We hypothesize that the strong sensitivity of radar backscatter to surface dielectric properties, and hence to the phase (solid or liquid) of any water near the surface should make space-borne radar observations a powerful tool for large-scale spatial monitoring of the freeze/thaw state of the land surface, and thus ecosystem growing season length. We analyzed the NASA scatterometer (NSCAT) backscatter from September 1996 to June 1997, along with temperature and snow depth observations and ecosystem modeling, for three BOREAS sites in central Canada. Because of its short wavelength (2.14 cm), NSCAT was sensitive to canopy and surface water. NSCAT had 25 km spatial resolution and approximately twice-daily temporal coverage at the BOREAS latitude. At the northern site the NSCAT signal showed strong seasonality, with backscatter around -8 dB in winter and -12 dB in early summer and fall. The NSCAT signal for the southern sites had less seasonality. At all three sites there was a strong decrease in backscatter during spring thaw (4--6 dB). At the southern deciduous site, NSCAT backscatter rose from -11 to -9.2 dB during spring leaf-out. All sites showed 1--2 dB backscatter shifts corresponding to changes in landscape water state coincident with brief midwinter thaws, snowfall, and extreme cold (Tmax<-25 ¿C). Freeze/thaw detection algorithms developed for other radar instruments gave reasonable results for the northern site but were not successful at the two southern sites. We developed a change detection algorithm based on first differences of 5-day smoothed NSCAT backscatter measurements. This algorithm had some success in identifying the arrival of freezing conditions in the autumn and the beginning of thaw in the spring. Changes in surface freeze/thaw state generally coincided with the arrival and departure of the seasonal snow cover and with simulated shifts in the directions of net carbon exchange at each of the study sites. ¿ 1999 American Geophysical Union