Impact of leaf area index from various sources on estimating gross primary production in temperate forests using the JULES land surface model

The effects of various types of leaf area index (LAI) on simulations of stand scale gross primary production (GPP) in evergreen needleleaf (TCK) and deciduous broadleaf (TBK) temperate forests were analyzed using the Joint UK Land Environment Simulator (JULES) land surface model (LSM). LAI was observed during the period from 2015 to 2017 with an LAI-2200 plant canopy analyzer. Other LAIs included the enhanced vegetation index (EVI) of two satellites with different temporal and spatial resolutions (Landsat and MODIS), the LAI product of MODIS, and the JULES phenology model. These LAIs were compared with in situ observations of LAI, and their effects on GPP simulation were compared with simulated GPP using in situ observed LAI (GPPobs). Our results show that the JULES phenology model does not adequately express the LAI of either TCK or TBK forests. This misrepresentation was improved through simple modifications. LAI data from a remote-sensing EVI corresponded relatively well with the observations, whereas the MODIS LAI product, with a relatively low spatial resolution (500 m), could not discern the complex heterogeneity of a temperate forest and provided a 31% higher maximum value. Compared with GPPobs, the difference in annual GPP was smaller than in LAI, ranging from -3.48% to 6.0% in TCK and -6.52% to 10.9% in TBK. The differences in seasonal GPP estimates were up to 14.6% in spring for TCK and -17.4% in autumn for TBK. Our study identified limitations of the JULES leaf phenology model, and suggested modifications to reduce the differences in GPP estimation for temperate forests at the stand scale. In addition, we suggested using sufficiently high spatial resolution data, especially for LAI, as inadequately represented forest cover types resulted in seasonal GPP discrepancies.