Every number below is either a published model or a threshold from an agronomic source, and each one is cited. If you think one of them is wrong, the source is right there to argue with.
Growth potential
Growth is scored with the PACE Turf Growth Potential model: a Gaussian on mean daily air temperature about the optimum for cool-season grasses, then modified by day length.
It is the established turf-industry standard, and it captures the thing calendar advice misses — that a mild December and a mild April are not the same, because the light is not the same.
Growth is treated as a gate, not as a score. It answers a different question from the rest of the model: wetness, firmness, temperature and wind say whether a cut would go well, and growth says whether one is wanted at all. Averaging the two gave a number that meant neither — a dry, firm, still afternoon in October came out as worse *conditions* than the identical afternoon in June, when the weather was the same and only the need had changed. So growth carries no weight in the score and none in the odds. Below the threshold the site says there is nothing to cut, and says separately whether the weather would have been any good.
Soil temperature is reported at 6cm — germination depth — and compared against a species-aware threshold. The default is a ryegrass and fescue blend in the UK and Ireland, and a bluegrass and tall fescue blend in the US and Canada.
A single germination number is wrong in both markets. UK lawns lean perennial ryegrass and fescue, which germinate around 8–10°C; US cool-season lawns lean Kentucky bluegrass and tall fescue, which need 10–13°C. Four degrees is several weeks of spring.
Water is tracked in three connected stores, hour by hour: free water on the leaf, free water in the thatch at the base of the sward, and wetness in the soil beneath. Rain fills the leaf first and falls through to the thatch and then the soil; dew settles on the leaf alone. Evaporation takes the leaf first and reaches the thatch only with whatever demand is left over. Water also moves back upwards — the thatch wicks into the leaf, and saturated ground feeds the thatch. Dew burn-off is an output of that balance rather than a fixed time of day.
Wet grass tears instead of cutting, clogs the deck and spreads disease, so this carries the heaviest weight. "It has not rained today" is not good enough: a clear, still night lays dew that holds until mid-morning, while a breezy night lays none at all. Nor is a single store good enough. An earlier version of this model capped surface water at 3mm, which meant 8mm of overnight rain and 25mm left the lawn in exactly the same state and dried in exactly the same time. What separates them is not the leaf, which dries at much the same rate either way, but the ground underneath: once the soil is saturated the sward has nowhere to drain to, and stays wet for days.
An hourly wetness index for the soil — a decaying sum of the water that actually reached it, less evapotranspiration, with a ten-day memory — combined with the provider's modelled soil moisture in the top 9cm. The index leads, four parts to one. Both are read against your own soil rather than an assumed one: a year of that grid cell's moisture record gives the water content at saturation, which identifies its texture class, and the class supplies the field capacity everything is measured from.
Soft ground ruts under a mower and badly under a ride-on, and the damage lasts months. A single moisture reading understates the memory of a wet fortnight, so the index leads and the moisture reading corroborates it. Three details matter. The index steps hourly rather than daily: a daily one gave every hour of a day the same value derived from that whole day's rain, so nine in the morning was penalised by rain that fell at nine at night, and the ground could never recover across a sunny afternoon. It counts the water that reached the soil rather than the water that fell — a light shower on a warm day evaporates off the leaf and softens nothing. And the soil is identified rather than assumed, because volumetric moisture is not comparable between places: it is capped by the pore space of whatever soil is there, so a single threshold reads a clay as permanently waterlogged and a sand as permanently parched. Measured against a loam, Glasgow, Tralee, Fort William, Oban, Kilmarnock and Falkirk were blocked as waterlogged every hour of every forecast, and could never show a mowable hour. Where the record cannot be had the site falls back to a loam and says so.
Air temperature is scored with a penalty at both ends: below about 10°C the score falls away towards the frost block, and above 22°C it falls again, reaching half by 26°C. Each profile also sets a minimum temperature score that a weighted mean cannot override.
Cold grass is brittle and recovers slowly from the cut. Above roughly 26°C the plant is already shedding water faster than it can take it up, and cutting adds a wound to a plant that is losing the argument. Temperature carries only a tenth of the weight by design, so without a floor a hot afternoon where everything else is perfect would still score as a good day — which is why the floor exists.
Wind is scored, lightly. Calm through a stiff breeze reads as fine; above about 13 m/s the score reaches zero. It never carries more than a twentieth of the weight, and it never decides a day on its own.
Wind affects where the clippings land and how well a mulching deck works. That is all. It is included because it is a real, if minor, part of whether a cut goes well, and excluded from mattering much because pretending otherwise would be dishonest about what actually governs a good cut.
Growth potential is converted into millimetres of leaf per day, then accumulated forward day by day until the sward has put on enough to warrant a cut.
Accumulating day by day rather than dividing by an average matters: a cold snap in the middle of the week genuinely delays the next cut, and an average would smooth that away. Where growth never reaches the threshold inside the forecast, the site says so rather than extrapolating a date it cannot support.
Mowing before 8am on a weekday, before 10am on a Sunday or bank holiday, or after 8pm on any day is discouraged. It is a preference applied to the score, never a block, and it can be switched off.
There is no UK law setting mowing hours. Councils act under the statutory nuisance provisions of the Environmental Protection Act 1990, which weigh a persistent pattern rather than a single Sunday morning. Many councils publish advisory "considerate hours"; those are what this encodes. It is etiquette, not law.
Any hour at or below 1°C air temperature is blocked outright. So is any hour where the grass canopy itself is estimated to be at or below freezing, even when the air is warmer.
Mowing frosted grass ruptures the leaf cells, and the damage shows for weeks. Grass radiates heat to a clear sky and sits several degrees below air temperature on still nights — which is why a 3°C clear morning can carry a frost that an air-temperature check would miss.
Every recommendation on this site is a count, not a claim. The forecast is fetched as an ensemble — dozens of equally plausible model runs rather than one best guess — and the whole water balance above is run separately against each one. The figure shown is the share of those runs in which the hour came out mowable on your profile. A window's figure is the share in which it stayed mowable for the window's whole length, counted run by run rather than multiplied hour by hour, because consecutive hours are strongly related: the runs that rain at two are largely the runs that rained at one. The hour-by-hour detail behind each day is drawn from the same runs — the chance of rain, the rainfall expected, and the state of the grass and the ground are all averages across them — so the conditions shown and the figure above them always describe the same forecast rather than two different ones.
A forecast is a distribution, and the useful question before wheeling the mower out is not "will it be dry" but "how likely is this to hold". Counting runs also fixes something a single forecast cannot express. Rain at eleven is the same run's wet grass at two, so the risk carries forward through the day instead of each hour being judged on its own — which is why an earlier version of this site could recommend an afternoon with rain plainly on the way.
Open-Meteo does publish a ready-made rain probability and this site deliberately does not use it. At a UK location the rainfall figures come from the Met Office 2km model, whose ensemble is not open data, so that probability is quietly filled in from a 25km global ensemble instead — a different model at a tenth of the resolution, with no indication of which hours are affected. It is also not really hourly: it is a count over the ensemble's own three-hour step, smoothed onto an hourly scale, which flattens the peak and leaks rain into hours that are genuinely dry. Counting members directly avoids all three problems.
The odds describe the weather and nothing else. Whether the grass is growing enough to be worth cutting is decided separately and stated separately — it is the same in every model run, so folding it in could only have shifted the percentage without adding any uncertainty to it, and part of "70%" would have quietly meant "the grass is growing slowly", which is not a chance of anything.
One limitation, stated plainly: only rainfall is sampled. Temperature, growth and wind are taken from the single deterministic forecast, because they are far more predictable at these ranges and vary far less between runs. So the odds are a little more confident about heat and growth than they strictly should be. And these are model-derived frequencies that have not yet been checked against what actually happened — until they have been, treat 70% as "most runs, not all" rather than as a calibrated seven-in-ten.
A second soil temperature is computed from air temperature alone — an exponentially weighted mean of daily mean air temperature, with a time constant set by depth — and compared against the provider’s modelled figure. The comparison is shown on every soil temperature panel.
The reported figure comes from one land-surface model, and a single number presented alone invites more confidence than it has earned. This second estimate uses different physics and different inputs, so agreement is real evidence and disagreement is a warning worth passing on rather than hiding. Soil at depth follows the surface with its swings damped and its timing lagged, and for a periodic signal that lag grows roughly linearly with depth — which is the whole model.
Forecasts come from Open-Meteo: hourly air temperature, dewpoint, humidity, precipitation, wind, shortwave radiation, evapotranspiration, and modelled soil temperature and moisture. Alongside that, an ensemble of precipitation runs — MOGREPS-UK at 2km and ICON-EU at 7km where they reach, and the ECMWF ensemble everywhere — which is what the odds are counted from. Cached and refreshed through the day. Open-Meteo data is used under CC-BY 4.0; the rain probabilities shown here are derived by this site from the ensemble members rather than taken from Open-Meteo's own probability field.
Soil temperature is modelled rather than measured — there is no dense network of soil probes in anyone’s garden, and the site says so wherever the figure appears.
We do not state a verdict without the reason. Every day and every window exposes its factor breakdown and names the limiting factor.
We do not fabricate a fallback. If the forecast service is down, or the soil never reaches your seeding threshold inside the forecast, the site says exactly that instead of estimating.
We flag low confidence. Anything beyond about seven days is marked as such, because a ten-day forecast is not a ten-day promise.
Profiles never override the physics. They change how much risk you are willing to take. Frost, rain, snow, frozen ground and waterlogging block mowing on every setting.
We use your soil, not a standard one. How wet is too wet depends on what the ground is made of, so the site works out the soil texture where you are from a year of that spot's own moisture record. If it cannot, it assumes a loam and says so.