The timing of the last freeze in spring and the first freeze in fall of the year has implications for agriculture, ecosystems, and water resources. In Thomas and Bosilovich (2026), indices used to represent these dates are computed from NASA’s MERRA-2 satellite data retrospective analysis.
On average, the timing of the last spring freeze (LSF) ranges from early January in the southernmost parts of the country to after May 31 in the Intermountain West (Fig. 1, upper left). The climatological date of the first fall freeze (FFF) ranges from early August in parts of the Intermountain West to December in the Southern US (Fig. 1, upper right). The standard deviation in these dates is about 1-2 weeks in most regions, with higher variability in the LSF date in the Southern US.
Figure 1. Climatology (top row) and standard deviation (days; bottom row) in the dates of the last spring freeze (left column) and the first fall freeze (right column) over 1991-2020 from MERRA-2. Grey shading indicates regions where it is too warm to compute a climatological LSF or FFF.
In general, MERRA-2 shows that over 1980 – 2023 the last spring freeze date has trended earlier, and the first fall freeze date has trended later, resulting in a longer freeze-free season (Figure 2). The strongest trends in the LSF and FFF are up to 0.4 days/year, suggesting the LSF (FFF) has shifted up to 2 weeks earlier (later) over this period. The largest trends occur in the Northwest and Northeast US, where the freeze-free season has lengthened by ~3 weeks over 1980-2023.
Figure 2. Trends (days/year) in the date of the last spring freeze, date of the first fall freeze, and length of the growing season over 1980-2023 from MERRA-2. Hatching indicates trends that are significant at the 95% confidence level using the Mann Kendall test.
Understanding drivers of interannual variability in the timing of the LSF and FFF could offer insights into the predictability of these events, so this study also examines connections with modes of climate variability. The main novelty of this work is the finding that the East Pacific/North Pacific (EP/NP) pattern is significantly correlated with the timing of both the LSF and FFF in most regions of the US (Figure 3). The EP/NP pattern is a teleconnection pattern primarily active in spring through fall with anomaly centers over the North Pacific, Alaska, and the Eastern United States (
https://fluid.nccs.nasa.gov/reanalysis/modes/epnp).
Figure 3. Correlations over 1980-2023 between the index representing the EP-NP pattern and (upper left) timing of the LSF, (upper right) timing of the FFF, (lower left) spring (March, April, May; MAM) T2MMIN, (lower right) fall (September, October, November; SON) T2MMIN. Statistical significance levels for each correlation value are shown above the color bar.
Forecasting of anomalies in the timing of the LSF and FFF on the subseasonal-to-seasonal (S2S) time scale would be crucial for agricultural planning. The connections found in this study may help in identifying forecasts of opportunity for S2S prediction models, and computing metrics such as the LSF and FFF date from these models is an area of active research.
References:
Thomas, N.P. and M.G. Bosilovich, (2026). Variability and Trends in Dates of the Last Spring Freeze and First Fall Freeze over the United States. Journal of Applied Meteorology and Climatology, Vol. 65 (pp 1017-126) - DOI: 10.1175/JAMC-D-25-0170.1
