2026-07-02

Reading the Dry:

California Can Learn to See Which Ground Is Ready to Burn. The Harder Question Is Whether It Can Water It.

A Water Policy Series — July 2026

Introduction

The best clue to where the West’s next great fire will start — and how fast it will run once it does — may not be in the trees at all. It is in the ground beneath them. Over the last few years, fire scientists working one range after another have kept turning up the same counterintuitive result: the moisture in the soil under a forest or grassland tracks fire risk better than the air above it — better than temperature, better than wind, better than the drought indices agencies have leaned on for decades.¹ The reason is almost tactile. Soil moisture is the weather’s ledger. It records not just the rain that fell but the water that evaporated back out, and whatever is left in the ground is what decides whether the grass, duff, and roots on top of it are ready to burn.²

It is a real finding and an easy one to oversell, so hold it to its true size. Soil moisture is not a crystal ball; it is the most useful new ingredient in a forecast that still needs every other one — weather, fuels, terrain. In the study that first put numbers to it, folding soil moisture and plant-water stress into a model already fed weather and fuels raised the share of daily fire growth it could explain from about 27 to 36 percent — a real gain that nonetheless leaves most of the day-to-day swing to everything else.² And every result of this kind so far has come from somewhere that is not California: the northern Rockies, the Oklahoma grasslands, the Utah sagebrush steppe. Whether the same signal holds in the state’s chaparral and oak woodland, with their different fuels and Mediterranean fire season, is a question the existing science has not yet asked. That is a reason to test the idea in California — not evidence that it already works here.

Even held to that modest size, the finding is a hopeful one, and it invites a tidy sequence of steps that runs something like this: California is the most fire-exposed state in the country, with more than two million acres of local land now mapped as high or very high fire hazard and roughly one in eight residents living inside the two most dangerous zones.³ California also already operates a statewide grid of automated stations — the California Irrigation Management Information System, or CIMIS, 145 of them, wired since 1982 to measure the weather and the ground.⁴ So: bolt soil-moisture probes onto CIMIS, aim the densest coverage at the fire-hazard maps, watch the dry spots surface as fire season builds — and then, having found the parched basins, rehydrate them, using the one kind of water California has in surplus: the off-peak, wet-season flood flows the state is already learning to sink back into the ground through managed aquifer recharge. Predict the fire, then water the ground so it won’t catch fire. It is a clean, satisfying arc, and the reader who has followed this series will already feel the shape of where it is going to fail.

It fails in the same place these stories always fail: the measuring is the easy part, and the fixing is a different problem wearing the measuring’s clothes. Adding soil moisture to CIMIS is cheap, tractable, and mostly a matter of will. But the second half of the arc — water the fire-prone ground — quietly assumes that the water California can spare and the ground that actually burns are in the same place. They are not. The state’s marquee recharge program pours floodwater into deep-valley aquifers beneath farmland, fighting a real and serious problem that has nothing to do with fire; the ridgelines and forest slopes where megafires start sit far above that water table and are untouched by it. There is a form of recharge that genuinely lowers fire risk — but it is small, distributed, mostly on federal land, and funded by no one whose job is either water or fire. This piece is about that gap: how easy it would be to see the dry, and how hard it is to wet it.

I. The Cheap Half: What CIMIS Already Is, and the One Thing It Doesn’t Report

Begin with the good news because it is real and unusually actionable. California is not starting from nothing. CIMIS is a forty-year-old network of automated stations built by the Department of Water Resources and UC Davis in 1982, originally to help farmers schedule irrigation.⁴ Each station already measures solar radiation, air temperature and humidity, wind, precipitation — and soil temperature.⁵ From those inputs it computes reference evapotranspiration, the ETo number that tells a grower how much water a crop lost yesterday and needs today.⁴ The network is public, free, statewide, and maintained; the poles are in the ground, and the telemetry works.

Here is the gap, and it is a narrow one. CIMIS was built to answer an agricultural question — how much water should I apply? — and for that question the crucial output is evapotranspiration, a demand estimate derived from the air. What CIMIS does not publish as a standard variable is the thing the fire scientists want: soil moisture itself, the actual water content in the soil column at depth.⁵ The stations sense the ground’s temperature but not, as a routine reported product, its wetness. So California has the poles, the power, the data pipes, and the statewide footprint — and is missing only the one sensor and the one data field that would let the same network feed fire forecasting in addition to irrigation. That is the key point: this is an expansion, not a substitution. Adding soil-moisture probes takes nothing away from the agricultural mission CIMIS already serves — every station keeps computing the evapotranspiration a grower needs — and layers a second, fire-relevant data product on top of the first, at the marginal cost of a sensor and its calibration. One network, two jobs. The Oklahoma Mesonet shows a soil-moisture network can carry that second job at exactly this scale: 120 stations, statewide, thirty-minute data feeding a fire-danger model that outperformed the drought-index surrogate it replaced — not as a standalone oracle, but as the sharpest single input in a weighted blend of weather, fuels, and moisture.¹²

The expansion has two parts, and it is worth being honest about the second. The first is the cheap retrofit above: instrument the existing stations, and the ag network starts producing fire-relevant soil-moisture data statewide tomorrow. The second is coverage. CIMIS was built to serve irrigated agriculture, so its stations cluster in valleys and on farmland, and the fire signal matters most in the foothills, forest margins, and wildland-urban interface where the stations are thin. Reaching that ground means extending the network beyond its agricultural core — a phased build-out of new fire-oriented stations aimed at the hazard maps. That is a real cost and a real design question: how many upland stations, sited where, to sample terrain whose moisture varies wildly across a single slope. No one has yet done that siting study, and until they do, “expand CIMIS to cover fire country” is a direction, not a costed plan. California has more stations than Oklahoma and a fire problem Oklahoma can only imagine — but the fire half of the network still has to be built where the fire is, not just switched on where the farms are.

Why not just use satellites and skip the hardware? Because the resolution isn’t there for this job. NASA’s SMAP satellite measures global soil moisture, but its native radiometer resolution is about 36 kilometers, and even blended with radar it reaches only a few kilometers, refreshed every two to three days.⁶ Fire ignites at the scale of a single dry slope. The Colorado researchers in the story that prompted this piece spent three days collecting fifteen hundred handheld probe readings across a hundred-square-meter patch precisely because they found enormous variability within a single hillside — what they call “slope granularity” — a swing that a 36-kilometer pixel averages into meaninglessness.¹ Satellites tell you the region is drying. They cannot tell you which draw above which town is the one that’s ready. That is a ground-sensor question, and it is why the cheap half of this proposal is worth doing even knowing the hard half is hard: better data on where the tinder is has value regardless of whether you can do anything about the tinder, because it sharpens evacuation, prescribed-burn timing, and where you spend the fuel-reduction dollars you already have.

And that ground layer need not be built entirely out of professional stations. The coverage problem has a cheaper second answer, one that borrows its logic from distributed volunteer projects like SETI@home: a citizen-sensing tier — a CoCoRaHS for soil moisture. CoCoRaHS is the volunteer rain-and-snow network that grew to more than 20,000 backyard observers across every state and now feeds the official U.S. Drought Monitor, built on cheap standardized gear and heavy training.⁷ Its soil-moisture equivalent is buildable now because the hardware has arrived: open-source, consumer-electronics stations that cost tens of dollars rather than thousands and, once installed by a homeowner, rancher, resource-conservation district, or school, report themselves automatically — the passive, install-and-forget contribution that made distributed computing work, translated to a physical sensor.⁸ It would not replace CIMIS; it would hang off it, using the professional stations as the calibrated reference the cheap probes are corrected against.

Here the geography does something unexpectedly useful. Citizen sensors cluster where people live — and for wildfire, where people live is the priority zone: the wildland-urban interface, where homes meet fuel, is exactly the ground the CAL FIRE hazard maps care most about and exactly where CIMIS’s valley-agriculture siting is thinnest.³ A crowd-built network is naturally densest where the property-and-life risk is highest, and it lets volunteers absorb the maintenance labor that quietly kills grant-funded sensor networks. But the limits are real. Cheap sensors are worthless uncalibrated — raw low-cost soil-moisture data is unusable for modeling until it is corrected for local soil, salinity, and temperature, which moves the cost from hardware to data engineering.⁸ The crowd covers the interface but not the roadless backcountry, where many large fires build and no one lives to host a probe. And volunteer networks skew toward affluent, engaged communities, leaving the poorer rural and tribal fire zones — the monitoring deserts the original reporting flagged — still dark. A citizen tier is a genuine answer to density and maintenance, a weak one to the backcountry and to data quality, and it earns its keep only anchored to the professional backbone.

II. The Expensive Half: The Water Is Real, but It’s in the Wrong Place

Now the hard half, and this is where the clean arc breaks. Suppose the sensors are in and the map lights up: here are the fire-prone basins, and they are dangerously dry. The proposal is to fix that with managed aquifer recharge — specifically Flood-MAR, California’s strategy of taking high wet-season flows from rain and snowmelt and spreading them across land to sink into the aquifer, instead of letting them run to the sea.⁹ It is one of the most promising water ideas the state has. It fights groundwater overdraft, it reduces flood risk downstream, it builds drought reserves, it slows the land subsidence that is cracking canals across the San Joaquin Valley, and at a median cost around $410 an acre-foot it is roughly a fifth the price of new surface storage.¹⁰ Everything good you have heard about it is true.

But look at where that water goes, because the geography is the whole argument. Flood-MAR, as California actually practices it at scale, recharges the deep alluvial aquifers beneath valley farmland — the flat, permeable, over-pumped basins of the Central Valley where the water table has dropped tens or even hundreds of feet and where there is pore space to bank a flood.⁹ That is exactly the right place to fight overdraft and subsidence. It is exactly the wrong place to fight fire. California’s megafires do not start on the valley floor. They start on chaparral ridgelines and in montane forest, on slopes and in the wildland-urban interface — ground that sits high above the valley aquifer, on thin soils over bedrock, where there is no deep basin to recharge and where the valley’s rising water table is entirely irrelevant to the moisture in the root zone of a manzanita stand a thousand feet up.³ You can bank a million acre-feet under Fresno County and not add a drop of moisture to the slope above Paradise. The water California can spare and the fuel that actually burns are separated by geology and elevation, and no amount of valley recharge closes that distance.

This is the mismatch the clean arc hides. “Rehydrate the dry basins” sounds like one action, but “the basins we can recharge” and “the basins that burn” are two nearly non-overlapping sets. The valuable thing Flood-MAR does — refilling depleted aquifers under irrigated agriculture — and the thing this proposal wants it to do — wetting fire-prone uplands — are different projects in different places for different reasons, and conflating them is how a good idea about monitoring turns into a hand-wave about mitigation.

III. The Honest Bridge: Where Water and Fuel Share the Same Ground

So is there any real version of “recharge the ground to lower fire risk”? Yes — but it is not the valley, and naming it correctly is the point. The place where a rising water table genuinely changes fire behavior is the place where the recharge feature and the fuel occupy the same ground: California’s mountain meadows.

A healthy Sierra meadow is a natural sponge and a natural firebreak at once. In an intact meadow, the stream is connected to a wide, flat floodplain; high flows spread out, slow down, and soak in, lifting the local water table and releasing that water slowly through the dry summer instead of flushing it downstream in a spring pulse.¹¹ The Department of Water Resources and the Sierra Nevada Conservancy have both made meadow restoration a water-security strategy for exactly this reason — a restored meadow extends late-season flow and recharges the shallow groundwater that keeps the meadow and its immediate margins green deeper into fire season.¹¹¹² Keep the scale honest: that effect is a property of the meadow’s own footprint and the narrow band around it, not a moisture halo cast over the surrounding forest; the fire benefit is a firebreak in a specific place, not a hydrated hillside. And crucially, a wet meadow with a high water table grows lush, low, herbaceous vegetation and no dense conifers — which is to say it is, physically, a fuel break. The Conservancy’s restoration strategy explicitly targets expanding wet meadows that already function as fuel breaks, precisely to widen the fire-resistant seams running through the forest.¹² The Forest Service’s “lost meadows” modeling suggests the Sierra’s potential meadow habitat is nearly three times its current extent — a large latent inventory of ground that could be restored to hold water and break fire at the same time.¹³

Here the mechanism connecting groundwater to flammability finally becomes concrete, and it is worth being precise about it because it is the load-bearing science under this whole idea. Vegetation moisture — the thing that determines whether a plant is fuel or firebreak — tracks the depth to the water table wherever roots can reach it. Groundwater-dependent vegetation typically needs the water table within about one to five meters of the surface; when it’s there, roots draw a steady “groundwater subsidy” that keeps foliage hydrated through drought, and when the table drops far below that range, the same plants suffer crown dieback and mortality — in the starkest studies, 92 to 100 percent of saplings dying when the water table fell out of reach.¹⁴ That is the physical chain the whole proposal rides on: shallow water table → hydrated roots → moist foliage → lower flammability. But be clear about how much of that chain is actually nailed down. Each individual link is well supported — water-table depth controls groundwater-dependent vegetation health, and live-fuel moisture tracks soil water more tightly than remote-sensing indices do.²¹⁴ What is not established is the full chain, end to end, as a measured reduction in fire behavior at landscape scale: the mortality studies are riparian-woodland and phreatophyte systems, the fuel-moisture work is plot-scale, and no study cited here demonstrates that restoring a meadow’s water table produces a documented drop in fire spread or severity across a real burn. The chain is plausible and each link is real; the whole is an inference, not a proven result. And even granting it, the chain only closes where the water table is shallow enough for roots to touch — in meadows, riparian corridors, and valley bottoms, not on the bedrock ridgelines. The mechanism’s geography is narrow: recharge can lower fire risk where the water table is within a root’s reach of the surface, and essentially nowhere else.

So the honest version of the arc is not “sense the dry basins and Flood-MAR them.” It is: sense the dry ground with an expanded CIMIS — its professional stations thickened by a CIMIS-calibrated community soil-moisture network reaching into the fire-prone country the official grid misses; where that ground is a degraded meadow or a headwaters wet system, restore it so it holds off-peak water and breaks fire in the same act; and understand that this is a distributed, small-parcel, mountain program — the opposite of the big valley recharge basins — carrying none of the valley program’s scale and none of its funding.

IV. Stakeholders and Money: Everyone Owns a Piece, No One Owns the Problem

This, surfaces the pitfall that is bigger than any hydrology problem: no single agency is set up to do this, because it falls in the seam between two entire fields that do not share a budget, a mandate, or a map.

Sort the players and the seam becomes obvious. On the water side: the Department of Water Resources runs CIMIS and champions Flood-MAR; the State Water Resources Control Board controls the water rights that decide whether anyone may divert an off-peak flow to recharge; and the local Groundwater Sustainability Agencies created under the 2014 Sustainable Groundwater Management Act are the ones actually planning and permitting recharge — but their legal mandate is to end overdraft, full stop. A GSA has no authority, and no funding line, to spend water or money to lower a fire risk; fire is not in its statute. On the fire side: CAL FIRE and the Office of the State Fire Marshal draw the hazard maps and run suppression, but they are not water managers and do not operate recharge. And straddling both, owning most of the ground that actually burns, is the U.S. Forest Service — because the montane forests and the meadows worth restoring are largely federal land, a different government entirely from the state agencies that run the water. The most fire-relevant recharge sits on federal land, is measured by a state irrigation network, would be permitted by a local groundwater agency with no fire mandate, using water rights adjudicated by a state board, to reduce a hazard mapped by a state fire agency. Six institutions, no overlap in mission, and no line item anywhere that reads “spend water to prevent fire.”

The costs split along the same seam, and it’s the cheap half that has a plausible home — for a reason worth drawing out, because the dual-use design is what makes it fundable. Retrofitting CIMIS with soil-moisture probes is genuinely modest — sensors and calibration on poles that already exist — and, crucially, the data it produces is useful to irrigators, drought and groundwater planners, air-quality modelers, and researchers, not only to fire agencies. That many-constituency value is not just a nice-to-have; it is the funding argument. A single-purpose wildfire sensor network has to win scarce fire dollars every budget cycle and defend them against suppression and fuels-reduction line items that always look more urgent, which is exactly how monitoring networks die — not at installation but at maintenance, when the launch grant runs out and no one owns the recurring cost of keeping sensors calibrated and telemetry alive. The Colorado researchers in the story that prompted this piece named that trap directly: installation and upkeep are expensive, and what they were hunting for was sustained state and federal funding rather than a one-time build.¹ A network that serves several missions at once spreads both the capital and the maintenance across several budgets and several beneficiaries — DWR’s irrigation and drought programs, GSAs under SGMA, air-quality agencies, and fire — so no single agency carries the whole cost and no single budget cut kills it. Dual use is how you pay for expansion and keep the lights on afterward. That advantage, though, attaches specifically to the sensing network. The mitigation is where the numbers get hard and homeless. Flood-MAR at scale is a multi-billion-dollar buildout — the recharge projects proposed in SGMA groundwater plans already total more than 2.5 million acre-feet a year at a projected $3 billion in capital¹⁰ — but that spending is aimed at valley overdraft and would happen with or without a fire rationale. Meadow restoration, the piece that actually reduces fire risk, is comparatively cheap per project but slow, permit-heavy, and scattered across thousands of small federal and private parcels, funded today by a patchwork of conservancy grants, bond dollars, and USFS programs that is nowhere near the scale of the three-million-acre restoration opportunity.¹³ The money that exists is pointed at the wrong basins; the money for the right basins doesn’t have an owner.

V. The Pitfalls, Named Plainly

Set the failure modes out in the open, because each one is a place a well-meaning version of this proposal quietly dies.

The measurement will outrun the mandate. This is the Oklahoma lesson turned into a warning. “Getting the data, having the data, and using the data are all separate issues,” one of the Mesonet researchers said; what’s missing, he added, is “a few champions… for implementing these new tools” in practice.¹ California could light up a beautiful statewide soil-moisture map and change nothing, because a map is not a mandate and no agency is obligated to act on the dry pixels. Better data with no one empowered to spend against it is a more precise way of watching the same fires.

The off-peak water only exists in wet years — and fire peaks in dry ones. Flood-MAR runs on surplus: flows above senior water rights and environmental minimums, available mostly in wet winters when farm demand is low.⁹ But the state has already conceded there isn’t enough flood water to correct overdraft, let alone spare a share for fire.¹⁵ Worse, the timing is perverse. The years you most want the ground wet — hot, dry, low-snowpack fire years — are precisely the years there is no flood flow to recharge with. This series has seen this exact trap before on the Klamath, where the flushing flows that break the salmon parasite depend on wet winters the sky may not deliver; a mitigation that requires the weather to cooperate fails in the very conditions that make it necessary.

Point sensors don’t cover slopes. The slope-granularity finding cuts both ways: it is the reason satellites can’t do this job, and it is the reason ground sensors are expensive to scale. If moisture varies wildly across a single hillside, then a few hundred CIMIS stations, however retrofitted, undersample two-plus million acres of hazard.¹³ The network sharpens the regional picture; it does not put a probe on every draw, and modeling the gaps reintroduces exactly the uncertainty the ground sensors were meant to remove.

The evidence is borrowed, and California hasn’t been tested. This is the caveat that sits under all the others. The predictive-skill numbers come from the northern Rockies, the southern Great Plains, and the Great Basin; the meadow-and-fire logic is drawn from Sierra restoration projects valued mostly for water supply and habitat, not measured as fire mitigation. None of it is a California fire result. Chaparral is not conifer forest, a Mediterranean fire season is not a Rockies one, and a proposition that pencils out in Oklahoma may or may not survive contact with the San Gabriels. Treating the out-of-state skill as if it were a California guarantee is precisely the overreach this piece is trying to avoid.

Wetter is not always safer. The tidiest version of this idea assumes hydration only ever lowers fire risk, and the ecology is not that obliging. Wetter ground can grow more biomass — more grass, more brush — which after it cures becomes more fuel; a long study of managed wildfire in Sequoia–Kings Canyon found the hydrology-vegetation relationship stubborn and the soil-moisture response to management surprisingly muted.¹⁶ “Rehydrate the landscape” is not a monotonic dial toward safety; in some regimes it trades a drought-stressed, ignitable landscape for a lush, high-load one, and which you get depends on species, timing, and management you’d have to get right.

The governance seam is the deepest pit of all. Every technical problem above is survivable if some institution owns the whole chain. None does. Until a water agency is funded to buy fire risk down, or a fire agency is empowered to manage water, the sensors and the meadows and the flood flows sit in three different departments’ budgets, and the seam between them is where the whole idea falls through.

Conclusion

Return to the arc we started with, because the appeal of it was never wrong — only incomplete, and worth stating in bounded terms. What is actually established is narrower than the pitch: that soil moisture is a valuable added input to fire models built and tested in other regions; that California already operates a 145-station network measuring most of what such a model needs, minus the soil-moisture field itself; that soil-moisture sensors are mature, off-the-shelf hardware; and that Flood-MAR and meadow restoration are real, funded programs — the first aimed at valley overdraft, the second valued mainly for water and habitat. What remains unproven, and would have to be tested before anyone spends against it, is the part that matters most here: whether the out-of-state predictive skill transfers to California’s fuels and terrain; whether the existing CIMIS footprint samples fire-prone ground or would need a costly upland build-out; and whether restoring a meadow’s water table produces a measurable drop in fire behavior rather than a plausible one. The honest recommendation follows the line between those two lists. The cheap, low-regret move — add soil-moisture probes to CIMIS and run a California validation against the fire record — is worth doing on its own merits, because the data helps drought, air-quality, and irrigation work regardless of how the fire question resolves, and because that very breadth of use is what would fund the network’s expansion and, harder still, its long-term maintenance. The expensive move — rehydrating landscapes to change fire behavior — should stay a hypothesis under test, not a program under construction, until California evidence exists.

Soil moisture is, at best, an early signal of which ground is primed to burn — not a forecast of where a fire will start, and not on its own. California really does own most of the network it would take to read that signal statewide, and finishing that network is the rare climate-adaptation move that is cheap, fast, and useful even to people who don’t care about fire. If the state does one thing from this, it should be that: finish the easy half, get honest eyes on the dry, and find out whether the signal even holds here before betting water on it.

It is the second half — and then water it — where the ambition has to be re-cut to fit the ground. The water California can spare pools under its farmland, and the fires burn on its ridgelines, and no volume of valley recharge climbs that grade. The place the two ever meet is the mountain meadow, where a restored sponge holds the off-peak water and breaks the fire in the same green seam — and that work is small, slow, mostly federal, and owned by no one whose job is water or whose job is fire. So the real finding underneath the hopeful one is a governance finding: California could learn to read which ground is primed to burn — cheaply, and soon. What it cannot yet do is decide who is allowed to spend water to stop it. The sensor is a retrofit. The seam between the agencies is the hard part, and it is the part that will still be there after every probe is in the ground.

Sources