Symptoms are clues, not diagnoses
Observe patterns, preserve multiple possible causes, and confirm the cause when the decision warrants it.
Earth Project reviewed collection · v1.6
Nine Foundations for Better Soil Research
These reviewed Earth Project topics are reasoning tools, not treatment claims. They help readers observe carefully, measure more than one dimension, distinguish a quick response from durable resilience, preserve method context, separate sequence from causation, interpret statistical thresholds without overstating them, keep conclusions within the settings the evidence supports, distinguish a plausible mechanism from a demonstrated outcome, and avoid treating an indicator as the outcome it is intended to represent.
Publication gate
Recovered does not mean validated.
The Earth Project archive preserves original wording, historical numbering, alternative branches, sources, contradictions, negative findings, and unresolved questions. Public summaries are created separately and released only after a new source and scope review.
Public collection
Nine foundations that can support later topics
Soil health needs multiple indicators
Evaluate physical, chemical, biological, plant, and visual evidence in context rather than relying on one number.
Response is not automatically resilience
Separate an immediate change from persistence, resistance to stress, recovery, and long-term system function.
Keep the number attached to what produced it
Preserve the analyte, value, unit, method or extractant, reporting basis, laboratory flag, sample context, and report date.
A change after treatment is not proof of causation
Use an appropriate comparison and design before attributing an observed change to a treatment.
Statistical significance is not practical importance
Interpret effect size and uncertainty instead of treating one p-value threshold as a verdict.
A result from one setting is not automatically transferable
Keep conclusions within the soils, climates, crops, treatments, and time periods the evidence supports.
A plausible mechanism is not a demonstrated outcome
Use mechanisms to form and test causal explanations, not as substitutes for measured field outcomes.
An indicator is not automatically the outcome it represents
Use indicators as evidence about defined functions, not as automatic substitutes for final outcomes.
Public record IDs organize these edited releases. They do not replace or renumber the permanent historical IDs in the recovered Earth Project archive.
EP-PUB-001 · ReviewedDiagnostic reasoning
Symptoms are clues, not diagnoses.
A visible plant symptom is a system output. Similar symptoms may arise from insects, pathogens, nutrient conditions, soil properties, moisture, weather, light, chemical injury, cultural practices, or interacting causes.
What the evidence supports
Start with the whole plant and its setting. Record timing, distribution, recent management, weather, soil and moisture conditions, and differences between affected and unaffected plants. Use patterns and signs to narrow possibilities, then test or seek qualified diagnosis when warranted.
What it does not establish
A single symptom does not automatically identify a nutrient deficiency, disease, pest, toxicity, or failed soil component. A response after treatment also does not by itself prove that the suspected cause was correct.
Penn State Extension — Diagnosing a Plant Problem 101
Explains that plant symptoms can have numerous biotic, abiotic, environmental, cultural, and interacting causes, and recommends verification through Extension or a diagnostic clinic when needed.
Review the diagnostic methodUC Statewide IPM — Diagnosing the Causes of Unhealthy Plants
Provides a structured approach for examining the plant, site, symptom pattern, timing, signs, and possible combinations of biotic and abiotic factors.
Review the UC IPM questionsScope reviewed: August 2026. Open question: how accurately can a simple gardener field record narrow causes without increasing false confidence?
EP-PUB-002 · ReviewedMeasurement
No single measurement defines soil health.
Soil health describes continued capacity to perform multiple functions. It is not measured directly; indicators provide evidence about physical, chemical, biological, plant, and visual conditions.
What the evidence supports
Choose indicators that relate to a defined function, use consistent methods and sampling conditions, preserve units and context, and look for patterns across indicators and through time. Interpret results within the soil, climate, land use, and management system being studied.
What it does not establish
One organism count, nutrient result, respiration value, organic-matter percentage, crop yield, or visual observation does not provide a complete or universally comparable soil-health verdict.
USDA NRCS — Soil Health Assessment
States that soil health cannot be determined from one outcome and describes useful physical, chemical, biological, plant, and visual indicators.
Review the indicator frameworkUSDA NRCS — Dynamic Soil Properties for Soil Health
Describes work to collect consistent, replicable metrics across diverse soils, regions, and management systems while developing regionally relevant interpretations.
Review the standardized-metrics effortScope reviewed: August 2026. Open question: which smallest set of affordable measurements provides useful information for a specific garden decision without pretending to be a universal score?
EP-PUB-003 · ReviewedTime and resilience
A short-term response is not the same question as resilience.
A rapid change may be real and useful, but resilience asks whether a system resists disturbance, maintains function, recovers, and remains effective across time and relevant conditions.
What the evidence supports
Define the stress and the function of interest before testing. Preserve the baseline, comparison, timing, magnitude of response, recovery trajectory, site and climate context, and repeated observations across seasons or years when making a resilience claim.
What it does not establish
One favorable observation, one growing season, or a simple before-and-after comparison does not automatically demonstrate durability, causation, reduced dependence, or resilience under a different stress, soil, crop, or climate.
USDA ARS — Long-Term Agroecosystem Research Network
Uses coordinated measurements across multiple agroecosystems and long-running research sites to connect local and national agricultural and natural-resource questions.
Explore the LTAR research approachUSDA ARS — LTAR Common Experiment
Coordinates long-term research across diverse working landscapes, supporting comparison across regions and production systems rather than relying on a single short observation.
Review the common experimentScope reviewed: August 2026. Open question: which early field indicators can predict later resistance and recovery with useful, independently validated error rates?
EP-PUB-004 · ReviewedMethod and context
Keep the number attached to what produced it.
A soil-test result should travel with its analyte, original value, unit, method or extractant, reporting basis, laboratory flag, sample context, and report date. Those fields define what was measured and how the result may be interpreted or compared.
What the evidence supports
Preserve the laboratory wording and reported value before any conversion. Record the method or extractant, preparation, reporting basis, flags, reference ranges, and sample context. Compare results only when the measured property, method, unit, basis, and relevant sampling conditions are compatible—or mark the comparison as unresolved.
What it does not establish
Matching units or a documented unit conversion do not make unlike methods equivalent. Different extractants, soil-to-solution ratios, preparation steps, measurement procedures, reporting limits, and calibrations can produce results that represent different measured fractions or support different interpretations.
University of Minnesota Soil Testing Laboratory — Extractable Phosphorus
Documents distinct Bray-1 and Olsen procedures. Both report phosphorus in ppm, but they use different extractants, soil-to-solution ratios, shaking times, and reporting limits.
Review the phosphorus methodsUniversity of Minnesota Soil Testing Laboratory — Our Methods
Lists laboratory analyses with their named extractants and procedures, showing why the method must remain attached to the reported result.
Review the laboratory methodsUniversity of Minnesota Extension — Nutrient Management Calculator
Requires the phosphorus test method and reported value as separate inputs and directs users to select Olsen or Bray according to the soil-test context.
Review the method-aware inputScope reviewed: August 2026. Open question: which minimum laboratory-result fields are necessary for reliable human and machine comparison across time without implying that unlike methods are equivalent?
EP-PUB-005 · ReviewedCausal reasoning
A change after treatment is not proof of causation.
A change observed after a soil amendment, product, or practice may be real, but timing alone cannot separate the treatment effect from weather, field variation, measurement variation, other management, or chance.
What the evidence supports
Define the treatment and outcome before testing. Use an appropriate control or comparison, real independent replication, random assignment where feasible, consistent measurement, and a design that accounts for known field variation. Preserve null results, uncertainty, departures from the design, and limits on where the result may apply.
What it does not establish
Sequence alone does not establish causation. One treated plot, one before-and-after comparison, repeated measurements from the same experimental unit, or an uncontrolled demonstration cannot by itself separate a treatment effect from natural variability, co-occurring changes, measurement error, or chance.
University of Minnesota Extension — How to do research on your farm
Explains the need for control treatments, independent replication, randomization, attention to field variability, and restraint against extrapolating one experiment to other fields.
Review the on-farm research methodSARE — Basics of Experimental Design
Describes how replication, blocking, and randomization reduce experimental error and help separate treatment effects from natural field variability.
Review the experimental-design basicsNIST — Completely Randomized Designs
Documents randomized assignment, replication, balance, and response comparison as defining elements of a completely randomized experiment.
Review the randomized-design referenceScope reviewed: August 2026. Open question: which practical small-scale garden designs provide the strongest causal information while remaining affordable and usable for first-time experimenters?
EP-PUB-006 · ReviewedStatistical interpretation
Statistical significance is not practical importance.
Statistical significance addresses how data compare with a specified model and null hypothesis. It does not measure effect size, practical value, or the probability that a scientific claim is true.
What the evidence supports
Report the estimated effect and its uncertainty, not only whether a p-value crosses a threshold. Preserve the study design, sample size, model assumptions, outcome definition, analysis choices, and practical threshold that would make the effect useful. Consider the full body of relevant evidence rather than treating one dichotomized result as a verdict.
What it does not establish
A small p-value does not by itself show that an effect is large, important, probable, causal, reproducible, or applicable elsewhere. A large p-value does not by itself show that an effect is absent or unimportant. Large samples can make small effects statistically detectable, while small samples can leave practically important effects uncertain.
American Statistical Association — Statement on p-Values
States that p-values do not measure effect size or importance, do not provide the probability that a hypothesis is true, and should not alone determine scientific or policy decisions.
Read the ASA statementAmerican Statistical Association — Moving to a World Beyond p < 0.05
Explains why a threshold label cannot establish the truth, presence, or importance of an effect and argues for interpretation in scientific context without bright-line declarations.
Read the ASA editorialNIST — Practical Versus Statistical Significance
Shows how sample size can make a small difference statistically detectable or leave a large practical difference statistically uncertain.
Review the NIST guidanceScope reviewed: August 2026. Open question: which effect sizes and practical thresholds are meaningful for specific soil and garden outcomes across different climates, soils, crops, time scales, and decision costs?
EP-PUB-007 · ReviewedExternal validity
A result from one setting is not automatically transferable.
A result may be valid for the field, soil, climate, crop, treatment, dose, management system, outcome, and time period that were studied without establishing the same result elsewhere.
What the evidence supports
Define the setting or population to which a conclusion is intended to apply. Preserve soil, climate, crop, treatment and dose, management history, outcome, timing, scale, and sampling conditions. Distinguish direct evidence from extrapolation, examine effect modifiers, and use compatible multi-site, multi-season, or local validation when broader application matters.
What it does not establish
Internal validity, statistical significance, or one successful local trial does not automatically establish that an effect will transfer to another setting. Combining unlike sites without examining compatibility can hide important differences. Failure to transfer does not necessarily invalidate the original local result; it may reveal a scope boundary or effect modifier.
University of Minnesota Extension — How to do research on your farm
Explains that an on-farm experiment represents a subset of conditions and warns against assuming that one experiment must be valid for other fields.
Review the limits of on-farm inferenceUSDA ARS — Spatial and Temporal Variability
Describes how soil properties vary within and among fields across space and time and may be influenced by vegetation, management history, and weather.
Review the variability contextUSDA ARS — LTAR Common Experiment
Coordinates long-term research across diverse working landscapes so findings can be compared across sites and translated beyond a single local observation.
Review the multi-site research approachScope reviewed: August 2026. Open question: which soil, climate, crop, treatment, management, and time variables most strongly determine whether a result transfers to a new setting?
EP-PUB-008 · ReviewedMechanistic reasoning
A plausible mechanism is not a demonstrated outcome.
A scientifically plausible pathway can explain how a product, amendment, organism, or practice could affect a soil or plant process. It does not by itself show that the pathway operated at a relevant dose, produced the claimed outcome, or will do so under field conditions.
What the evidence supports
Use mechanistic evidence to define a testable causal pathway, identify intermediate measurements, and predict what should occur if the pathway is operating. Then seek evidence that the exposure, intermediate steps, and defined outcome occur at relevant magnitudes and times. Weigh the mechanism with direct outcome evidence, alternative causes, controls or comparisons, replication, uncertainty, and field context.
What it does not establish
A plausible pathway, ingredient function, laboratory reaction, biomarker, intermediate process, or explanatory story does not by itself establish that the final soil or plant outcome occurred, was caused by the intervention, was practically important, or will occur at a field-relevant dose. Absence of a known mechanism does not by itself refute an observed effect; an implausible or contradicted mechanism can weaken a claim.
US EPA — About Causal Assessment
Describes causal assessment as evaluation of all relevant evidence, including alternative causes, rather than reliance on a single type of evidence.
Review the causal-assessment frameworkUS EPA — Evidence of Exposure or Biological Mechanism
Explains that evidence of exposure or a biological mechanism can strongly support a candidate cause but is not convincing by itself because it does not show that the exposure or mechanistic action was sufficient to cause the effect.
Review the mechanism-evidence boundaryUS EPA — Step 5: Identify Probable Causes
Evaluates candidate causes through the strength and consistency of evidence across multiple causal characteristics and preserves uncertainty when the evidence is incomplete.
Review the weight-of-evidence stepScope reviewed: August 2026. Open question: which intermediate measurements most reliably predict meaningful field outcomes for specific soil and plant processes?
EP-PUB-009 · ReviewedIndicator validity
An indicator is not automatically the outcome it represents.
An indicator can provide useful evidence about a defined soil property, process, or function. It does not automatically replace direct measurement of the broader outcome or prove that a change in the indicator predicts a meaningful change in that outcome.
What the evidence supports
Define the outcome or function of interest before selecting an indicator. Preserve the indicator's measurement method, context, timing, scale, uncertainty, and intended use. Test whether it is sensitive to the relevant change and whether its relationship with the outcome is reliable in the soils, climates, crops, management systems, and decisions where it will be used.
What it does not establish
A change in soil respiration, microbial biomass, enzyme activity, nutrient concentration, organic matter, plant color, or another intermediate measure does not by itself establish improved soil health, crop performance, resilience, environmental benefit, or decision value. Correlation, biological plausibility, or analytical accuracy alone does not validate an indicator as a substitute for a different final outcome.
USDA NRCS — Soil Health Assessment
Explains that soil health is not measured directly, that indicators provide clues about soil functions, and that useful indicators should relate to changes in those functions and be interpreted through patterns and comparisons.
Review the soil-indicator frameworkUS EPA — Assessment Endpoints
Describes endpoints as measurable expressions linked to ecological conditions and management goals, with relevance, sensitivity, measurability, and adequate data among the characteristics needed for their use.
Review the endpoint characteristicsFDA–NIH Biomarker Working Group — Validation
Distinguishes measuring a marker accurately from showing that it measures or predicts the relevant concept, and explains that validation is tied to a specified purpose and context of use.
Review the fit-for-purpose validation principleScope reviewed: August 2026. Open question: which affordable soil and plant indicators reliably predict defined garden outcomes, for which decisions and settings, and with what error rates?
Collection rule