Specific leaf area estimation from leaf and canopy reflectance through optimization and validation of vegetation indices

Specific leaf area (SLA), which is defined as the leaf area per unit of dry leaf mass is an important component when assessing functional diversity and plays a key role in ecosystem modeling, linking plant carbon and water cycles as well as quantifying plant physiological processes. However, studies of SLA variation across relevant spatial and temporal scales are lacking. While remote sensing is a fast and efficient approach to quantify vegetation parameters, there are insufficient studies estimating SLA from remotely sensed data. This article aims at finding efficient hyperspectral indices for fast and accurate estimation of SLA from leaf and canopy spectral measurements. Validation of our results with data measured at leaf and canopy scale as well as with experimental datasets simulated using PROSPECT (at leaf level) and INFORM (at canopy level) revealed SLA was predicted accurately by several indices, such as simple ratio and normalized index types. Most of the bands sensitive to SLA selected using these indices were in the 1300–1800 nm spectral region. At leaf level, a ratio index at bands1370 nm, 1615 nm performed strongly (R² = 0.93 and RMSE = 13.66 cm²/g) during cross-validation. The multiband indices with 920 nm, 1675 nm, 1335 nm and 1345 nm, 1675 nm, 1850 nm central wavelengths were also among the top performing indices, with R² > 0.90 for both measured and simulated leaf level data. At canopy level, the soil adjusted ratio vegetation index (with a band setting of 1537 nm, 1543 nm) showed that the hyperspectral data from HySpex airborne imagery accurately estimated SLA (R² = 0.88 and RMSE = 13.30 cm²/g). Generally at canopy level the potential indices for accurate retrieval of SLA from HySpex imagery were two-band indices with soil line information. The discrepancy observed in the performance and band combinations of vegetation indices at leaf and canopy scales can be explained by external factors such as canopy structure, soil background, illumination and sensor configuration which affect the signal when moving from leaf to canopy level. Our findings suggest the availability and suitability of a wide range of existing vegetation indices for assessing SLA accurately and rapidly from remotely sensed data.