Rain on 115 acres

This is the real precipitation field around the farm at 800 metre resolution. Every square is about 171 acres of ground, and the whole farm fits inside a single one of them. Annual totals in this one view run from to inches: a 34 inch spread inside a box 50 miles across. That is the first thing to understand about rainfall here. It is not a county-level number, it is a hillside-level number.

Switch to Terrain to see the ground underneath, and the two maps stop looking like a coincidence. Hovering reads both layers at once.

Elevation is the obvious explanation and it is mostly the wrong one. Across all 9,072 cells in the map above, height explains only 12 percent of the variation in annual rainfall. The lapse rate is a feeble 2 inches per 1,000 feet, and the scatter around it is wider than the signal.

Add a single extra term, how much higher a place sits than the land upwind of it, and the model explains 64 percent. The field is organised by exposure, not altitude. Two gauges 12 kilometres apart make the same point without any statistics: Blowing Rock and Banner Elk sit within 40 metres of the same elevation and are 14 inches apart, and Beech Mountain is 370 metres higher than Blowing Rock and 12 inches drier.

Two national gridded datasets cover this farm and they disagree by 8.5 inches, so the first job was deciding which to believe. Both were tested against the 11 gauges within 55 km that have 20 or more complete years in the 1991 to 2020 window, using each station's own observed annual totals rather than its published normal.

PRISM wins clearly: almost unbiased, and within about 1.5 inches. Daymet runs 2.4 inches wet at every single station. So the estimate is built on PRISM, then nudged by the local gauge network.

There is no dry season here. The wettest month, July, gets 5.0 inches and the driest, February, 3.2, a remarkably flat distribution by any standard. Summer carries 28.5 percent of the year and winter 21.9 percent.

That flatness hides the real constraint. Reference evapotranspiration, the water the atmosphere pulls back out of grass and soil, totals inches a year against 50 inches of rain. The surplus is only inches, and it does not arrive where the grass needs it: May, June, July and August all run a deficit. Most of the farm's usable water falls in the cold months, when nothing is growing.

Fifty inches is the long-run mean. The farm has never received exactly fifty inches. Scaling the 46 year Boone record to this location gives a standard deviation of 11 inches, a coefficient of variation of 23 percent, and a range from 30.3 inches in 1988 to 79.0 inches in 2018. The gap between the driest and wettest year on record is 152 million gallons, very nearly a whole extra average year of water.

An inch of rain on an acre is 27,154 gallons. On 115 acres every single inch is 3.12 million gallons, so the whole year comes to gallons. Spread evenly it would stand 4 feet 2 inches deep over the entire farm, and it arrives at an average of 297 gallons a minute, continuously, all year.

Every figure here traces to public measurement. Nothing is modelled from a regional average.

• Depth. PRISM 800 m 1991–2020 annual normal read at the farm centroid, 48.35 inches, corrected by the residuals of 20 gauges within 13 km, weighted by inverse distance squared and by record length. Six method variants span 48.4 to 54.5 inches; the median, 50 inches, is the estimate. Leave-one-out validation on long-record gauges gives a bias of 0.07 inches and an RMSE of 2.0 inches, which is the stated uncertainty.
• Gauge normals. Volunteer gauges miss days, and the missing days are not random. Each station was compared with the 47 year Boone record using only the days both reported, then that ratio applied to Boone's own observed mean. This removes the missing-day bias entirely, which matters: the nearest gauge's raw complete-year average is 4 inches higher than its paired-day result.
• Terrain and exposure. Elevation is USGS 3DEP resampled to the same 800 m grid as the rainfall. The exposure term is how far a cell stands above the land 16 km upwind of it; bearing and distance were both scanned rather than assumed.
• Volume. 1 acre-inch = 27,154.286 US gallons = 102.790 m³. Water density at 15 °C. Drop counts use Marshall-Palmer with N₀ = 8,000 m⁻³mm⁻¹ and Λ = 4.1R⁻⁰·²¹.
• Resolution is the hard limit. One PRISM cell is 171 acres and the farm is 115, so the farm is smaller than the smallest thing any available dataset can resolve. Nothing can tell you how rainfall varies between the top and bottom of this tract. The local gradient is 3.26 inches per kilometre pointing east, implying roughly 2 inches of difference across the property, but that is at the resolution floor and inside the error bar. Treat it as indicative, not measured.
• Gauges undercatch, and so does PRISM. Unshielded gauges lose a few percent of rain to wind and considerably more of snow. PRISM is built from those same gauges and inherits it. True precipitation is likely 2 to 5 percent above every figure here, most of the shortfall in the winter snow months. These numbers are gauge-equivalent, which is the convention and the comparable quantity.