How We Rate Evidence: Agriculture

This page explains how the Agriculture & Farming section of CAB Direct judges the evidence behind the claims in its articles. The section covers regenerative farming, precision agriculture, GM and biotech crops, and agricultural markets and subsidies, with topics that range from no-till, cover crops and irrigation to Bt crops, GMO detection, crop origins, seed banks, yield gaps, food loss and waste, and commodity prices.

The approach described here is a working method for weighing sources claim by claim. It is not a formal academic grading system, and it does not give an article a single numeric score. The general, site-wide version of the approach is at /how-we-rate-evidence/; this page adapts it to agricultural questions, where a result often depends on the soil, weather, crop, management and prices of the place and year in which it was measured.

A claim-by-claim approach

An agricultural article usually contains several kinds of claim at once: a measured result from a trial, a figure from official statistics, an estimate from a model and, sometimes, a statement made by a company selling a product or service. Each claim is assessed against the evidence that supports it, rather than labelling a whole article as strong or weak. A sentence on national wheat yields, a sentence on how cover crops affected soil organic matter in one long-term trial and a sentence on the input savings claimed for a piece of machinery can appear in the same article and carry very different levels of certainty.

For every claim, the questions are the same: what kind of source supports it, what exactly was measured and under which conditions, and whether the wording of the article stays within what that source shows.

What we assess

The points below guide the assessment of an individual claim. Not every point applies to every source: a commodity price series raises different questions from a plot trial or a survey of farmers.

• Direct relevance: whether the source measured the outcome the article discusses, such as grain yield, gross margin, soil organic carbon or water use, rather than a related indicator.
• Study design: whether the result comes from a replicated field trial, an on-farm comparison, a farm survey, a model or a promotional claim.
• Scale and duration: the number of sites, plots and seasons, and whether the work ran long enough for slow processes such as changes in soil organic matter or tree growth in agroforestry.
• Conditions: soil type, climate and rainfall in the years studied, crop and variety, and the farming system in which the result was obtained.
• Units and baselines: whether figures are per hectare, per rotation, per kilogram of product or per farm, and what the comparison treatment actually was.
• Consistency: whether other trials, surveys and reviews point in the same direction or disagree.
• Uncertainty and limitations: ranges, variation between sites and years, and the limitations stated by the study authors.
• Source and interests: who published the figure, who funded the work where this is disclosed, and whether the source sells the product, input or service being assessed.

How different source types are used

Agricultural evidence comes from several kinds of source, and each answers a different question. The section uses them as follows.

Field trials and multi-site trials

Replicated field trials are the main source for claims about how a practice, input or variety affects yield, soil or crop health under defined conditions. We look at the trial design, the number of replicates, the treatments compared and the seasons covered. A result from a single site in a single season is reported as a result for that site and season. Multi-site trials, and trials repeated over several years, carry more weight for general statements because they show whether an effect holds across different soils and weather. Results from glasshouses, pots or laboratory studies are described as such and are not presented as field performance.

Farm surveys and official statistics

Farm surveys and official statistics describe what happens across many farms: areas planted, yields, input use, prices, incomes and support payments. They are the main source for figures on scale and trends, such as national crop production or the uptake of a practice. Articles state the year, the geography and the body that published the figure, and note changes in definitions or survey methods where these affect comparisons over time. Survey data show patterns and associations; on their own they do not show that a practice caused a difference in yield or income, because farms that adopt a practice often differ from other farms in several ways.

Meta-analyses and systematic reviews

Meta-analyses pool results from many studies and are often the best available summary for broad comparisons, such as organic and conventional yields, no-till and ploughed systems, or the effects of Bt crops on non-target organisms. We check which studies were included, how the comparison was defined and whether a pooled average hides large differences between crops, regions or management. A meta-analysis is only as reliable as the studies it draws on, so where the included trials are few, short or concentrated in one region, the article says so.

Models and economic projections

Crop models, yield gap estimates, digital soil maps and projections of commodity prices or of the effects of subsidy changes are estimates built on assumptions. They are useful for exploring scenarios and for filling gaps where measurements are sparse, but they are not observations. When an article draws on a model or projection, it identifies it as such, names the main assumptions where they affect the conclusion, and keeps the stated range or scenarios rather than presenting one number as a forecast.

Industry and manufacturer claims

Machinery manufacturers, input suppliers, technology companies and trade bodies publish figures on yield gains, input savings, water use or returns on investment. These figures can point to useful information, but they come from organisations with an interest in the result. Articles report them as the claims of the organisation that made them, look for independent trials or data that test the same claim, and do not present a manufacturer figure as an established finding when no independent evidence is available. This applies, for example, to claimed savings from drones, sensors, variable-rate application or vertical farming systems.

From one study to a broader conclusion

A single trial rarely settles an agricultural question. A cover crop that raised soil organic matter on one farm, or a precision system that reduced fertiliser use in one season, is reported as a finding from that setting. Broader statements, such as whether a practice generally improves yields or margins, need support from several trials in different conditions or from a well-conducted review.

Where studies disagree, the article describes the disagreement and its likely reasons, such as differences in soil, climate, trial duration or how the comparison was set up, rather than choosing the result that fits a preferred conclusion. A new study that contradicts earlier work is reported alongside that work, not as an automatic replacement for it.

Scope and applicability: soils, climates and farm systems

Agricultural results often travel less well than headlines suggest. A reduced-tillage result from a well-drained loam in a dry climate may not hold on heavy clay in a wet one. A yield comparison between organic and conventional systems depends on the crops, rotation and inputs involved. An economic result depends on the prices, costs and support schemes in force in the country and year studied.

Articles therefore state where and when the evidence was collected and which farm system it applies to, including farm scale, whether land was irrigated or rainfed, and whether results come from research stations or commercial farms. When a finding is applied to a different region or system, the article makes clear that this is an extrapolation. Figures on prices, subsidies and regulation are given with their date, because they change over time. Articles explain what the evidence shows; they are not agronomic, financial or legal advice for a particular farm.

Matching language to evidence strength

The wording of an article should signal how strong the evidence behind each claim is. The section keeps a small set of distinctions:

• “Found” or “showed” is used for results measured in the study being described, with the setting stated.
• “Is associated with” is used for patterns in surveys or observational data where cause and effect have not been established.
• “Estimated” or “projected” is used for the outputs of models, yield gap analyses and price or policy scenarios.
• “Claimed” or “reported by” is used for figures from manufacturers, suppliers or trade bodies that have not been independently tested.
• “Suggests” or “early evidence” is used when support comes from a few small, short or single-site studies, or from glasshouse and laboratory work.

Articles avoid words such as “proven”, “guaranteed” or “always” for farming outcomes, because results vary with season, site and management. They also avoid turning a finding about one crop, region or system into general advice for all farms.

Review, updates and corrections

Claims and figures in Agriculture articles are checked against traceable sources, and every number is given with its source and date. Each article names its author with a stated role. When a cited source is updated, corrected or withdrawn, or when newer evidence changes the picture, the affected wording or figure can be revised. Corrections are dated and remain visible in the article.

This page does not set a fixed interval for re-checking evidence. For the wider publishing rules, see /editorial-policy/; for how published fixes are recorded, see /corrections/. Readers who think a claim overstates or misreads its source can raise it through /corrections/ or /contacts/.

Who applies this approach

The approach is applied by the authors of the Agriculture & Farming section, each within their stated coverage. All five are listed with the role of Author.

Harriet Colbeck covers regenerative farming, soil organic matter, cover crops, reduced tillage and nutrient management, where long-term trials and soil sampling methods matter. Tobias Mardle covers precision agriculture, including yield mapping, sensors, variable-rate application, drones and decision-support software, where manufacturer claims need to be separated from independent tests. Callum Hebden covers agricultural markets and prices, farm subsidies, land use, and the regulation and farm-level economics of GM and gene-edited crops, where official statistics and economic projections are central. Anjali Dhanjal covers plant breeding, the science of GM and gene-edited crops, GMO detection, crop origins, seed banks and Bt crops, where laboratory, glasshouse and field results have to be kept apart. Nuala Keaveney covers organic and conventional farming, irrigation, yield gaps, vertical farming, food loss and waste, and agroforestry trials, where the unit of comparison can change the result.

Author profiles and credentials are at /authors/.

Historical context

CAB Direct also has a separate history as a CABI research database platform. CABI traces its beginnings to an entomological committee in 1910 and records that, in 1973, the contents of its abstract journals were unified into the CAB Abstracts database. According to CABI, CAB Direct launched in 2003 as CABI’s own platform for CAB Abstracts. CABI’s help pages now describe searching the CABI Digital Library, in particular its main databases, CAB Abstracts and Global Health.

This page describes the evidence approach used by the current Agriculture & Farming editorial section. It should not be read as a claim that the historical CAB Direct database used this framework, or that CABI publishes, reviews or endorses this section.

CABI history: .

CABI Digital Library help: .

Related pages and contact

For background on the project, see /about/, and for the section hub, see /agriculture/. The site-wide evidence page is /how-we-rate-evidence/ and the editorial rules are at /editorial-policy/. Questions about how evidence is handled in a published Agriculture article can be sent through /contacts/.