Planting and first harvest dates of tomato (Lycopersicon esculentum Mi
ll.) from 2 seasons in 3 years at eight locations in Georgia, North Ca
rolina and South Carolina formed 38 environments which were used to de
termine the most reliable method to predict first harvest date of toma
to based on daily maximum and minimum air temperature. Eleven methods
of calculating heat units were chosen for comparison based on their pe
rformance as described in the literature. The most reliable method was
defined as the one with the smallest coefficient of variation (CV). C
Vs were calculated for each method over both seasons and locations, fo
r each season over all locations, each location over all seasons, and
for each season at each location. All heat unit summation methods had
smaller coefficients of variation (CV) than the standard method of cou
nting days from planting to first harvest. Heat unit summation methods
improved harvest date prediction accuracy compared with the counting
day method for tomatoes in the South Atlantic Coast (SAG) region. Pred
iction using location/season specific models were less variable than t
he models over all seasons and locations. Incorporating daylength impr
oved model prediction accuracy when applied over all locations and sea
sons, all locations by season, and all seasons by location, Based on t
he results of this study, the heat unit summation technique recommende
d for this region (where the location and season specific models are n
ot available) is the reduced ceiling method multiplied by daylength. (
C) 1997 Elsevier Science B.V.