation units of approximately 1,000 km² (corresponding to 10 model cells of 10×10 km) are selected to represent different agro-ecological zones and production systems. Power calculations typically suggest 20 community interviews per validation unit provide adequate statistical power for equivalence tests. The validation process compares model cell values to field-collected data, with sampling frames stratified by settlement type, agro-ecological zone, and administrative boundaries. Surveys collect data on livestock populations, herd composition, feed sources, and feeding practices to validate both supply and demand intermediate model outputs. 

Key Issues 

Feed quality is approximated using values from literature and databases, with browse resources being particularly uncertain due to limited understanding of utilisation rates and quality. Grassland and crop residue quality estimates also carry substantial uncertainty. The approach may systematically underestimate feed availability by not accounting for cross-border transhumance, purchased feed, imported feed, and the spatial distribution of water points. Conversely, it may overestimate by not fully capturing the effects of drought stress, nutrient deficiencies, pests, and diseases on actual biomass production. 

Examples 

• A study titled Feed balances for ruminant livestock: gridded estimates for data−constrained regions (Fraval et al. 2024) assessed feed balances for ruminant livestock in Ethiopia and Burkina Faso using remotely sensed products. The study estimated biomass availability at 8.6 tonnes dry matter/ha in Ethiopian highlands, 3.2 tonnes dry matter/ha in Ethiopian lowlands, 2.9 tonnes dry matter/ha in Burkina Faso’s Sudanian zone, and 1.0 tonne dry matter/ha in the Sahel. Feed scarcity hotspots were identified in the Ethiopian highlands and Sahelian zone of Burkina Faso, providing spatially explicit information for targeting interventions. 
• A presentation titled Livestock Feed Assessment and Forecasting with Earth Observation(Mutua 2026) examined a study that assessed feed balances for ruminant livestock in Nigeria. The study estimated 17.7 million tons (8.3%) of dry matter feed deficit, with significant regional disparities. Results revealed a 194.7 million tons supply against 212.4 million tons demand, enabling faster identification of critical shortages, especially in the north.
• A study titledBeyond livestock carrying capacity in the Sahelian and Sudanian zones of West Africa(Rahimi et al. 2021) used EO methods to assess livestock carrying capacity in the Sahelian and Sudanian zones of West Africa, demonstrating the scalability of the approach across multiple countries. 

Resources 

• Livestock Feed Assessment and Forecasting with Earth Observation(Mutua 2026) provides insights on livestock feed assessment and forecasting with earth observation. 
• A research brief titled, AU-IBAR uses feed assessment modelling to support Nigeria and Cameroon livestock investments (Mutua et al. 2025) provides details on feed assessment modelling to support Nigeria and Cameroon livestock investments.
• A presentation titled Ruminant livestock feed balance models – insights from Nigeria(Mutua 2025).
• A study titled Feed balances for ruminant livestock: gridded estimates for data−constrained regions(Fraval et al. 2024) provides a methodology for EO-based feed balance assessment.
• Copernicus Land Monitoring Service provides dry matter productivity estimates.
• Digital Earth Africa provides cropland extent maps.
• NASA’s MODIS Land Cover Dynamics products provide phenology data.