Last updated: August 13, 2026
Every number on your dashboard comes from somewhere — a published turf-science model, a rule of thumb, an AI estimate, or something you typed in yourself. This page tells you which is which, honestly, so you know how much weight to put on each one.
Forecasts, current conditions, and hourly observation history come straight from the National Weather Service (api.weather.gov) — the same source meteorologists use, not a third-party aggregator. Your course's coordinates are geocoded once from its city/state and cached.
We cache each fetch for 15 minutes per course. Within that window, loading a page reuses the cached data instead of re-hitting NWS; after 15 minutes, the next page load triggers a fresh pull. If NWS is unreachable, we serve the last known data rather than show nothing.
We use a separate source from the rest of the weather page for this one: NWS station observations frequently report no precipitation reading at all — even in clear weather — so summing them would leave real gaps. Instead, actual daily rainfall comes from Open-Meteo's blended forecast/reanalysis model at your course's exact coordinates, refreshed daily and backfilled to January 1st using their archive of historical data — so you get a true year-to-date total from day one, not just days tracked since setup.
The 10-year average is computed once per course (then cached indefinitely — climate normals don't change day to day) as the mean daily rainfall on each calendar date over the past 10 years at that same location, accumulated from January 1st. "Ahead" or "behind" compares your real, tracked total against that historical baseline for today's date.
This map isn't a sensor reading — no soil probe network exists at this scale. It's the top 4 inches of soil temperature from NOAA's GFS weather model (the same national forecast model behind most weather apps), which computes soil state as part of running its forecast. It updates 4 times a day; we refresh our copy every 6 hours.
The resolution is coarser than the radar or rainfall data — about 17 miles per grid cell, versus radar's 1km — so treat it as a regional trend, not a precise reading for your exact location. Rendering happens outside the main app on a schedule, since decoding weather-model data isn't something a typical web server is built to do.
We use the standard simple-average method with a 50°F base: GDD = ((high°F + low°F) / 2) − 50, floored at zero. It's computed once per day from that day's forecast high/low and added to your season total, which always starts from January 1st — backfilled from historical weather data if your course was set up later in the season, the same way rainfall tracking is, so a mid-season signup doesn't understate the real season-to-date total.
Dollar Spot PEER-REVIEWED — the Smith-Kerns model (Koch et al., 2018, PLOS ONE), a logistic regression built from real disease-outbreak data. It takes the mean temperature and relative humidity over the trailing 5 days of hourly observations and outputs a probability of outbreak. We flag action-needed at 20% probability, matching the model's published threshold.
Pythium Blight RULE OF THUMB — a threshold model (after Nutter-Shane): elevated when the trailing 24 hours had a high above 86°F, a low above 68°F, and at least 14 hours above 90% relative humidity. Rule-based, not a fitted statistical model.
Brown Patch RULE OF THUMB— we'll be direct about this one: there's no formally validated predictive model for Brown Patch in the turf-science literature we're aware of. We flag elevated risk when overnight lows stay above 68°F with 6+ hours above 95% humidity, which is a reasonable qualitative signal, not a calibrated probability like Dollar Spot's.
This module doesn't predict anything — it tracks what you tell it. Nutrient deficiency flags compare your most recent soil test against standard reference ranges (pH 6.0–7.0, phosphorus 25–50 ppm, potassium 100–200 ppm, iron 80–120 ppm). Your annual nitrogen target is whatever you enter for the fiscal year; the progress bar is just your logged applications summed against that number.
All three models below key off season-to-date GDD (base 50°F, as above) rather than the calendar — pest development tracks heat accumulation, not dates.
Crabgrass and Annual Bluegrass Weevil (cool-season grasses only) RULE OF THUMB — GDD windows corroborated against public extension guidance (Purdue, Michigan State GDDTracker, UW-Madison).
White Grub RULE OF THUMB — LOWER CONFIDENCE— we'll flag this one honestly too: these GDD windows come from general industry guidance rather than a primary university extension source, so treat the timing as directional.
Tonight's ET (evapotranspiration) target uses the FAO-56 Hargreaves equation PEER-REVIEWED— the standard approximation for when you don't have a full weather station (solar radiation, wind speed, vapor pressure) on site, using just temperature, your course's latitude, and day of year. That inches-of-water figure is converted to gallons using your maintained acreage.
Soil moisture readings YOUR DATAare logged by hand — we don't have sensor integration yet — and shown as a separate signal alongside the ET target rather than blended into one number. You're meant to read both, not just one.
Suggested maintenance schedules AI ESTIMATE — when you ask us to suggest a schedule for a piece of equipment, Claude generates it from general knowledge of typical maintenance intervals for that type of machine. It is notlooking up your specific manufacturer's service manual — always cross-check against the actual documentation for your equipment before relying on it.
5-year replacement planning SIMPLE HEURISTIC— every piece of equipment is planned for replacement 5 years from its purchase date, flat, regardless of type. It's a usable default for capital planning, not a model tuned to mower duty cycles vs. utility vehicles vs. sprayers.
Both features are Claude (Anthropic), given real data pulled from your course each time — current weather and disease risk, GDD-based pest status, your recent pest and fertilizer applications, today's scheduled tasks, equipment maintenance status, and (in the chat) your budget, labor, and irrigation history too. It's answering from your actual records, not guessing.
We explicitly instruct it not to invent facts and not to recommend specific pesticide products or application rates — for those, it'll point you to a licensed applicator or the product label rather than make the call for you.
If something looks off, or you want more detail on any of these, reach out at mikeconley7@gmail.com. See also our Terms of Service and Privacy Policy.