Evaluation of Land Surface Models in Reproducing Satellite-Derived LAI over the High-Latitude Northern Hemisphere. Part I: Uncoupled DGVMs

• Published in: Remote sensing (Basel, Switzerland) Vol. 5; no. 10; pp. 4819 - 4838
• Main Authors: Murray-Tortarolo, Guillermo, Anav, Alessandro, Friedlingstein, Pierre, Sitch, Stephen, Piao, Shilong, Zhu, Zaichun, Poulter, Benjamin, Zaehle, Soenke, Ahlström, Anders, Lomas, Mark, Levis, Sam, Viovy, Nicholas, Zeng, Ning
• Format: Journal Article
• Language: English
• Published: Basel MDPI AG 01.10.2013; MDPI
• Subjects: Continental interfaces, environment; Earth and Related Environmental Sciences; Geovetenskap och miljövetenskap; growing season; LAI; land surface models; Natural Sciences; Naturgeografi; Naturvetenskap; northern hemisphere; Ocean, Atmosphere; phenology; Physical Geography; Remote sensing; Sciences of the Universe; trendy
• ISSN: 2072-4292; 2072-4292
• DOI: 10.3390/rs5104819

Leaf Area Index (LAI) represents the total surface area of leaves above a unit area of ground and is a key variable in any vegetation model, as well as in climate models. New high resolution LAI satellite data is now available covering a period of several decades. This provides a unique opportunity to validate LAI estimates from multiple vegetation models. The objective of this paper is to compare new, satellite-derived LAI measurements with modeled output for the Northern Hemisphere. We compare monthly LAI output from eight land surface models from the TRENDY compendium with satellite data from an Artificial Neural Network (ANN) from the latest version (third generation) of GIMMS AVHRR NDVI data over the period 1986–2005. Our results show that all the models overestimate the mean LAI, particularly over the boreal forest. We also find that seven out of the eight models overestimate the length of the active vegetation-growing season, mostly due to a late dormancy as a result of a late summer phenology. Finally, we find that the models report a much larger positive trend in LAI over this period than the satellite observations suggest, which translates into a higher trend in the growing season length. These results highlight the need to incorporate a larger number of more accurate plant functional types in all models and, in particular, to improve the phenology of deciduous trees.