Tree height is an important variable which can be used in estimates of forest biomass and carbon uptake. Application of airborne light detection and ranging (LiDAR) technique has become the method of choice for remote estimates of forest Net Primary Production (NPP) and carbon uptake. LIDAR accuracy, however, depends on calibration with ground-based observations, which are time-consuming and costly. We propose an alternative calibration, using a network of more than 300 internet cameras (PHENOCAM), to estimate tree height extension using a simple triangulation technique from onsite measurements of marker points. Our PHENOCAM site (E New York, 42.5266 N; 74.1587 W) uses a fixed camera focused on a pure stand of mature white spruce (Picea glauca). During 2015-2016 we conducted daily measurements of heights of 7 trees, as well as their radial growth, determined using two-point high resolution dendrometers. In situ sensors recorded soil temperature and moisture, and air temperature, precipitation, wind speed and solar radiation were recorded at an automated weather station located 70 m distant.
According to our estimates, a single measurement of tree height using a PHENOCAM camera can be performed with an accuracy of ±3 cm. Daily variability, however, is ± 6 cm, compared with a seasonal increase in average height of about 20 cm in 2015 and 15 cm in 2016. Direct comparisons of average daily height with wind speed revealed significant negative correlations, with stronger winds leading to shorter height estimates. In addition, we determined that, while vertical growth ended in late June, radial growth continued for another 2-3 months. If these phenological patterns are consistent and general, they will need to be considered when height is used as a surrogate for new biomass. The underlying reasons for the wind effect we observed over growing seasons remains ambiguous. Strong winds can either bend stem tops or cause mechanical damage, in either case reducing apparent height. Overall, our observations caution against comparison of ground-based measurements of tree height with high resolution remote sensing images and LiDAR data without reference to wind speed. Wind-related error in tree height measurements of 5 cm could lead to a 20-30% error in estimates of annual NPP.