Leaf spectra predict water potential but not hydraulic vulnerability in Amazonian trees

Introduction:

Hydraulic traits enable predictions about tree mortality and the associated impacts on forest carbon and water cycling under climate change. However, water potential and hydraulics measurements are time consuming. High throughput indicators of hydraulic traits would allow more rapid assessments of forest-scale drought vulnerability. It is well established that leaf reflectance spectra are associated with leaf physiology, but linkages between leaf reflectance spectra and hydraulic traits are relatively unexplored. We expect coordination between branch hydraulics and water potential (e.g. high photosynthetic capacity and stomatal conductance associated with larger vessels and therefore low embolism resistance). In this way, changes in water potential may be associated with not only leaf water content, but also stomatal behavior and investment in photosynthesis.

Question:

Can leaf spectra predict water potential and hydraulic traits for Amazonian trees?

Hypothesis:

Considering the coordination between leaf physiology and hydraulics traits we expect that spectra would also predict leaf water potential and hydraulic vulnerability.

Methods:

We sampled 254 individual Amazonian trees for water potential, measured leaf reflectance for the same leaves, and performed hydraulic percent loss conductivity curves (i.e., P50). We examined relationships between leaf spectra and hydraulic traits through remote sensing indices and partial least square regression (PLSR).

Results and Discussion:

Our preliminary results show a relationship between leaf spectra and leaf water potential, but not a relationship for P50, based on PLSR. The water absorption bands of the leaf spectra showed high variable influence on projection scores, and the Water Band remote sensing index also predicted leaf water potential. This indicates that leaf water content underlies the relationship between leaf spectra and leaf water potential, making upscaling with remote sensing imagery difficult due to atmospheric water absorption. However, wavelengths in the visible domain were also influential, and warrant further investigation.