U.S. Department of Agriculture: Forest Service -- National Agroforestry Center


Document Type


Date of this Version



J.R. Butnor et al. / Agricultural and Forest Meteorology 192–193 (2014) 121–131; http://dx.doi.org/10.1016/j.agrformet.2014.03.005 0168-1923/Published by Elsevier B.V.


U.S. Government Work


Soil frost depth in forest ecosystems can be variable and depends largely on early winter air temperaturesand the amount and timing of snowfall. A thorough evaluation of ecological responses to seasonallyfrozen ground is hampered by our inability to adequately characterize the frequency, depth, durationand intensity of soil frost events. We evaluated the use of ground penetrating radar to nondestructivelydelineate soil frost under field conditions in three forest ecosystems. Soil frost depth was monitoredperiodically using a 900 MHz antenna in South Burlington, Vermont (SB), Sleepers River Watershed,North Danville, Vermont (SR) and Hubbard Brook Experimental Forest, New Hampshire (HBEF) duringwinter 2011–2012 on plots with snow and cleared of snow. GPR-based estimates were compared to datafrom thermistors and frost tubes, which estimate soil frost depth with a color indicating solution. In theabsence of snow, frost was initially detected at a depth of 8–10 cm. Dry snow up to 35 cm deep, enhancednear-surface frost detection, raising the minimum frost detection depth to 4–5 cm. The most favorablesurface conditions for GPR detection were bare soil or shallow dry snow where frost had penetrated to theminimum detectable depth. Unfavorable conditions included: standing water on frozen soil, wet snow,thawed surface soils and deep snow pack. Both SB and SR were suitable for frost detection most of thewinter, while HBEF was not. Tree roots were detected as point reflections and were readily discriminatedfrom continuous frost reflections. The bias of GPR frost depth measurements relative to thermistors wassite dependent averaging 0.1 cm at SB and 1.1 cm at SR, and was not significantly different than zero. Whenseparated by snow manipulation treatment at SR, overestimation of soil frost depth (5.5 cm) occurredon plots cleared of snow and underestimation (−1.5 cm) occurred on plots with snow. Despite somelimitations posed by site and surface suitability, GPR could be useful for adding a spatial component topre-installed soil frost monitoring networks.