Retrieval of leaf area index in different plant species using thermal hyperspectral data

Leaf area index (LAI) is an important variable of terrestrial ecosystems because it is strongly correlated with many ecosystem processes (e.g., water balance and evapotranspiration) and directly related to the plant energy balance and gas exchanges. Although LAI has been accurately predicted using visible and short-wave infrared hyperspectral data (0.3–2.5 μm), LAI estimation using thermal infrared (TIR, 8–14 μm) measurements has not yet been addressed. The novel approach of this study is to evaluate the retrieval of LAI using TIR hyperspectral data. The leaf area indices were destructively acquired for four plant species: Azalea japonica, Buxussempervirens, Euonymus japonicus, and Ficus benjamina. Canopy emissivity spectral measurements were obtained under controlled laboratory conditions using a MIDAC (M4401-F) spectrometer. The LAI retrieval was assessed using a partial least squares regression (PLSR), artificial neural networks (ANNs), and narrow band indices calculated from all possible combinations of waveband pairs for three vegetation indices including simple difference, simple ratio, and normalized difference. ANNs retrieved LAI more accurately than PLSR and vegetation indices (0.67 < R² < 0.95 versus 11.54% < RMSE_cv < 31.23%). The accuracy of LAI retrieval did not differ significantly between the vegetation indices. The results revealed that wavebands from the 8–12 μm region contain relevant information for LAI estimation, irrespective of the chosen vegetation index. Moreover, they demonstrated that LAI may be successfully predicted from TIR hyperspectral data, even for higher values of LAI (LAI⩾5.5). The study showed the significance of using PLSR and ANNs as multivariate methods compared to the univariate technique (e.g., narrow band vegetation indices) when hyperspectral thermal data is utilized. We thus demonstrated for the first time the potential of hyperspectral thermal data to accurately retrieve LAI.