Check a result
Use these checks to investigate whether your own results and methods hold together, whether the work is in progress or already published. They test reported figures against what the method allows, and highlight questions worth answering before you share your work. Nothing you enter leaves your browser.
How to cite: Flynn, T. (2026). Check a result: plausibility checks for soil and environmental results, version 1.0 [software]. https://doi.org/10.5281/zenodo.23736495. Please also cite the method references given with each check. All citations →
Finding the right pixel size
Grid resolution is often chosen by habit (30 m because that’s Landsat). Hengl (2006) set out rules that tie it to what the data can actually support. For each aspect of a dataset they give a coarsest legible resolution (the largest pixel that still represents it), a finest legible resolution (below which pixels “are probably just waste of memory”) and a recommended compromise. Fill in whatever you know; the overall range is where the rules agree.
Reference
Hengl, T. (2006). Finding the right pixel size. Computers & Geosciences 32, 1283–1298. doi:10.1016/j.cageo.2005.11.008
Back-transformation
A mean calculated on log values, then exponentiated, is the median (geometric mean), not the mean. For skewed soil properties the difference is often 10–30%. Enter the value on the log scale.
References
Aitchison, J., Brown, J.A.C. (1957). The Lognormal Distribution. Cambridge University Press.
Journel, A.G. (1980). The lognormal approach to predicting local distributions of selective mining unit grades. Mathematical Geology 12, 285–303. doi:10.1007/BF01029417
Webster, R., Oliver, M.A. (2007). Geostatistics for Environmental Scientists, 2nd ed. Wiley.
Carbon stock arithmetic
Rebuild a reported stock from its inputs. If it doesn’t match, the check tries the common slips (% vs g/kg, stones ignored or on the wrong basis, organic matter instead of carbon, kg/m² vs Mg/ha) and tells you which one reproduces the reported number.
References
Poeplau, C., Vos, C., Don, A. (2017). Soil organic carbon stocks are systematically overestimated by misuse of the parameters bulk density and rock fragment content. SOIL 3, 61–66. doi:10.5194/soil-3-61-2017
Pribyl, D.W. (2010). A critical review of the conventional SOC to SOM conversion factor. Geoderma 156, 75–83. doi:10.1016/j.geoderma.2010.02.003
Area from a confusion matrix
Counting pixels in a map gives biased class areas, because every map has errors. With a stratified accuracy sample, the confusion matrix gives unbiased areas with confidence intervals (Olofsson et al. 2014). Rows are map classes, columns are reference classes, and the last column is each map class’s mapped area.
Reference
Olofsson, P., Foody, G.M., Herold, M., Stehman, S.V., Woodcock, C.E., Wulder, M.A. (2014). Good practices for estimating area and assessing accuracy of land change. Remote Sensing of Environment 148, 42–57. doi:10.1016/j.rse.2014.02.015
Effective sample size
Samples close together carry overlapping information. For estimating an average, a cluster of ten nearby samples can be worth little more than one. This uses the variogram to work out how many independent samples your layout is really worth.
References
Cressie, N.A.C. (1993). Statistics for Spatial Data, revised ed. Wiley. Section 1.3.
Webster, R., Oliver, M.A. (2007). Geostatistics for Environmental Scientists, 2nd ed. Wiley.
Spatial cross-validation and leakage
Random cross-validation on spatial data usually tests points close to the training data, so it reports accuracy the map won’t achieve between samples. Tick anything that applies to your study.
References
Roberts, D.R. et al. (2017). Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography 40, 913–929. doi:10.1111/ecog.02881
Meyer, H., Pebesma, E. (2021). Predicting into unknown space? Estimating the area of applicability of spatial prediction models. Methods in Ecology and Evolution 12, 1620–1633. doi:10.1111/2041-210X.13650
Milà, C., Mateu, J., Pebesma, E., Meyer, H. (2022). Nearest neighbour distance matching leave-one-out cross-validation for map validation. Methods in Ecology and Evolution 13, 1304–1316. doi:10.1111/2041-210X.13851
Independence and study design BETA
Independence is a property of the study design and the question being asked, not a box that a single diagnostic can certify. Use this guided check to identify common sources of dependence and mismatches between the sampling unit, statistical analysis, and validation design. It flags issues to investigate; it does not automatically invalidate a study. If you know how many observations and groups there are, it also shows how much the grouping shrinks the effective sample size.
References
Hurlbert, S.H. (1984). Pseudoreplication and the design of ecological field experiments. Ecological Monographs 54, 187–211. doi:10.2307/1942661
Roberts, D.R. et al. (2017). Cross-validation strategies for data with temporal, spatial, hierarchical, or phylogenetic structure. Ecography 40, 913–929. doi:10.1111/ecog.02881
Meyer, H., Pebesma, E. (2021). Predicting into unknown space? Estimating the area of applicability of spatial prediction models. Methods in Ecology and Evolution 12, 1620–1633. doi:10.1111/2041-210X.13650
Kish, L. (1965). Survey Sampling. Wiley, New York.
Moulton, B.R. (1990). An illustration of a pitfall in estimating the effects of aggregate variables on micro units. Review of Economics and Statistics 72, 334–338. doi:10.2307/2109724
Statistical assumptions check BETA
A guided check for matching assumptions to the statistical method. Select the method and the diagnostics you have checked, then use the prompts to identify anything that may need further investigation. If you have the residuals, paste them: the check reports skewness, outliers, unequal spread, curvature and, for ordered data, autocorrelation, with a residual plot.
Scope note
This beta tool is a structured checklist, not a test of model assumptions. The relevant assumptions depend on the exact statistical procedure and intended inference. An unchecked item only means that you have not marked the check as reported. The residual diagnostic screens one set of residuals; it does not replace a full model diagnosis.
Jarque, C.M., Bera, A.K. (1987). A test for normality of observations and regression residuals. International Statistical Review 55, 163–172. doi:10.2307/1403192
Durbin, J., Watson, G.S. (1950). Testing for serial correlation in least squares regression. I. Biometrika 37, 409–428. doi:10.1093/biomet/37.3-4.409
Model comparison audit BETA
Use this checklist when your work claims that one model performs better than another. It checks comparability of the evaluation design. With paired fold scores it also tests the difference properly (cross-validation folds share training data, so a plain paired t-test is too optimistic), and with the counts of cases where two classifiers disagree it runs McNemar’s test.
Scope note
This beta check examines the comparability of model evaluations based on your selected answers, and tests a difference only from the paired scores or counts you enter. For spatial, temporal or grouped data, validation design should reflect the intended prediction task; see Roberts et al. (2017), Ecography, 40, 913–929. doi:10.1111/ecog.02881.
Nadeau, C., Bengio, Y. (2003). Inference for the generalization error. Machine Learning 52, 239–281. doi:10.1023/A:1024068626366
Bouckaert, R.R., Frank, E. (2004). Evaluating the replicability of significance tests for comparing learning algorithms. In: Advances in Knowledge Discovery and Data Mining (PAKDD 2004), Lecture Notes in Computer Science 3056, 3–12. Springer.
McNemar, Q. (1947). Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika 12, 153–157. doi:10.1007/BF02295996
Dietterich, T.G. (1998). Approximate statistical tests for comparing supervised classification learning algorithms. Neural Computation 10, 1895–1923. doi:10.1162/089976698300017197
Lab method conversions
Results from different methods aren’t directly comparable. These are the common conversion factors, with the assumption behind each stated, so you can see what you’re accepting when you convert.
References
Walkley, A., Black, I.A. (1934). An examination of the Degtjareff method for determining soil organic matter, and a proposed modification of the chromic acid titration method. Soil Science 37, 29–38.
Pribyl, D.W. (2010). A critical review of the conventional SOC to SOM conversion factor. Geoderma 156, 75–83. doi:10.1016/j.geoderma.2010.02.003
Ahern, C.R., Baker, D.E., Aitken, R.L. (1995). Models for relating pH measurements in water and calcium chloride for a wide range of pH, soil types and depths. Plant and Soil 171, 47–52. doi:10.1007/BF00009563
Oxide and unit conversions are exact, from molar masses.
Coordinates and distances
A degree of longitude shrinks with latitude, so distances calculated directly on longitude and latitude are distorted, and “metres” computed from degrees are wrong. Convert a distance, or paste coordinates to check them.
Method
No paper needed: standard spherical geometry (haversine formula, mean Earth radius 6,371 km).
Equation check BETA
Paste the equations from a paper, one per line, in plain text, Word’s linear format or LaTeX. The check reads each one symbolically, flags anything that can’t be read or evaluated, tests whether rearranged equations really follow from each other, and, given a worked example, solves for the missing value so it can be compared with the paper. No AI: the same equations always give the same answer.
Method
Symbolic parsing and evaluation with math.js: de Jong, J. (2013–2024). math.js: an extensive math library for JavaScript and Node.js. https://mathjs.org. Equivalence is tested numerically at random points, a standard approach to symbolic identity testing (Schwartz, J.T. (1980). Fast probabilistic algorithms for verification of polynomial identities. Journal of the ACM 27, 701–717. doi:10.1145/322217.322225).