Introduction

Part I: Soil 101 (a brief overview of pedology)

Part II: Do Central Texas soils predict their plant distributions?

  • Central Texas: 42 counties, ~25 million acres, four physiographic regions
  • 395 soil series and their official survey data (SSURGO)
  • 141 NRCS ecological sites: what should grow there
  • ~785,000 iNaturalist observations: what we actually see growing there

Part III: Live demo (play along at home)

What is soil?

  • Not just dirt: Soil is a living, layered body that forms over millennia
  • A loamy topsoil is roughly:
    • ~45% mineral particles (sand, silt, clay)
    • ~25% water and ~25% air, held in the pore spaces
    • ~5% organic matter: roots, fungi, microbes, decaying plants
  • Half of a healthy soil is empty space (where roots, water, and life go)
  • (Just idealized numbers: Real soils vary a lot!)

Soil formation

Five factors: CLimate, Organisms, Relief, Parent material, Time

  • Parent material does the most work in central Texas:
    • Ancient Precambrian granite (Llano Uplift), ~1 billion years old
    • Old, hard Cretaceous limestone (Edwards Plateau), ~ 100 mya
    • Slightly younger, softer Cretaceous chalks (Blackland), ~ 75 mya
    • Younger Eocene sands (Post Oak Savannah), ~50 mya
    • New Quaternary alluvium (river bottoms), < 2 mya
  • Relief (Geography): Thin, stony soils on slopes (hill country), deep soils on flats (East)
  • Climate: Rainfall increases going east, so leaching/carbonate change with it

Soil horizons

A soil profile is a stack of layers (horizons), each formed by a different process

Horizon What it is
O Leaf litter and organic debris
A Topsoil: minerals darkened by organic matter, where most roots are
E Leached zone: clay, iron and organic matter washed out
B Subsoil: where the washed-down clay, iron or carbonate accumulates
C Weathered parent material, little changed by soil formation
R Bedrock

Not every soil has every horizon! Some just have one or two aside from the bedrock.

Soil horizons

  • Eckrant: a dark A horizon directly on limestone bedrock less than 10” down
  • Houston Black: 5+ ft of deep clay, churned by shrinking and swelling
  • Padina: 3+ ft of sand above a clayey B horizon (3+ ft of clay below)

Soil composition

  • Sand / silt / clay
  • Sand: ~thickness of printer paper
  • Silt: ~thickness of spider silk
  • Clay: smaller than wildfire smoke particles, bacteria
  • Texture matters: sand drains fast, clay holds water (and nutrients)
  • Sand particles are roughly 1000x the size of clay particles
    • sand particle : basketball :: clay particle : grain of salt
  • Some clays shrink and swell: crack when dry, heave when wet

Where central Texas soils fall

Soil color

  • Dark brown to black: organic matter (grassland Mollisols, Blackland clays)
  • Red: iron oxides in well-drained soils (granite of Llano Uplift, “redland” sites)
  • Gray or mottled: waterlogging, where iron has been chemically reduced
  • White: calcium carbonate (caliche, chalky subsoils)

Beyond texture: chemistry & physics

Chemistry

  • pH: acid sands out east (~5.7) to alkaline limestone (~7.9)
  • Calcium carbonate (CaCO₃) “lime”
    • defining characteristic Hill Country soils
    • primary cause of iron chlorosis in out-of-range species
  • Nutrient holding (CEC): clay and organic matter hold nutrients; sand barely does
  • Organic matter: usually 1–2% of the root zone, 5%+ in some
  • Salt and sodium: rare here, more common down south

Physics

  • Shrink-swell: separates Blackland clays from everything else
  • Rooting depth: under 10 inches on the stony limestone; no restriction within 6 feet on the deep clays and sands
  • Water capacity: plant-available water in the top meter: clay > sand > thin rocky soils
  • Drainage: water moves through sandy soils ~500× faster than through heavy clay
  • Bulk density: how tightly packed; denser soil is harder for roots to push through
  • K factor: soil erodibility index
    • ease with which soil is detached by splash during rainfall or by surface flow

Shrink-swell: the Blackland’s signature

  • Houston Black is the official state soil of Texas
  • Its expansive clays swell when wet and shrink when dry
    • Summer cracks open deep enough to lose a boot in
    • Surface soil falls into the cracks, so the soil is self-plowing
    • Gilgai: a hummocky micro-relief of mounds and hollows
  • Houston Black’s shrink-swell (linear extensibility) is 12.1, the highest of the nine focal series
  • Hard on foundations, roads and pipes; very productive for crops

Soil taxonomy

  • There are 12 soil orders (broadest level of classification)
  • Texas features 9 of them
  • Central Texas has 6:
    • Alfisols
    • Entisols
    • Inceptisols
    • Mollisols
    • Vertisols
    • Ultisols

Order → suborder → great group → subgroup → family → series

Soil orders

Order In a phrase Central Texas example
Vertisols Shrink-swell clays that crack and churn Houston Black (Blackland Prairie)
Mollisols Dark, fertile, grassland topsoil Eckrant, Rumple (Hill Country)
Alfisols Clay has washed down into a subsoil layer Crockett, Edge, Padina (Post Oak)
Inceptisols Young soils, only weakly developed Brackett (limestone hills)
Entisols Barely any development: fresh alluvium, sand, rock 33 series in the region
Ultisols Old, weathered, low-nutrient acidic forest soils Cuthbert (Pineywoods)

Reading a soil’s full name

Houston Black: Fine, smectitic, thermic Udic Haplusterts

  • -erts: a Vertisol
  • -ust-: ustic moisture, dry for part of the year
  • Hapl-: the simplest (“haplo”) kind
  • Udic: on the moister side of ustic
  • Fine: clayey particle size
  • Smectitic: dominated by smectite, the shrink-swell clay mineral
  • Thermic: warm soil temperature regime

Vertisols are unique

  • Only about 2.4% of the world’s ice-free land
  • India, Australia, and Sudan hold the most
    • India’s Deccan Plateau: the “black cotton soils”
  • In the US, about 2% of land, found mostly in Texas
  • They need a particular recipe:
    • parent material that weathers to smectite clay (Cretaceous chalk/marl here; basalt in India)
    • climate that historically swings between wet and dry

Soil series

Brackett–Lampasas (Lampasas County): a block diagram shows where soil series sit relative to each other, the landscape, and often often the rock beneath them

  • A series is a named, specific soil with a documented profile (the Official Series Description)
  • A map unit is what gets drawn on the map: a series plus slope and texture
    • “Houston Black clay, 1 to 3 percent slopes”
  • Central Texas has 395 series in 2,540 map units and 264,796 polygons

How soils get mapped

  • Soil scientists walk the land, take core samples, and describe profiles
  • They draw boundaries on aerial photos where the landscape changes
  • Results were published by county as printed soil surveys, now digitized as SSURGO
  • A map unit is rarely pure: it usually contains other soils too
    • Everything in this talk uses the dominant soil of each map unit

The study area

Central Texas (unique!)

  • 4+ regions meet in one area:
    • Post Oak Savannah: acid sands and claypans
    • Llano Uplift: granite, an acidic island in a carbonate region
    • Blackland Prairie: deep shrink-swell clays on chalk and marl
    • Edwards Plateau: thin, stony soils on limestone
  • The Balcones Escarpment is the limestone–Blackland boundary

Four landscapes

Edwards Plateau

Blackland Prairie

Llano Uplift

Post Oak Savannah

Regions

Region (example counties) pH CaCO₃ % Shrink-swell Root depth % restricted
Post Oak Savannah (Leon, Freestone) 5.7–6.1 0.1–0.5 3.0–3.8 101–125 cm 2–14%
Llano Uplift (Llano, Mason) 6.7–6.9 2–4 2.6–2.8 65–68 cm 81–88%
Blackland Prairie (Falls, Williamson, Travis) 7.3–7.9 7–26 6.8–7.7 72–133 cm 32–47%
Edwards Plateau (Kerr, Bandera, Comal) 7.8–7.9 23–32 4.5–5.8 44–53 cm 57–66%


  • Across counties, pH spans 2.27 units and carbonate 33.5 points

  • Llano Uplift: acidic island in a carbonate region

  • “% restricted” is share of soils with a restrictive layer (e.g., rock) in the root zone

    • root zone refers top 100 cm (~40 inches)

Nine soils that span the range

Series Order Area (ha) pH CaCO₃ % Clay % Sand % Depth (cm)
Brackett Inceptisols 550,189 7.9 61.3 28 34 40
Eckrant Mollisols 544,590 7.5 5.6 46 17 28
Houston Black Vertisols 296,206 7.9 24.0 54 20 200
Crockett Alfisols 237,742 6.9 0.3 39 40 139
Edge Alfisols 210,127 5.7 0.0 35 39 198
Padina Alfisols 150,448 6.5 0.0 6 85 200
Rumple Mollisols 85,940 7.2 0.0 53 23 67
Sunev Mollisols 64,535 8.2 45.6 29 31 200
Cost Alfisols 2,003 8.5 0.0 30 39 200
  • Chosen by cluster analysis on 16 features; each is largest series in each cluster

Taxonomy alone can’t tell them apart

Crockett vs Edge

  • Both Alfisols, both claypans, both ~39% sand, both deep
  • Split on pH 6.9 vs 5.7
  • Edge carries the acid-sand flora: Baptisia bracteata, eastern redcedar

Brackett vs Sunev

  • Both high-carbonate limestone soils
  • Split on depth: 40 cm vs 200 cm
  • Same chemistry, opposite water-holding capacity

All soil features used in the analysis

  • 16 features measured for all soil series (dominant)
  • All are root-zone (0–100 cm) values unless noted:
  1. pH
  2. CaCO₃ (lime)
  3. EC (salinity)
  4. SAR (sodium)
  5. CEC (nutrient holding)
  6. Organic matter
  7. Shrink-swell (LEP)
  8. Ksat (drainage)
  1. Bulk density
  2. Available water capacity
  3. Clay
  4. Sand
  5. Rock fragments
  6. Rooting depth
  7. K factor (topsoil erodibility)
  8. Slope

Principal Component Analysis (PCA)

  • Every series is described by 16 measurements: pH, CaCO₃, clay, sand, depth, rock… so each soil is a point in 16 dimensions
  • PCA rotates that cloud to find the directions it varies along most
  • Each new axis (a principal component) is a blend of the 16 measurements
  • We reduce the set of 16 to a few PCs that capture most of the variation


  • Example plot to help build intuition (right):
    • Just 3 of the 16 dimensions
    • 3D can be pretty easily collapsed into 2D
    • Most of the cloud’s variation is along two dimensions
    • PC1 spans widest extent of cloud, PC2 spans next widest

PC1 vs PC2: texture and depth

Other PCs: PC3 (12%) salinity/sodium, mainly one soil (Cost) · PC4 (9%) lime/pH · PC5 (7%) slope/erodibility vs. depth

PCA recovers taxonomy

Central Texas in One County?

Central Texas in One County?

From the 1983 soil survey

The round rock: Brushy Creek carved this limestone pedestal, which gave the town its name. Oakalla soils line the creek.

Wagon ruts worn into the limestone bed of Brushy Creek in pioneer days: the rock that underlies the Eckrant soils.

The cover: preparing a seedbed on Branyon clay, a typical Blackland landscape.

West: the limestone

Eckrant–Georgetown: thin stony soils and rock outcrop over hard fractured limestone

Denton–Eckrant–Doss and Oakalla–Sunev: Eckrant on the ledges, deeper Denton and Doss on slopes, Sunev and Oakalla in the valleys

Middle: the Blackland

Austin–Houston Black–Castephen: soils on marine chalk and marl, Houston Black in the broad low areas

Branyon–Houston Black–Burleson: deep shrink-swell clays on old clayey alluvium and marine clays

East: claypans, sands and floodplains

Wilson–Crockett–Behring: claypan soils on noncalcareous marine clays

Axtell–Rosanky–Rader: sandy Post Oak soils on marine sands and clays

East: claypans, sands and floodplains

Tinn: clayey floodplain soils along the county’s rivers and creeks

We have the soils, now for the plants!

Three views of the vegetation

  • USDA’s Natural Resources Conservation Service (NRCS)

    • Ecosystem Dynamics Interpretation Tool (EDIT) descriptions
    • Basically, expert reference plant communities, 141 sites, 715 species
    • Defined per ecological site, which spans several series
    • Overindexes on native plants good for forage
  • NRCS Official Series Descriptions: a short list of dominant plants per series

  • iNaturalist: 785,019 research-grade observations, 2,659 species

    • De-duplicated to one record per species per 1 km cell: the top 1% of cells held 37% of records

De-duplication matters: without it, Eckrant’s top species were Ashe juniper and mountain laurel from repeat photography of one greenbelt. With it they are stonecrop and lace hedgehog cactus.

How the data fit together

  1. Soil map and data: NRCS soil survey (SSURGO) for 42 counties, 264,796 polygons; each takes the properties of its dominant soil series, averaged over the root zone (0–100 cm)
  2. Expected plants: each polygon’s ecological site links it to its NRCS reference plant community
  3. Observed plants: 785,019 iNaturalist research-grade plant records, each placed on the soil polygon it falls inside, so every record gets a soil series, an ecological site and that soil’s properties
  4. One record per species per 1 km cell so a heavily photographed spot counts once

Then ask both ways: for each species, what soil is it found on? For each soil, which species stand out?

More about this in the app demo later

Does it recover known ecology?

Blind test: 13 species with published soil affinities

Species CaCO₃ % Sand % Depth cm
Calcicoles (lime-lovers)
blackfoot daisy 36.6 26 66 ✓
Lindheimer muhly 34.0 28 89 ✓
Texas red oak 32.1 26 75 ✓
twistleaf yucca 30.3 25 70 ✓
Ashe juniper 26.6 25 81 ✓
escarpment live oak 23.0 26 89 ✓
Calcifuges (lime-haters)
blackjack oak 4.3 47 146 ✓
black hickory 0.2 68 169 ✓
loblolly pine 0.9 57 175 ✓
farkleberry 0.4 50 172 ✓
old plainsman 2.3 60 178 ✓
Bottomland
bald cypress 25.2 31 152 ✓
sycamore 27.6 32 145 ✗

Before trusting this, check it against what botanists already know

  • Pass rules, fixed in advance:
    • lime-lovers above the median CaCO₃
    • lime-avoiders above the median sand
    • bottomland species deeper than 150 cm
  • 12 pass, 1 threshold artifact
  • Lime-loving species on 23–37% CaCO₃; lime-avoiding species on 0.2–4.3%
  • Bald cypress and American sycamore found in deeper floodplain soils

Does soil shape whole plant communities?


Method What it does Result
Mantel r correlation between soil/plant distance matrices 0.30, p = 0.001
Procrustes do soil/plant similarity maps have the same shape 0.63, p = 0.001
db-RDA how much variation in plant distribution due to soils ~31%


  • Yes! Soils and floras are arranged in a similar way

  • If you take any two series, the more their soil measurements differ (soil distance), the more their observed species differences tend to differ (floristic distance)

  • Note: 69% is unexplained (land use, dispersal, climate, etc)

So soil matters, but which part matters most?

db-RDA predicts variation in plant composition from 16 soil features

We can ask, which features account for the greatest variation in composition


Soil property F p
pH 7.05 0.005
bulk density 2.25 0.020
slope 2.18 0.025
CaCO₃ 1.95 0.030
clay 1.96 0.060
11 others ≤ 1.5 n.s.
Unique share of variation
Chemistry 8.2%
Texture 3.6%
Depth / slope / shrink-swell 2.3%
Shared 13.3%
Unexplained 68.8%


- pH’s effect is 3× the next property’s. Chemistry explains more than twice as much as texture. - (Bulk density higher if soil compacted, lacks organic matter/aggregation; inverse of porosity)

Soils decide which species…

What about how many?

  • Raw richness (# of unique species per series)
    • Raw species counts mostly measure how often a soil gets visited
    • So plants on soils in Travis and Williamson County over-represented
  • Rarefied richness fixes this; varies only 18% between soil groups
    • Sandy soils highest (624), Blackland lowest (527)
  • Raw counts would have said Eckrant was the most diverse…
    • But Eckrant has ~56,000 species-cells (way more land) vs Padina, ~3,800
    • After downsampling/standardization, it’s middle of the pack

Scale matters

So far we have focused on big-picture, plant community findings…

Zooming in on individual species changes the picture

  • pH is the top variable for explaining which mix of plants grows on a soil
    • mainly because pH differs a lot from one region to another
  • But it’s actually the weakest for which specific species does (!)
    • Within a county, pH varies little and few species show a pH preference
  • Conversely, a soil property can be irrelevant to plant communities, yet still decide where one species grows…

Example: Texas bluebell, Vertisol specialist

Eustoma russellianum: 238 records, 27 counties, 170 observers

Shrink-swell, a property that barely matters for communities, decides where this one species grows

Soil property Compared with its own county
Shrink-swell +3.2 (the strongest of any axis)
Clay +6.4 points
Rooting depth +34 cm
CaCO₃ −3.2 points (avoids it)

  • 4.5× more common on Blackland than expected: the highest of any species
  • Shrink-swell is the weakest of all 16 properties for communities…
    • yet it is decisive for this one species
  • Zero records in the NRCS reference communities

Oaks sort by carbonate

Quercus provides range of species preferences within one genus, where close relatives can be compared

Oak CaCO₃ % Sand %
Texas red oak (Q. buckleyi) 32.1 26
chinkapin oak (Q. muehlenbergii) 27.8 23
bur oak (Q. macrocarpa) 24.2 24
escarpment live oak (Q. fusiformis) 23.0 26
post oak (Q. stellata) 6.6 39
blackjack oak (Q. marilandica) 4.3 47
water oak (Q. nigra) 0.9 50

  • A 35-fold carbonate range within one genus, and it holds within counties
    • Texas red oak beats post oak in 11 of 12 counties
    • So real niche separation, not geography

Sitting on limestone isn’t the same as choosing it

Species Raw CaCO₃ vs. its own county
Texas red oak 32.1% +7.3 (selects carbonate)
escarpment live oak 23.0% −1.0 (just abundant in limestone country)
post oak 6.6% −5.9 (actively avoids it)
common sunflower 20.6% −0.1 (the pure generalist)
  • Subtracting the mean of all obs in the same county separates “prefers” from “happens to be common where it occurs”
  • The large Q. fusiformis below is actually in San Saba county on sandier soil

Oaks split on lime, sunflowers on something else

Species Sand % vs. its own county CaCO₃ vs. county
Helianthus praecox 71 +37.8 −12.8
H. debilis (beach / cucumberleaf) 73 +25.2 −4.0
H. argophyllus (silverleaf) 70 +23.0 −5.8
H. maximiliani (Maximilian) 26 −2.4 +1.9
H. annuus (common) 24 −3.0 −0.1

  • Three sand specialists, two generalists: a 48-point gap in sand
  • H. debilis is sandier than common sunflower in every county where both grow (6 of 6)
  • Test more than one soil property: a pH screen would call this genus null

One genus, two soils: the junipers

  • Ashe juniper: the limestone indicator (Brackett, Eckrant)
  • Eastern redcedar: distinctive of Edge’s acid sand (5.3×)
County Ashe juniper CaCO₃ % Eastern redcedar CaCO₃ %
Travis 33.1 19.1
McLennan 46.8 20.8
Coryell 30.8 6.9
Bell 27.6 18.5
Williamson 15.4 24.1
  • Survives the within-county check: 4 of 5 counties
  • Wilco flip is an artifact of where the county line falls

Flip the question: start from soil, ask: which plants?

  • Lift = (share of plant records for each series) / (share across all 42 counties)

Padina (85% sand)

  • Reference: Texas bullnettle
  • Observed:
    • Palafoxia hookeriana, 52× more likely than elsewhere in Central Texas
    • Helianthus debilis, 39×
    • Hymenopappus artemisiifolius (old plainsman), 27×
  • Only 9.5% introduced species, among the most intact floras in the region

Brackett (61% CaCO₃)

  • Reference: tall grama, seep muhly, Texas red oak
  • Observed:
    • Melampodium leucanthum (blackfoot daisy)
    • Scutellaria wrightii (Wright’s skullcap)
    • Stillingia texana, Yucca rupicola

Eckrant: the data found the rock

  • Sometimes the plants reveal something about the soil that its label doesn’t
  • Eckrant: a shallow, stony Mollisol over Hill Country limestone
  • Reference plant list: agarita, cane bluestem, stretchberry (shrubland)
  • iNaturalist
    • raw abundance: Ashe juniper and mountain laurel
    • 1 record per species per 1 km cell: yellow stonecrop and lace hedgehog cactus
  • A trap: both read as carbonate “avoiders” (−12.9 and −13.8 points)
  • …but +4.1 and +3.5 on rock fragments
  • Thin soil over limestone gets leached of carbonate: a limestone glade

Ways of measuring species “distinctiveness”

Metric Core Question Needs Predefined Groups? Formula Intuition
TF-IDF How important is this feature in one item compared to all items? No Local frequency × global rarity
Z-Score How many standard deviations is a value from the dataset average? No (or compares group mean vs. overall) (Value − Mean) / Standard Deviation
IndVal How exclusive and consistent is this feature within a specific group? Yes Specificity (group concentration) × Fidelity (group presence %)
Lift How much more likely do two things occur together than by random chance? Yes (target event vs. baseline) Observed co-occurrence / Expected independent occurrence

Expected vs. actual

  • Only 46.5% of observations are of a species found in any NRCS reference community
  • 84% of observed species are absent from every reference list
Soil Reference says Observations say
Brackett, Eckrant (limestone) tall grama, seep muhly Ashe juniper
Houston Black (Blackland) vine mesquite, Virginia wildrye honey mesquite

Eastern gamagrass: a century later

  • Reference tables give it up to 3,275 lb/acre on Blackland
  • Observed: concentrated in bottomlands (2.2×), depleted on Blackland (0.83×)
On Blackland Enriched Depleted
Grazing increasers buffalograss, Texas wintergrass, bermudagrass
Grazing decreasers gamagrass, switchgrass, sideoats, little bluestem

Cultivation alone would suppress all of them together; this is a grazing-response signal.

Deep fertile soils host more invasives

Site group % introduced
Blackland 24.0
Upland 21.3
Bottomland 21.1
Sandy 15.0
Shallow 14.7
Stony/steep 14.5

Invasives ignore the pH gradient that structures the native flora: they follow land use, not soil.

The Lost Pines vs. a planted stand

Loblolly pine (Pinus taeda): 304 records, three-quarters of them in Bastrop County

County Records pH CaCO₃ % Sand %
Bastrop (the Lost Pines) 235 6.0 0.1 56
Leon 15 5.0 0.0 39
Lee 13 6.1 0.1 66
Travis 14 8.0 52.0 11

  • Travis: 13 of 14 records are on one chalky Blackland map unit.
  • Loblolly doesn’t grow wild on chalk; these are plantings
  • Counting one record per 1 km cell shrinks them to 1 of 141 cells and the species’ carbonate figure barely moves

Rare plants test the method’s limits

Records Counties Soil series In reference lists? What the data can say
Texas madrone (Arbutus xalapensis) 646 12 21 4 sites strong signal; both sources agree
Alabama croton (Croton alabamensis) 22 2 6 no describable, not analyzable
Blanco crabapple (Malus ioensis) 1 1 1 no invisible to both sources


  • Madrone: a shallow-limestone specialist: 41% CaCO₃ and 55 cm deep, both among the top 30 of 823 species. Reference lists put it on the same steep limestone sites
  • Crabapple: one record in 42 counties, and absent from the reference communities too. Absence from the data is not absence from the land

Soil explains habitat, not range

  • Alabama croton and Texas madrone grow on the same limestone soils (Brackett and Eckrant)
  • Madrone is in 12 counties; Alabama croton in 2 (Travis and Coryell)
  • Alabama croton’s six soil series cover 1.5 million ha, about 16% of the study area

Its soil is everywhere. Something else, such as dispersal or history, keeps it rare.

Central Texas Soils Explorer

  • Map soil properties (pH, carbonate, depth, crop productivity…) across 42 counties
  • Click a polygon: series description, reference plant community, observed species
  • Search any soil series, or any plant species and see which soils it prefers

Demo: soil properties on the map

  • Soil map units colored by pH, then CaCO₃, then rooting depth
  • Switch to shrink-swell: the Blackland lights up
  • Turn on the county outlines; try another basemap

Demo: regions and ecological sites

  • MLRA layer: the physiographic regions appear
  • Ecological site layer: each site’s dominant plants in the popup
  • Ecological site group: the coarser grouping used for invasion and diversity
  • General soil map: soil associations, the level between the regions and the series

Demo: one soil, all three sources

  • Click a Brackett polygon: the popup
  • Details tab:
    • all 91 soil fields
    • the Official Series Description and the plants it names
    • the reference plant community, the cluster’s indicator species
    • the species iNaturalist observers find there

Demo: why we count places, not photos

  • Click an Eckrant polygon
  • Toggle per observation ↔︎ per 1 km cell
  • Per observation: Ashe juniper and mountain laurel. Per cell: stonecrop and lace hedgehog cactus

Demo: search a series

  • Details tab → Find a soil series → Houston Black
  • Every map unit of the series, the counties it’s mapped in, its ecological sites
  • Show on map: the series outlined across the selected counties

Demo: where does a plant grow?

  • Fetch Juniperus ashei, tick Shade soils where the species occurs: ~97% of polygons light up
  • Fetch Eriogonum multiflorum: 2%
  • Species tab: which soils it prefers, after correcting for where people look

Demo: the species table

  • All 2,887 species, reference and observed, in one table
  • Sort by Δ CaCO₃: the strongest lime-selectors in the region
  • Look up Texas red oak vs. escarpment live oak, and Texas bluebell

Three things to take home

  1. Soil shapes the flora, moderately: about a third of the pattern, with pH and chemistry leading
  2. Common on a soil ≠ preferring it: hundreds of thousands of observations can tell the two apart (Texas red oak vs. live oak)
  3. What grows now has drifted from the reference: brush, invaders and grazing history write over the soil signal

What the data can’t tell us

  • 69% of floristic variation is unexplained: land use, dispersal, history, climate
  • Soil explains habitat, not range: the Lost Pines, Alabama croton
  • Canyon and seep species are invisible: a soil map can’t resolve a canyon wall
  • Observations are biased toward showy plants, cities and roadsides
  • Absence isn’t evidence: rare plants can be missed by both sources

Learn more

  • https://www.nrcs.usda.gov/resources/education-and-teaching-materials/soil-education
  • Web Soil Survey: https://websoilsurvey.nrcs.usda.gov
  • Ecological site descriptions (EDIT): https://edit.sc.egov.usda.gov
  • Official Series Descriptions: https://soilseries.sc.egov.usda.gov
  • UC Davis SoilWeb: https://casoilresource.lawr.ucdavis.edu/gmap/
  • iNaturalist observations via GBIF: https://doi.org/10.15468/dl.344s5x
  • https://passel2.unl.edu/view/lesson/2eafec8dd762/15

Image credits

  • Title, section and soil photos: bartlett-houston-black.jpg, caliche, soil-horizons.gif, soil profiles, vertisol-cracks, gilgai.jpg, vertisols-distribution: add sources
  • Soil color: USDA NRCS Texas, public domain (Wikimedia Commons)
  • Charts and study-area map: the author, from NRCS SSURGO and MLRA data
  • Balcones Fault trend: Alan K. Clark & Ted A. Small, USGS, public domain (Wikimedia Commons)
  • Four landscapes: Pedernales Falls, Aleksomber, CC BY-SA 4.0 · Enchanted Rock, K B, CC BY 2.0 · Clymer Meadow, Wilafa, CC BY-SA 4.0 · Post Oak Savannah, William L. Farr, CC BY-SA 4.0 (all Wikimedia Commons)
  • Texas bluebell: Michelle W., CC BY, iNaturalist observation 168781752
  • Sunflower (H. debilis): Alice Chodura, public domain (Wikimedia Commons)
  • Texas bullnettle: sonnia hill, CC BY 2.0 · Blackfoot daisy: Sue in az, public domain (Wikimedia Commons)
  • Yellow stonecrop: Cody Stricker, CC BY, iNaturalist 42913284 · Lace hedgehog cactus: Turner Brockman, CC BY-SA, iNaturalist 209400263
  • Texas red oak: Katherine Wagner-Reiss, CC BY-SA 4.0 (Wikimedia Commons)
  • Ashe juniper: katsrcool, CC BY 2.0 (Wikimedia Commons) · Eastern redcedar: Linda Jo Conn, CC BY, iNaturalist 260325101
  • Lost Pines: Larry D. Moore, CC BY 4.0 (Wikimedia Commons)
  • Texas madrone: Gary M. Stolz, U.S. Fish and Wildlife Service, public domain (Wikimedia Commons)
  • Alabama croton: Derek Ramsey, GFDL 1.2 (Wikimedia Commons)
  • Block diagrams: Soil Survey of Williamson County, Texas (USDA Soil Conservation Service, 1983), via the UC Davis SoilWeb Series Data Explorer
  • Williamson County map: the author, from NRCS SSURGO
  • Brackett–Lampasas block diagram: Soil Survey of Lampasas County, Texas (USDA SCS, 1991), via the UC Davis SoilWeb Series Data Explorer
  • Round rock, wagon ruts and cover photos: Soil Survey of Williamson County, Texas (USDA SCS, 1983)
  • Padina block diagram: Soil Survey of Milam County, Texas (USDA NRCS, 2004), via the UC Davis SoilWeb Series Data Explorer
  • Section backgrounds: soil pit, USDA NRCS Texas, public domain · Pedernales Falls, Aleksomber, CC BY-SA 4.0 · San Gabriel River, Georgetown, Tony Webster, CC BY-SA 2.0 · Bluebonnets near Marble Falls, Carol M. Highsmith (Library of Congress), public domain (all Wikimedia Commons) · soil map, the author