A Simple Method for Retrieving Understory NDVI in Sparse Needleleaf Forests in Alaska Using MODIS BRDF Data

Remote Sensing

Wei Yang²  Hideki Kobayashi²  Rikie Suzuki²  Kenlo Nishida Nasahara¹  Santonu Goswami³  Daniel J. Hayes³  Guido Grosse³  Benjamin Jones³ 

[1] Faculty of Life and Environmental Sciences, University of Tsukuba, Tsukuba, Ibaraki 305-8572, Japan; E-Mail:;Department of Environmental Geochemical Cycle Research, Japan Agency for Marine-Earth Science and Technology, Yokohama 236-0001, Japan; E-Mails:Department of Environmental Geochemical Cycle Research, Japan Agency for Marine-Earth Science and Technology, Yokohama 236-0001, Japan;

关键词: understory NDVI;    boreal sparse forests;    MODIS;    BRDF;   

DOI  :  10.3390/rs61211936

来源: mdpi

【 摘 要 】

Global products of leaf area index (LAI) usually show large uncertainties in sparsely vegetated areas because the understory contribution is not negligible in reflectance modeling for the case of low to intermediate canopy cover. Therefore, many efforts have been made to include understory properties in LAI estimation algorithms. Compared with the conventional data bank method, estimation of forest understory properties from satellite data is superior in studies at a global or continental scale over long periods. However, implementation of the current remote sensing method based on multi-angular observations is complicated. As an alternative, a simple method to retrieve understory NDVI (NDVIu) for sparse boreal forests was proposed in this study. The method is based on the fact that the bidirectional variation in NDVIu is smaller than that in canopy-level NDVI. To retrieve NDVIu for a certain pixel, linear extrapolation was applied using pixels within a 5 × 5 target-pixel-centered window. The NDVI values were reconstructed from the MODIS BRDF data corresponding to eight different solar-view angles. NDVIu was estimated as the average of the NDVI values corresponding to the position in which the stand NDVI had the smallest angular variation. Validation by a noise-free simulation data set yielded high agreement between estimated and true NDVIu, with R² and RMSE of 0.99 and 0.03, respectively. Using the MODIS BRDF data, we achieved an estimate of NDVIu close to the in situ measured value (0.61 vs. 0.66 for estimate and measurement, respectively) and reasonable seasonal patterns of NDVIu in 2010 to 2013. The results imply a potential application of the retrieved NDVIu to improve the estimation of overstory LAI for sparse boreal forests and ultimately to benefit studies on carbon cycle modeling over high-latitude areas.

【 授权许可】

CC BY   
© 2014 by the authors; licensee MDPI, Basel, Switzerland.