Guide

The BonaRes Knowledge Library structures published research on agricultural soil management. For every paper it holds not only the bibliography but the study behind it: where it was done (sites, soils, climate), what was compared (treatments), what was measured, and what was found, each result as practice → soil property with its direction and significance.

Soil processes depend on site conditions. That is why the library keeps them with every result: you can ask not only "what does zero tillage do to soil carbon?" but "on fine-textured soils, in temperate climate, in the topsoil?".

The library is a curated selection, not a complete index of the literature. For a systematic review, use it for scoping and screening alongside database searches (Web of Science, Scopus, CAB Abstracts).

The search reads titles, authors, DOIs, abstracts and the terms of each study, so "no-till" also finds a study whose treatments are linked to "Zero tillage". When your words match a catalogue concept, the library offers it as a filter: a concept filter includes everything narrower (Tillage system includes zero, reduced and conventional tillage).

• Practices, soil properties and land use: catalogue concepts found in the studies.
• Study type: field or laboratory experiment, meta-analysis, model or method study.
• Soil group, texture and climate zone of the sites; year of publication.
• "Read in detail": only papers whose study is in the library, not just the bibliography.

Every search is a URL: bookmark it, share it, or cite it ("Cite this search" copies text with date and link).

A paper page shows the findings grouped by measured property, the set-up (treatments with their levels, measured variables), the sites with climate, soil and layers, and similar papers (those sharing rare concepts). Open a finding to see its numbers and the passage of the paper it was read from.

The box "At a glance" says how the study came into the library: entered by hand in the old library, or read by the library and checked by a person, and whether that check is complete.

The evidence page answers one question at a time: what does a practice do to a soil property? It counts one result per study, so a paper with forty sampling dates does not outvote ten others, and gives the share of studies with a 95 % confidence interval, the typical size of the effect where studies printed numbers, and how much of it was checked by a person.

• When does it work: the result split by precipitation, temperature, texture, depth, duration and study type, with a permutation test (corrected for testing several conditions) that names the conditions that make a difference.
• Under my conditions: enter the precipitation, temperature, texture and depth of your site and the answer is drawn only from comparable studies.
• What was never studied: precipitation × texture combinations without a single study of the question.
• For your manuscript: method, result and the studies as a paragraph to adopt, with the date you consulted the library.

This is vote counting, not a meta-analysis: a small pot trial counts as much as a 40-year field experiment, and "no significant effect" often means "too little power". Differences between conditions are hints, not controlled comparisons. Use it to find, group and weigh studies; export the findings (CSV, with numbers where the paper gives them) for a proper analysis.

The map shows where studies were done; it takes the same filters as the search, and the list beside it follows what is in view.

Explore shows the library as a space of worlds (topic fields) whose currents are the findings between them. Click a world to unfold its concepts, a concept to see its effects, a relation to see every finding as a bead you can point at, each a paper. A one-minute guided tour shows the main results.

Chains of effects: from any concept, choose a target and see how the first may act on the second through one or two intermediate effects, next to what was studied directly. A chain joins findings from different studies; it is a hypothesis, not proof, and steps that rest on correlations only are marked.

• Sign in and drop the PDF (several at once are fine). Nothing to type: the bibliography comes from the DOI on the first page (Crossref, OpenAlex).
• If the paper is already in the library, you are told at once, and nothing is read twice.
• The library reads the paper with its tables and figures, records its main findings, checks each of them once more against the paper and verifies the quotes against the PDF. This takes a minute or two; you can leave the page.
• You then check only what the library is unsure of, usually about ten items, each with its source page beside it. Correct, wrong, or choose a better catalogue term; then publish.

The PDF is used only for reading and checking, is never shown to others and is deleted when you publish or discard the paper.

Every item read by the library carries the page and quote it came from, and a confidence: sure (two readings agree, the quote is in the PDF, the numbers fit), check, or doubtful. Only check and doubtful items are shown to the person who checks; on public pages, unconfirmed items are marked as such.

Every decision made in the check is stored. It is used to recalibrate the confidence and to improve how terms are linked to the catalogue, so the library gets better the more it is used.

Older entries were entered by hand in the first version of the library; they have no page references and are marked "entered by hand".

• References as RIS (Zotero, EndNote, Citavi, Mendeley) or BibTeX, for a search, your saved papers or one paper.
• Papers as a table (CSV) and findings as a table: one row per finding with driver, comparator, response, direction, significance, depth, site conditions, numbers and source.
• Machine-readable metadata for harvesters: schema.org JSON-LD at /schema/v1/soildocs.

Reading needs no account. With one you can add papers, save papers to your reading list and export it. Accounts of the old library work with the same username and password.