An innovative way to monitor leaf age demographics in a tropical evergreen forest
Leaf age is an important characteristic for controlling ecological processes of plants and is associated with the life trajectory of leaf physical, chemical, and physiological properties. Understanding how leaf traits change over time will allow more accurate predictions of plant growth patterns, and a more comprehensive understanding of life evolution of leaves. Tagging leaves is a common method for accurately monitoring leaf age, but this is tedious and time consuming. Therefore, a fast and efficient technique for monitoring leaf age is desired.
We explore the following questions in this study by using the data collected over one full year in an evergreen tropical forest at Tapajos km67 site (near Santarem, Brazil), including (Q1) Can leaf spectra efficiently and accurately predict leaf age? (Q2) Which spectral wavelengths are most associated with changes in leaf age? Are these age-sensitive spectral regions similar to that of other key traits, e.g. specific leaf area (SLA), water percentage? (Q3) Is the proposed spectral-age approach generalizable across species, regardless of interspecific variation or intraspecific variation induced by microhabitats (i.e. radiation regime)?
Our preliminary results suggest that leaf spectra can precisely predict leaf age across 10 tropical species studied (R2=0.90; RMSE=27.3 days) spanning the age range from 20 days to 365 days, which is better than traditional leaf trait approach (either by SLA or water percentage). The spectra-age model regression coefficients demonstrate that three spectral domains are highly correlated with leaf age, including the visible domain (centers in green peak ~550 nm), red edge (centers ~720nm) and water related spectra domain (i.e. ~1420 nm and 1950nm). The age-sensitive spectral domains are also in a good agreement with that of SLA and leaf water percentage, providing a potential underlying mechanism as to spectra-age relationship. We also test the generalizability of this model within and across plant communities.
Last, we will also discuss the broad impact of this work, including how this study could (1) complement conventional plant functional trait studies, as leaf age by itself is an important phenological trait for plants; (2) improve analyses of leaf trait correlations by allowing for normalizing traits for the leaf age effect; (3) provide a conceptual remotely sensed approach to predicting and monitoring life evolution of leaves, from the development perspective of leaf physical, chemical, and physiological properties; (4) test the hypothesis as to the fundamental mechanisms regulating the life-time resource investment and return in leaves.