Why Your Soil Health Targets Might Be Unfair (and How to Fix Them)

A western Nebraska field can appear 4.6 percentage points behind in soil organic matter when it is graded against a humid eastern Nebraska reference. Compare that same field with an ecologically matched western reference, and the gap is 1.8 points. The soil did not change. The benchmark did.

That is why this question has stayed with me since 2020: improving soil health compared with what?

Working across Nebraska, I met farmers who were trying to do right by their soil but could not get a fair answer from a soil-health score. A grower could hand me a test and ask, “Is this good?” Yet a number by itself could not tell us whether the soil was close to its potential, whether management was helping, or whether the target belonged to a completely different soil and climate.

The problem was not a lack of recommended practices. It was the fairness of the measurement.

Two soils under similar management can behave very differently because their attainable ranges differ. Texture, mineralogy, precipitation, potential evapotranspiration, landscape position, and ecological history all shape the starting point and the ceiling before management enters the picture.

Our research group began treating soil health less like a universal score and more like a distance from an ecologically relevant reference. That idea grew across three peer-reviewed papers and one preprint: Soil Health Gap, Cropland Reference Ecological Units, ecological reference-site selection, and a provisional Two-Tier assessment framework.

Those last three figures come from a Nebraska preprint spanning two Major Land Resource Areas. They show useful dimensional compression within that dataset; they do not yet prove that four indicators predict yield, profitability, or ecosystem services across the country. That distinction matters, and I return to it below.

Part 1: Soil Health Gap — turning “improvement” into a measurable idea

We defined soil health improvement as the closing of a gap:

Soil Health Gap (SHG) = soil health under a reference condition − soil health under the managed (cropland) condition.

In equation form:

SHGₓ = (SH)ₙ − (SH)ₘ

where (SH)ₙ is the reference condition, (SH)ₘ is the managed condition, and x is the indicator of interest. This direct subtraction is intuitive for indicators where more generally means better, such as soil organic carbon within a relevant range. Indicators with the opposite direction, such as excessive bulk density, require direction-aware scoring rather than blindly applying the same sign.

If a gap shrinks across repeated, comparable measurements, that is evidence the soil is moving toward the reference condition. It is not automatically proof that one practice caused the change: season, sampling location, laboratory variation, and the absence of a matched control can all alter the result. The original Soil Health Gap paper was a peer-reviewed concept illustrated with four locations, not a replicated management trial.

The gap is still physically meaningful. A synthesis of cultivated soils estimated that conversion from native vegetation reduced soil carbon by roughly 20–40%, with a central estimate near 30% (Davidson & Ackerman, 1993). With the gap concept in hand, we began pairing managed cropland with reference sites and asking how much of that distance remained.

Then reality pushed back. Every soil is different, and its texture and environment set its potential.

Part 2: Why statewide comparisons can misdiagnose a field

Within a few sampling rounds, the flaw was obvious. Comparing a sandy, semi-arid soil in western Nebraska with a humid, fine-textured soil in eastern Nebraska is not an honest test of management — even when both soils are healthy in their own settings.

Each soil has its own ceiling. A sandy soil under semi-arid conditions may never reach the SOC or aggregate stability of a humid, fine-textured soil. Not because the farmer is failing — because the constraints are different. A soil's carbon-storage capacity is set first by its minerals and texture, then modulated by climate (Georgiou et al., 2022); on average, cropland soils sit near 31% of that capacity versus 46% for soils under natural vegetation — the deficit itself varies by soil and region.

So even though the Soil Health Gap was the right idea, it needed one more thing:

a way to compare soils only against other soils that share the same ecological constraints. That became the Cropland Reference Ecological Unit (CREU).

This is not a small statistical adjustment. In our 2024 analysis, the wrong reference made western conventional cropland look 4.6 SOM percentage points behind. A matched comparison put that gap at 1.8 points. Long-term manure management was 0.8 points from the matched reference. Benchmark choice changed the diagnosis and therefore could change the advice, incentive payment, or perception of farmer performance.

Part 3: CREU — making comparisons fairer by matching potential

CREU classifies land into units with similar soil-forming and climatic context, so that reference benchmarks actually mean something.

Conceptually:

CREU = f(MLRA, ES, Precipitation, Soil)

where MLRA is the Major Land Resource Area, ES is the Ecological Site, Precipitation is the rainfall range (which often needs careful grouping, especially in semi-arid systems), and Soil is the texture- or series-level grouping.

The point of a CREU is not the label. Comparing soils within the same CREU reduces ecological confounding, so observed differences are less dominated by geography. It does not eliminate confounding or turn an observational comparison into a causal experiment.

The peer-reviewed CREU paper proposed the method and demonstrated it in Nebraska MLRA 67A, where 45 CREUs were identified. That is a useful proof of method, but the framework still requires collaborative testing and calibration in other regions.

Now we had two of the pieces: a measurable definition of “improvement” (a shrinking gap), and a fairness unit that controls for ecological potential.

But the hardest operational question was still open: where do the reference sites actually come from?

Part 4: Reference sites — the fence-line problem (and why it matters)

Everyone reaches for fence lines, and I understand why — they look undisturbed, they’re easy, and they’re right there. But fence lines are often poor references. They can be shaped by runoff, dust and nutrient drift, compaction, invasion, and a long, invisible disturbance history; in windy regions, aeolian processes rework them outright.

More fundamentally, a fence line doesn’t necessarily represent the productive ecological potential of that land.

So we asked a different question: how do you identify a reference condition that is ecological, defensible, and repeatable?

This is where State-and-Transition Models (STMs) became essential. Instead of guessing, we used STMs and Ecological Site Descriptions to identify the reference plant community and the expected functional state for a given ecological context.

Reference selection then becomes a method, not a convenience: match soil, ecological site, and vegetation reference state, then sample it systematically alongside a paired managed site from the same CREU.

Our measured comparison showed why this matters. One disturbed fence-line reference had 2.7% SOM — the same as nearby long-term-manured cropland — while the ecological reference was 0.7 points higher. A convenient reference could therefore make an altered site look as though it had already reached its potential.

That gave us the third piece: a reference isn’t a location you settle for. It’s an ecological state you define.

So far the system looked like this — define improvement as closing a gap (SHG), define the comparison unit by potential (CREU), and define the reference ecologically (STM/ESD).

Then we hit the fourth problem — the one farmers and researchers both feel immediately.

Part 5: What do we measure, realistically?

Soil health has too many indicators to measure everything, everywhere. Ask farmers or programs to track 25–40 indicators on repeat, and adoption collapses — so do budgets and time. We needed a minimal set: small enough to be operational, strong enough to explain most of the variation in how a soil functions. We weren't alone in wanting this — evaluating more than 30 indicators across 124 North American sites, the Soil Health Institute converged on just three core measurements (Bagnall et al., 2023).

That led to a provisional Two-Tier framework, currently available as a preprint rather than a peer-reviewed final paper.

Tier 1 is a candidate Minimum Data Set intended to be affordable and scalable. Tier 2 interprets those measurements against CREU-based reference distributions. A fungi-to-bacteria ratio can be added as an optional Tier 1+ biological measure when budget and purpose justify it.

Across 30 indicators measured at 16 native reference sites in two Nebraska MLRAs, the candidate Tier 1 set converged on organic matter, pH, nitrate-N, and bulk density. Re-running principal component analysis on those four variables explained 88.6% of the variance in the first three components of the reduced dataset.

Tier 2 then draws provisional interpretation ranges from reference distributions, usually the 25th–75th percentile for each available MLRA × texture context. In the present dataset, precipitation is nested within MLRA rather than independently validated across a broad climate gradient.

This matters because “good organic matter” is not a universal number.

In our published reference-site comparison, average organic matter was about 1.5% in one sandy, semi-arid context and 6.2% in a humid clay-loam context. Both represented healthy reference conditions for their own ecological potential.

Same word. Different number. Same ecological truth.

A quick example: the carbon gap becomes a decision-ready number

Take one dataset from the semi-arid High Plains. SOC ran 4.4% in native grassland, 2.2% under no-till, 1.8% under conventional tillage, and 0.7% in exposed subsoil.

Turn those into Soil Health Gaps for carbon and the story sharpens: SHG_C (no-till) ≈ 44 − 22 = 22 g C kg⁻¹; SHG_C (conventional till) ≈ 44 − 18 = 26 g C kg⁻¹; SHG_C (exposed subsoil) ≈ 44 − 7 = 37 g C kg⁻¹.

Now the farmer doesn’t just hear “your SOC is low.” They hear: this is the size of the gap relative to your soil’s potential — and that gap can be tracked, up or down, as management changes. That’s a measurable story of degradation and recovery.

What this means for farmers

For a farmer, the practical shift is to stop treating a generic threshold as a verdict. Instead, ask four questions.

• What CREU am I in? — what are my soil and climate constraints?
• What’s my reference-informed potential range? — not a national “ideal,” but my local ecological baseline.
• How big is my gap right now? — SHG makes “improvement” quantitative.
• Is my management shrinking that gap over time? — that’s the real definition of soil health improvement.

A native reference is a compass, not a demand that every field be restored to native vegetation. Full restoration may be biologically or economically unrealistic in working cropland. A better decision pathway uses hierarchical attainable targets: first compare with similar farms using feasible practices, then evaluate whether stacking practices can move the field closer to its ecological potential.

Success becomes measurable progress toward a fair reference while maintaining the farm functions that matter — yield stability, nutrient-use efficiency, water movement, erosion control, profitability, and risk. Closing a laboratory gap without improving a relevant outcome is not enough.

What this means for researchers and programs

If you’re a researcher, consultant, or program manager, this framework buys you defensible interpretation:

• SHG gives you a metric that is explicitly baseline-referenced.
• CREU reduces, but does not eliminate, confounding from agroecological variability.
• STM/ESD-based reference selection avoids the fence-line trap by grounding the reference in ecological state.
• The provisional two-tier design offers a testable route to reduce indicator overload, but still needs outcome and geographic validation.

This matters for incentives and carbon markets. If thresholds are not ecologically matched, programs can reward geography or inherited soil stocks rather than the change a farmer creates. Baseline design determines additionality: whether a payment funds new improvement or pays for what would already have existed (Kannegieter & Medlock, 2026). A forest-offset study estimated 29% over-crediting under loose baselines (Badgley et al., 2022); that is an analogy, not direct validation of CREU for agricultural carbon markets, but it shows why benchmark design cannot be a footnote.

What is established — and what is still being tested

This sequence is the part I am proudest of: each paper exposed a limitation in the previous step, then tried to solve it. But it should be read as a developing framework, not a finished national standard.

Where this needs to go next

This is where Soil Health Exchange becomes more than a place to talk. Today, CREU is still conceptual in many places. The next step is to make it operational.

• Digitize CREU — a map layer that identifies the CREU for any parcel from geospatial inputs (MLRA + a precipitation zone based on potential evapotranspiration + soil texture/series).
• Build reference distributions for each CREU — the reference-based thresholds that drive Tier 2 interpretation.
• Validate the Tier 1 indicators in the real world — paired, longitudinal datasets that show the MDS tracks meaningful change across management transitions.
• Make it farmer-usable — a grower enters location, soil type/texture, and a few lab values (OM, pH, NO₃-N, BD), and gets back:
• “This is your CREU.”
• “These are your reference-informed thresholds.”
• “This is your soil health gap.”
• “And here’s what it implies — the management pathways that typically close that gap in your context.”
• Connect scores to consequences — test whether smaller gaps predict yield stability, nutrient-use efficiency, greenhouse-gas outcomes, water quality, and farm profitability.

The broader field is moving in the same direction. Production-environment benchmarks in New York stratify results by texture and cropping system, while the national SEMWISE model adjusts indicator expectations for clay and climate. These approaches are complementary: production data can define what is currently attainable; ecological references can show the longer-term potential and the distance still remaining.

That’s the goal: simple measurement, careful interpretation, and ecological fairness. Soil health should not be a universal scorecard. It can become a consistent method for asking a fairer question.

Explore the CREU Map to see how ecological context changes the benchmark, or start a soil health assessment and bring your own field into the conversation.

References

The framework behind this article

• Maharjan, B., Das, S., & Acharya, B. S. (2020). Soil Health Gap: A concept to establish a benchmark for soil health management. Global Ecology and Conservation, 23, e01116. doi.org/10.1016/j.gecco.2020.e01116
• Das, S., & Maharjan, B. (2022). Cropland Reference Ecological Unit: A land classification unit for comparative soil studies. Ecological Indicators, 144, 109468. doi.org/10.1016/j.ecolind.2022.109468
• Das, S., Hird, A., Maharjan, B., Stephenson, M., & Kariyawasam, L. (2024). Reference site selection based on state-and-transition models for soil health gap evaluation within cropland reference ecological units. Soil Security, 16, 100142. doi.org/10.1016/j.soisec.2024.100142
• Das, S., Kariyawasam Hetti Gamage, L., Stephenson, M. B., Acharya, U., Daigh, A., & Maharjan, B. (2025). A Two-Tier Framework for Soil Health Assessment: Minimum Indicator Sets and Ecologically Dynamic Thresholds. SSRN preprint (not peer reviewed). doi.org/10.2139/ssrn.5437463

Supporting literature

• Davidson, E. A., & Ackerman, I. L. (1993). Changes in soil carbon inventories following cultivation of previously untilled soils. Biogeochemistry, 20(3), 161–193. doi.org/10.1007/BF00000786
• Georgiou, K., et al. (2022). Global stocks and capacity of mineral-associated soil organic carbon. Nature Communications, 13, 3797. doi.org/10.1038/s41467-022-31540-9
• Bagnall, D. K., et al. (2023). A minimum suite of soil health indicators for North American agriculture. Soil Security, 10, 100084. doi.org/10.1016/j.soisec.2023.100084
• Amsili, J. P., van Es, H. M., Aller, D. M., & Schindelbeck, R. R. (2023). Empirical approach for developing production environment soil health benchmarks. Geoderma Regional, 34, e00672. doi.org/10.1016/j.geodrs.2023.e00672
• Deel, H. L., Moore, J. M., & Manter, D. K. (2024). SEMWISE: A national soil health scoring framework for agricultural systems. Applied Soil Ecology, 195, 105273. doi.org/10.1016/j.apsoil.2024.105273
• Cassman, N. A., et al. (2025). A purpose-led framework for soil health assessment. Soil Science Society of America Journal. doi.org/10.1002/saj2.70144
• Kannegieter, B., & Medlock, K. B. III (2026). Additionality constrains investment in carbon sequestration. npj Sustainable Agriculture, 4, 40. doi.org/10.1038/s44264-026-00155-8
• Badgley, G., et al. (2022). Systematic over-crediting in California's forest carbon offsets program. Global Change Biology, 28(4), 1433–1445. doi.org/10.1111/gcb.15943