Classifications and when to use them

Classifications and when to use them

Soil texture classifications carve the sand-silt-clay simplex into different sets of named regions. There is no "best" system — each one encodes the agronomic / pedological priorities of the community that designed it. Picking the right one matters more than people usually admit: a "clay loam" in USDA is not the same physical material as a "clay loam" in Avery (England & Wales), and the soil-water curves you get from a PTF will reflect whichever system the regression was trained on.

Pedotri ships 14 built-in classifications. Here is what each one is actually good for, the regions it was designed against, and the gotchas you'll hit if you use it outside that intended scope.

Triangle (2-D) classifications

Key	Region	Use it when…	Watch out for…
USDA	USA + global default	You're publishing internationally, sharing data across labs, or unsure. The lingua franca.	Coarse near the sand-silt boundary; misses fine clay distinctions important in Mediterranean soils.
FAO	International — coarse	You need a quick coarse / medium / fine grouping (3 classes). Common in FAO yield modelling and global crop maps.	Too coarse for field-scale agronomy; collapses ~12 USDA classes into 3.
GEPPA	France	French soil-survey data, INRA / GIS Sol references. Modern refinement of Jamagne.	14 classes — finer than most international systems; "AL" / "ALo" boundaries may not round-trip cleanly with USDA.
JAMAGNE	France (legacy)	Reproducing pre-1995 French studies. Otherwise use GEPPA.	Superseded by GEPPA in current practice.
HYPRES	Europe (continental)	You're driving the Wösten (1999) PTF or any HYPRES-derived hydraulic-property database.	Only 5 classes; intentionally aligned with the Wösten parameter set, not for descriptive use.
KA5	Germany	German soil survey data, Bodenkundliche Kartieranleitung references.	31 subclasses — much finer than USDA. Many subclasses share a parent so think about whether you actually need the full split.
ISSS	International, classical	Older European literature; pre-1980 reference. Uses 0.02 mm sand-silt cutoff.	Cutoff differs from USDA — see Soil-conversions.
INTERNATIONAL	International (Leeper & Uren)	Australian / textbook references that pre-date Northcote.	Often confused with ISSS — verify which one your source uses.
AVERY	UK (England & Wales)	UK soil-survey data; current SSEW practice.	Slightly different boundary geometry from USDA in the loam region.
NORTHCOTE	Australia	Australian agronomy, especially heavy clays. Fine subdivision of the high-clay corner.	Less useful in sandy / loamy soils.
CHINA	China (GB/T 17296-2009)	Chinese national soil-survey data.	The 6-class split is closer to FAO's grouping than USDA's.
EMBRAPA	Brazil (SiBCS)	Brazilian agronomy / EMBRAPA datasets.	5 classes — coarse; mostly used as a grouping after measurement.
PTG	Poland	Polish soil-survey data (PTG 2008).	Six classes, regional vocabulary.

One-axis (1-D) classifications

Key	Region	Input	Use it when…
KACHINSKY	Russia & CIS	physical_clay — % particles < 0.01 mm	Reading Russian / Soviet-era soil literature (9-class scheme). Cannot be converted from sand/silt/clay alone — you need a sieve PSD. See Soil-conversions.
RUS2004	Russia (current)	physical_clay	Reproducing labels from the 2004 "Классификация и диагностика почв России" or the 2008 Полевой определитель почв России. Six classes; a strict coarsening of KACHINSKY (loose+cohesive sand → "песчаная"; light/medium/heavy clay → "глинистая"). Use this when you want the official Russian soil-survey "разновидности" label.

How to choose

1. Where is the data from? Use the local convention. USDA is not a safe default for French INRA data — it'll silently drop the GEPPA-specific distinctions your colleagues care about.
2. What are you going to do with it? If you're feeding a PTF, use the classification the PTF was trained against. HYPRES → Wösten. USDA → Saxton-Rawls.
3. What is the resolution of your input? Don't classify into 31 KA5 subclasses if your lab gave you sand/silt/clay rounded to the nearest 5 %. The classification reports false precision.

Quick decision tree

Got geographic data with a specific country/region?
├── Yes → use the regional classification (GEPPA, KA5, AVERY, …)
└── No → 
    ├── Need it for a hydraulic-property PTF?
    │   ├── Saxton-Rawls (USDA-trained) → USDA
    │   ├── Wösten / HYPRES (continental EU) → HYPRES
    │   └── Other → match the PTF's training set
    └── Just need a label?
        ├── Coarse summary (3 classes) → FAO
        └── Detailed (12 classes) → USDA

Classifying once against every scheme

Side-by-side multi-scheme reports — typical in agronomic web apps that want to surface USDA + FAO + the regional French / German label all at once — are a one-liner:

import pedotri

results = pedotri.classify_all(sand=27, clay=45, locale="fr")
{k: r.name for k, r in results.items()}
# {'USDA': 'argile', 'FAO': 'fine', 'GEPPA': 'argile lourde',
#  'KA5': 'Tl', 'ISSS': '...', ...}

classify_all calls classify(..., detailed=True) once per 2-axis classification (1-axis schemes like KACHINSKY are skipped — they take a different input). Pass schemes=["USDA", "FAO"] to restrict the set, or locale="fr" (etc.) for localised class names in one go.

Related

• Soil-conversions — moving (sand, silt, clay) between standards with different cutoffs
• Pedotransfer-functions — which classification each PTF expects
• Custom-classifications-in-practice — when none of the built-ins fit your data