Monthly · NOAA nClimDiv · —
US Climate Trends by County and Month
Every US county, every calendar month, 1895 to last month, from NOAA’s nClimDiv county dataset: average temperature, the daily high and the daily low (monthly means of maximum and minimum temperature), precipitation, heating and cooling degree days, and monthly mean dewpoint from PRISM. For each the page shows the trend per decade since 1970 (or 1950, or 1895) with its significance, the 1991–2020 normal, and the trend-adjusted normal, what normal is now if the last five decades’ trend holds. That last number is the one to keep in mind when reading any forecast on this site against a 1991–2020 base.
County map
Loading counties and data…
Trend is the ordinary least-squares slope over the chosen period in units per decade (precipitation in percent of the 1991–2020 normal per decade), fitted to the calendar-month or seasonal series; the Theil–Sen slope and a Mann–Kendall p-value are computed alongside and shown in the tooltip. Counties fade where the trend is not significant at the 5% level; the Mann–Kendall p-value is computed on the raw annual series with no correction for serial correlation, so where a series is autocorrelated the significance is optimistic — read the fading as a rough screen, not a test. “Normal shift” is the trend-adjusted normal minus the 1991–2020 normal.
Where and when: trend by state and month
The chosen variable and start year, one row per state grouped by NOAA climate region, one column per calendar month plus the seasons. The colour is the state-mean trend per decade. Read across a row for the seasonality of a state’s change, down a column for its geography.
One county over time
Click a county on the map, or choose one here. The monthly series for the chosen month or season with the fitted trend for each start year, the 1991–2020 normal band, and the latest value ranked against the full record.
What normal is now: state degree days by month
Heating and cooling degree days by month for a state: the official 1991–2020 normal from the state series, the 1970-start trend-adjusted normal for this year, and the difference in percent. The energy-weather reading of the trend.
Method and caveats
County values in nClimDiv are area means of NOAA’s 5 km nClimGrid, itself interpolated from homogenised station data (COOP, RAWS, SNOTEL and neighbours), not raw station records; the network behind them thins before 1950, so the 1895-start trends carry more uncertainty than their p-values suggest. A linear trend is a description of the chosen period, not a projection, and the start year is a choice: the 1970 start is the period over which the forced warming is close to linear, the others are there so the sensitivity is visible. Precipitation trends are in percent of the normal so dry western counties do not read as large in inches. Degree days come from NOAA’s own daily-based county series, not from monthly means. Dewpoint is not in nClimDiv: it comes from PRISM’s station-based 4 km monthly mean dewpoint grids, area-averaged over each county, so it covers the lower 48 only and its most recent eight months are provisional and are re-read on every run. Data are refreshed on about the 9th of each month, once NOAA’s monthly update has landed.