A combined GLAS and MODIS estimation of the global distribution of mean forest canopy height

Abstract

Mapping the global distribution of forest canopy height is important for estimating forest biomass and terrestrial carbon flux. In this study, we present a global map of mean forest canopy height at 500 m spatial resolution obtained by combining Geoscience Laser Altimeter System (GLAS) data acquired from 2005 to 2006 and 13 ancillary variables, including seven climatic variables and six remote sensing variables (nadir BRDF-adjusted reflectance at red and NIR bands, tree cover, anisotropic factor, accumulated Enhanced Vegetation Index, and elevation). The original contributions of this study include the following: (1) The wavelet method was applied to complement the GLA14 product to identify the ground peak and the top-canopy peak. We found that it was useful for dealing with waveforms with low reconstruction accuracy. (2) GLAS data from the leafless season were not used for nonevergreen forest because the height retrieval results exhibited underestimation and strong variations. (3) The anisotropic factor (ANIF), an indicator related to surface structure, was included as an ancillary variable for the first time and was determined to be important for height modeling in the Asian and North American regions. (4) The balanced random forest (BRF) algorithm was applied to register GLAS mean forest canopy height to a 500 m grid considering the small proportion of extreme height classes (tall and short trees), and it achieved good performance in terms of modeling accuracy (RMSE = 2.75 to 4.45 m) and preserving data variation. An inter-comparison among three global forest height maps [the present study, Lefsky (2010), andSimard et al. (2011)] was implemented in a pixel-by-pixel manner. High agreement (R 2 = 0.73, RMSE = 4.49 m) was determined between the present study and Simard et al., whereas the result from Lefsky was notably different from the other two results (R 2 = 0.14, RMSE = 8.92 m, compared with the present study; R 2 = 0.11, RMSE = 11.19 m, compared with Simard et al.). Large disparities were generally associated with evergreen broadleaf forests in South America, deciduous needleleaf forests in Europe and Russian North Asia, and evergreen needleleaf forests on the West Coast of North America. Differences in the height metric were a main factor affecting the disparities among the three results. Validation against field survey data acquired from the Distributed Active Archive Center indicated the accuracy of our mean forest canopy height map (R 2 = 0.63, RMSE = 4.68 m, n = 59).

Hyperspectral Reflectance Anisotropy Measurements Using a Pushbroom Spectrometer on an Unmanned Aerial Vehicle—Results for Barley, Winter Wheat, and Potato

“…For remote sensing purposes, information on the BRDF is important, for example, for the correction of viewing and illumination effects [2] or for the calculation of albedo [3,4]. In addition to this, the BRDF has been shown to contain additional information on vegetation parameters, such as canopy density, foliar water content, nitrogen content and leaf area index [5], canopy height [6], canopy clumping [7,8] and soil properties, such as soil surface roughness [9] and soil moisture content [10]. Moreover, BRDF information can be used to improve classification accuracies [11].…”

Global patterns and determinants of forest canopy height

“…Satellite Light Detection and Ranging (LiDAR) is a state-of-the-art remote sensing technique that provides worldwide measurements of forest canopy height (Harding and Carabajal 2005). Satellite LiDAR data have recently been used to quantify the global distributions (Simard et al 2011, Wang et al 2016) and determinants (Klein et al 2015, Zhang et al 2016) of forest canopy height. However, modeled satellite LiDAR data, instead of original data, are often utilized.…”

Tree-Lists Estimation for Chinese Boreal Forests by Integrating Weibull Diameter Distributions with MODIS-Based Forest Attributes from kNN Imputation

“…Our method framework could use a Weibull function to fit each species component separately and provide the information about species composition across large regions using a kNN model based on MODIS data. MODIS data have wide coverage and high temporal resolution, and have been widely used to generate regional maps of forest attributes [3,5,24,30,41,[46][47][48][49][50]. The advantage of our method framework is its use of MODIS data as predictor variables to estimate stand attributes.…”