Two separate periods of the LAI–VIs relationships using in situ measurements in a deciduous broadleaf forest

► NDVIHSSR and EVIHSSR capture the seasonal variations of LAI, but prematurely estimate the onset of the growing season because of the effect of the forest floor on the remote sensing images. ► The relationships between NDVIHSSR and EVIHSSR and in situ LAI were both better characterized as linear than non-linear for the one-period relationship. ► The two-period relationships better represented the seasonality of in situ LAI and in situ VI, and thus result in more accurate LAI estimations than the one-period relationship. ► EVI produced a better estimate of LAI than NDVI, especially in the leaf-expansion to -saturation period. Leaf area index (LAI) is an important parameter in many ecological models and is used to link remote sensing and ecological models. The relationship of LAI and vegetation indices (VIs) such as the normalized difference vegetation index (NDVI) or the enhanced vegetation index (EVI) are not as linear relationship especially for deciduous broadleaf forests due to leaf seasonality. To obtain the accurate LAI–VI relationship, in this study, in situ observation data of LAI and VIs were used to rigorously examine the LAI–VI relationship for a deciduous broadleaf forest. Spectral reflectance data from a hemispherical spectroradiometer (HSSR) system were used to calculate in situ NDVI and EVI. LAI was estimated using leaf seasonality (in situ measurement of sample shoots) and litter trap approaches. The LAI–VI relationship was analyzed in two ways: (1) as a one-period relationship using data for the entire year and (2) for two separate periods, leaf-expansion to -saturation and leaf-saturation to -fall. There were nearly linear correlations between both in situ NDVI and EVI and in situ LAI when looking at data for the entire year. When separated into two time periods, the LAI–EVI relationship was clearer than the LAI–NDVI relationship, with a linear regression (r2=0.96) in the leaf-expansion to -saturation period and a clear logarithmic curve (r2=0.97) in the leaf-saturation to -fall period. NDVI and EVI estimates of LAI derived from a moderate resolution imaging spectroradiometer (MODIS) were evaluated by applying the regression equations for both the one- and two-period LAI–VI relationships. The two-period relationships presented better results than the single relationship, because they responded directly to the leaf phenological cycle of the deciduous broadleaf forest. These results indicate that the two-period relationships estimate LAI mo