Scoring Methodology

1. Overview

Growable Ground (growableground.com) evaluates plant suitability for a specific parcel by comparing measured site conditions against published plant requirements across 10 environmental constraints. Site conditions are derived from 20+ federal data products across 14 agencies (USDA SSURGO, NRCS, USGS 3DEP, NOAA NCEI, NREL NSRDB, PRISM, EPA FRS, FEMA NFHL, USGS NHD, and others), complemented by 621+ state and county parcel sources for boundary resolution. Plant requirements are compiled from peer-reviewed horticultural literature and USDA databases for 1,112 species.

Each plant receives a suitability score (0–100) reflecting the degree of alignment between the site and the plant's biological requirements. The scoring engine is deterministic: identical inputs always produce identical outputs. No machine learning, no stochastic elements, no AI inference. Every recommendation traces to measured data and published thresholds.

The engine also identifies which constraints are amendable (correctable through cultural practices like raised beds, soil amendments, or irrigation) and which are fixed (climate zone, sun exposure), providing actionable guidance alongside each recommendation.

2. Constraint Specification

The scoring engine evaluates 10 environmental constraints organized into two tiers based on biological severity.

Pass/Fail Gates

These constraints represent fundamental survival requirements. Plants that fail a gate constraint cannot survive at the site under standard growing conditions, regardless of cultural practices.

Hardiness zone evaluation uses a three-tier resolution: (1) annuals skip the zone gate entirely, evaluated instead by frost-free days; (2) perennials with USDA zone data are compared directly; (3) perennials with only killing temperature data use an approximated zone. Bed type modifiers (cold frame through heated greenhouse, earth-sheltered walipini, raised bed) can extend the effective growing zone.

Sun minimum evaluation uses seasonal sun matching: cool-season crops are evaluated against spring/fall sun hours, warm-season crops against peak summer sun hours. A site marginal for tomatoes may be excellent for lettuce.

Graduated Assessment Constraints

These constraints represent conditions that affect plant performance on a continuum. Deviation from optimal conditions results in graduated scoring penalties proportional to the degree of mismatch.

3. Constraint Importance Rationale

Constraints are ranked by biological importance following the Analytic Hierarchy Process (AHP) framework (Saaty, 1980), a structured method for multi-criteria decision analysis based on expert pairwise comparison.

Primary Constraints

Sun exposure receives the highest importance because it cannot be amended by any cultural practice short of major infrastructure. A shaded site cannot be amended into a productive food garden. Sunlight is the primary driver of photosynthetic capacity (Janick, 1972).

Hardiness zone, soil pH, and drainage share equal importance at the next tier. Hardiness determines winter survival for perennials. Soil pH controls nutrient availability through mineral solubility chemistry (Brady & Weil, 2008). Drainage determines root zone aeration.

Secondary Constraints

Frost-free days and growing degree days both measure growing season adequacy but are partially correlated with each other and with hardiness zone. Their lower importance reflects this redundancy.

Tertiary Constraints

Precipitation receives lower importance because it is largely addressable through irrigation (FAO Paper 56, Allen et al., 1998). Soil texture receives the lowest importance because its effects are partially captured by the drainage and pH constraints.

Sensitivity Validation

Weight sensitivity analysis confirms that top-10 plant recommendations are robust to significant perturbation of constraint importance parameters.

4. pH Assessment Calibration

The pH assessment is calibrated against the published nutrient availability curve in Brady & Weil (2008, Figure 16-4), which maps nutrient availability as a function of soil pH across the 4.0–8.5 range.

• Within ±0.5 pH units of optimal: Most nutrients remain available. Plants generally perform well.
• 0.5–1.0 pH units from optimal: Mild reduction in phosphorus, iron, and manganese availability begins.
• 1.0–1.5 pH units from optimal: Moderate multi-nutrient impact. Phosphorus fixation becomes significant.
• 1.5–2.0 pH units from optimal: Severe nutrient lockout. Most crop species show significant growth failure.
• >2.0 pH units from optimal: Extreme conditions incompatible with most cultivated plants.

Amendment pathway: Agricultural lime raises soil pH approximately 0.5 units per application cycle in loam soils (Ritchey et al., 2016). Elemental sulfur lowers pH at a comparable rate. The practical amendment window is approximately 1.5 pH units over 2–3 years.

5. Precipitation Assessment and Drought Tolerance

The precipitation constraint evaluates annual rainfall against each plant's published precipitation range, with a drought tolerance modifier based on species-specific water use efficiency.

FAO Crop Yield Response Framework

The base assessment is grounded in the FAO crop yield response factor (Ky) framework (Steduto et al., 2012). FAO Paper 66 publishes Ky values representing the relationship between relative yield decline and relative evapotranspiration deficit:

• High sensitivity (Ky ≈ 1.0–1.2): Pepper, tomato, bean — yield drops roughly proportional to water deficit
• Medium sensitivity (Ky ≈ 0.7–0.9): Corn, sunflower — moderate yield resilience
• Low sensitivity (Ky ≈ 0.4–0.6): Wheat, cotton — significant drought tolerance

Drought Tolerance Modifier

The plant database includes a drought tolerance field (High / Medium / Low) that maps to the FAO Ky framework. High drought tolerance corresponds to low Ky values — these species require substantially less penalty for precipitation deficit. This modifier applies only to precipitation deficit, not to excess precipitation.

6. Drainage Assessment and NRCS Alignment

The drainage constraint uses a bidirectional compatibility assessment aligned with USDA NRCS drainage class definitions (National Soil Survey Handbook, Part 618, Subpart B).

The seven NRCS drainage classes form an ordered continuum: Very poorly drained, Poorly drained, Somewhat poorly drained, Moderately well drained, Well drained, Somewhat excessively drained, Excessively drained.

Two secondary modifiers adjust the assessment: anaerobic tolerance softens penalties for too-wet conditions, and drought tolerance softens penalties for too-dry conditions.

Amendment pathway: Raised beds (8–18 inches) immediately improve drainage for waterlogged sites (NRCS Conservation Practice Code 436).

7. Multi-Year Trajectory Rates

The scoring engine projects site improvement over a 3-year horizon for amendable constraints:

• Soil pH: ~0.5 pH units per year with proper liming or sulfur application (Ritchey et al., 2016). Rate depends on buffer capacity, texture, and organic matter.
• Organic Matter: ~0.35 percentage points per year with consistent compost and cover cropping (Magdoff & van Es, 2021). Rate varies by climate and management intensity.
• Drainage: Immediate with raised bed installation (NRCS Code 436). Landscape-scale drainage requires professional engineering.

8. Chill Hours Computation Method

Many temperate fruit and nut trees require a minimum period of winter cold to break dormancy. The engine estimates annual chill hours using a sinusoidal daily temperature model from PRISM 30-year monthly normals, counting hours within 32–45°F (0–7.2°C) per the Utah Model (Richardson et al., 1974).

The sinusoidal approximation is documented in Luedeling et al. (2009) and Luedeling & Brown (2011).

Scoring Basis

Graduated fulfillment thresholds based on Weinberger (1950): full fulfillment (≥100%) receives no penalty; slight deficit (90–99%) may delay bud break; moderate deficit (60–90%) reduces yield; severe deficit (<60%) causes near-total dormancy failure.

Chill hours are a fixed constraint — they cannot be amended. The recommended path is selection of low-chill varieties. Greenhouse environments reduce effective chill accumulation (UC ANR).

9. Contamination Risk Assessment

Growable Ground evaluates environmental contamination risk from 9 federal data sources covering 1.8M sites nationally. Each source type uses a physically-informed dispersal model matched to the actual transport pathway of that contamination type — not a uniform-radius proximity model.

Data Sources by Pathway

Air Dispersal Model

Airborne contamination (TRI emissions, mining tailings dust, CAFO ammonia, pesticide spray drift) is modeled using an inverse-square decay function informed by the far-field behavior of Gaussian plume models. Wind direction frequency is derived from NOAA Integrated Surface Database (ISD) hourly observations at the nearest weather station, computed as a 16-sector wind rose. Fractional downwind exposure represents the percentage of time a parcel sits in the contaminated air pathway of a given facility.

Caveat: the wind rose is parcel-centered using the nearest ISD station (typically 10–50 km away). Local terrain effects — coastal sea breeze, valley channeling, urban heat island — are not modeled. Citation: EPA SCREEN3 Model User's Guide (EPA-454/B-95-004, 1995); NOAA ISD.

Per-source plume radii anchor to federal screening models: EPA RSEI (TRI air releases, scaled log-with-emission-lbs), BLM/OSMRE fugitive-dust guidance (mining), EPA AERMOD + Iowa State odor research (CAFO), and USDA-ARS AGDISP (pesticide spray drift). Distances reflect conservative-end screening thresholds for surfacing exposure-relevant proximity, not regulatory exposure boundaries.

Groundwater Directional Model

For contamination types that migrate via groundwater (UST petroleum plumes, Superfund groundwater contamination, nitrate, brownfield leachate), flow direction is estimated from surface topographic slope as a proxy for water table gradient, derived from the USGS Elevation Point Query Service (3DEP). Transport velocity is weighted by USDA SSURGO saturated hydraulic conductivity — higher-permeability soils allow faster and farther plume migration. Method: Darcy's law applied to a uniform-gradient approximation (EPA DRASTIC framework, Aller et al., 1987).

Surface Water Pathway

Contamination that travels via waterways (TRI water discharges, mining acid mine drainage, CAFO nutrient runoff) is traced along the NHDPlus Version 2 stream network. Concentration decay is estimated from drainage area ratios — as additional tributary inflow joins the stream, the contaminant plume dilutes proportionally.

Near-Road Exposure Model

Major roads (TIGER PRISECROADS classified S1100 interstate or S1200 US/state highway) are modeled as line sources for vehicle-related lead, PAH, and particulate deposition. Plume extent follows EPA Near-Road Air Quality guidance (EPA-454/F-09-002, 2010) and the Karner et al. (2010) meta-analysis: 200 m downwind (elevated PM/UFP zone) + 50 m upwind back-drift, scaled by road class (S1100 interstate ×1.5, S1200 US/state ×1.0 baseline, S1400 local ×0.6) as a traffic-volume proxy.

Caveat: MTFCC-based scaling is a rough surrogate for traffic volume; FHWA HPMS AADT data is the rigorous input and is on the roadmap. Vehicle-mix (diesel-truck percentage) is not yet incorporated. Citations: EPA Near-Road Air Quality (EPA-454/F-09-002, 2010); HEI Special Report 17 (2010); Karner, Eisinger & Niemeier, “Near-Roadway Air Quality: Synthesizing the Findings from Real-World Data,” Environmental Science & Technology 44 (2010).

PFAS Service-Area Model

PFAS exposure in drinking water is modeled as a water system membership risk rather than a point-proximity risk. If a parcel's ZIP code is served by a public water system with PFAS detections, the parcel is at risk regardless of distance from the treatment facility. Detection data is sourced from EPA's Fifth Unregulated Contaminant Monitoring Rule (UCMR 5, 86 FR 73131, December 2021). EPA's 2024 MCL final rule sets PFOA at 4.0 ppt and PFOS at 4.0 ppt (89 FR 32532). Private-well users should test independently.

Pesticide Spray Drift

Pesticide spray-drift exposure combines USDA CropScape Cropland Data Layer (CDL) satellite imagery at 30-meter resolution to identify agricultural crops within 2 km of a parcel, USGS NAWQA EPest county-level application data to map each detected crop to commonly-applied active ingredients, and the USDA AgDRIFT spray drift model (Teske et al., 2002) calibrated to the EPA Spray Drift Task Force empirical dataset.

For each active ingredient, the engine runs a 1,000-draw Monte Carlo simulation across application-rate uncertainty (coefficient of variation ≈ 30%), applicator method (ground boom vs. aerial vs. backpack), droplet size distribution (fine / medium / coarse), and wind speed from the nearest ISD station. Output is p10 / p50 / p90 deposition bands per chemical, rendered as five isopleth polygons on the contamination map. Atmospheric stability at spray time is classified using Pasquill-Gifford stability classes (A–F), derived from solar radiation, cloud cover, and wind speed — stable conditions (classes E–F, typically dawn/dusk) produce longer, narrower plumes than unstable conditions (classes A–B, midday), which drives the AgDRIFT deposition curve selection.

Full-physics pesticide modeling is deployed on production for the California Central Valley (approximately 35.0–40.0°N, −122.0 to −119.0°W), where California's mandatory Pesticide Use Reports provide per-acre application rates at statewide coverage — the richest pesticide dataset of any US state. The Central Valley deployment is the proving ground: full-physics coverage extends state by state as independent validation against EPA's AgDRIFT v2.x desktop tool completes for each region's NASS-derived application profile. Parcels outside the pilot region receive proximity-based exposure through the 9-source contamination assessment above.

Per-Bed Attenuation

When a parcel has contamination risk, the effective exposure depends on cultivation method. Raised beds with imported soil reduce ground-contamination exposure by approximately 60%; container gardens reduce it by approximately 90% (air pathways remaining). Citation: EPA "Reusing Potentially Contaminated Landscapes: Growing Gardens in Urban Soils” (EPA 542-F-10-011, 2011).

All contamination models are screening-level assessments. They are not a substitute for Phase I or Phase II Environmental Site Assessments. A minimum residual risk floor is maintained for remediated and closed sites — natural attenuation reduces, but does not eliminate, all long-term risk.

10. Known Limitations

• Sun Exposure: We cast the sun across terrain, buildings, and a leaf-on tree-canopy model. Canopy height is modelled from aerial imagery — validated to within about 3 m against NEON LiDAR — so shade reads most reliably over dense stands and least certain around a single open-grown tree, where a midday look confirms it.
• Plant Database Scope: Species-level ranges from USDA and peer-reviewed literature. Cultivar-specific requirements may vary significantly. Consult local Cooperative Extension for commercial decisions.
• Chill Hours Uncertainty: Year-to-year variation is ±150–300 hours. Microclimatic effects (cold air pooling, urban heat island) can shift values ±100–200 hours.
• Soil Data Resolution: SSURGO data at 1:24,000 scale represents dominant soil component, not point-specific measurements. Site-specific soil tests are recommended.
• Precipitation: Assumes natural rainfall only. Sites with irrigation may disregard deficit warnings.
• GDD Estimation: Microclimate effects can shift effective GDD by 200–400 units from grid value.
• Pesticide Validation: Independent validation of the AgDRIFT implementation against EPA's AgDRIFT v2.x desktop tool is in progress, region by region. Internal tests cross-check against the same Teske 2002 empirical dataset the model is derived from; external validation completes before full-physics extends to a new region. CDL crop identification has 85–95% accuracy depending on crop type and region; application rates are drawn from NAWQA county aggregates outside the California PUR-covered pilot area.
• Native Region: Informational only — does not affect suitability score.

11. Scoring Version History

12. Proprietary Implementation

This document describes the methodology used by Growable Ground's scoring engine in sufficient detail for scientific review and reproducibility assessment. The specific software implementation — including constraint weighting coefficients, scoring curve parameters, gate thresholds, persona modifier tables, and the integration architecture that composes federal spatial data into parcel-level intelligence — constitutes trade secret and proprietary information of Growable Ground LLC.

Patent notice: The Growable Ground platform, including its parcel-level agricultural intelligence engine, scoring methodology, and spatial computation systems, is protected by US Patent Application No. 64/013,215 (filed March 22, 2026) and US Patent Application No. 64/055,547 (filed May 2, 2026), and other pending intellectual property rights.

Inquiries regarding licensing, collaboration, or research use of the scoring methodology should be directed to info@growableground.com.

13. References

1. Allen, R.G., Pereira, L.S., Raes, D., and Smith, M. (1998). Crop Evapotranspiration. FAO Paper 56.
2. Aller, L. et al. (1987). DRASTIC: A Standardized System for Evaluating Ground Water Pollution Potential Using Hydrogeologic Settings. EPA/600/2-87/035.
3. Baker, N.T. and Stone, W.W. (2015). “Estimated annual agricultural pesticide use for counties of the conterminous United States, 2008–12.” USGS Data Series 907.
4. Brady, N.C. and Weil, R.R. (2008). The Nature and Properties of Soils. 14th ed. Prentice Hall.
5. Daly, C. et al. (2008). “Physiographically sensitive mapping of climatological temperature and precipitation.” Int. J. Climatol. 28: 2031–2064.
6. EPA (1995). SCREEN3 Model User's Guide. EPA-454/B-95-004.
7. EPA (2011). Reusing Potentially Contaminated Landscapes: Growing Gardens in Urban Soils. EPA 542-F-10-011.
8. Janick, J. (1972). Horticultural Science. 2nd ed. W.H. Freeman.
9. Luedeling, E. and Brown, P.H. (2011). “A global analysis of the comparability of winter chill models.” Int. J. Biometeorol. 55: 411–421.
10. Luedeling, E., Zhang, M., and Girvetz, E.H. (2009). “Climatic Changes Lead to Declining Winter Chill.” PLOS ONE 4(7): e6166.
11. Magdoff, F. and van Es, H. (2021). Building Soils for Better Crops. 3rd ed. SARE Handbook 10.
12. McMaster, G.S. and Wilhelm, W.W. (1997). “Growing degree-days: one equation, two interpretations.” Agric. For. Meteorol. 87(4): 291–300.
13. NOAA NCEI. Integrated Surface Database (ISD). National Centers for Environmental Information.
14. Richardson, E.A., Seeley, S.D., and Walker, D.R. (1974). “A model for estimating the completion of rest.” HortScience 9(4): 331–332.
15. Ritchey, E.L., McGrath, J.M., and Murdock, L.W. (2016). “Improving Subsoil Acidity.” UK Extension AGR-219.
16. Saaty, T.L. (1980). The Analytic Hierarchy Process. McGraw-Hill.
17. Steduto, P. et al. (2012). Crop Yield Response to Water. FAO Paper 66.
18. Teske, M.E. et al. (2002). “AgDRIFT^®: A model for estimating near-field spray drift from aerial applications.” Environmental Toxicology and Chemistry 21(3): 659–671.
19. Thom, H.C.S. (1952). “Seasonal degree-day statistics.” Mon. Weather Rev. 80(9): 143–147.
20. Tisdale, S.L., Nelson, W.L., and Beaton, J.D. (1985). Soil Fertility and Fertilizers. 4th ed. Macmillan.
21. USDA NASS. CropScape Cropland Data Layer (CDL). 30-meter resolution annual crop classification.
22. USDA NRCS (2022). National Soil Survey Handbook. Part 618.
23. Weinberger, J.H. (1950). “Chilling requirements of peach varieties.” Proc. Am. Soc. Hortic. Sci. 56: 122–128.