Understanding the temporal patterns of leaf traits is critical in determining the seasonality and magnitude of terrestrial carbon, water, and energy fluxes. However, we lack robust and efficient ways to monitor the temporal dynamics of leaf traits. Here we assessed the potential of leaf spectroscopy to predict and monitor leaf traits across their entire life cycle at different forest sites and light environments (sunlit vs. shaded) using a weekly sampled dataset across the entire growing season at two temperate deciduous forests. The dataset includes field measured leaf-level directional-hemispherical reflectance/transmittance together with seven important leaf traits [total chlorophyll (chlorophyll a and b), carotenoids, mass-based nitrogen concentration (Nmass), mass-based carbon concentration (Cmass), and leaf mass per area (LMA)]. All leaf traits varied significantly throughout the growing season, and displayed trait-specific temporal patterns. We used a Partial Least Square Regression (PLSR) modeling approach to estimate leaf traits from spectra, and found that PLSR was able to capture the variability across time, sites, and light environments of all leaf traits investigated (R2= 0.6-0.8 for temporal variability; R2= 0.3-0.7 for cross-site variability; R2= 0.4-0.8 for variability from light environments). We also tested alternative field sampling designs and found that for most leaf traits, biweekly leaf sampling throughout the growing season enabled accurate characterization of the seasonal patterns. Compared with the estimation of foliar pigments, the performance of Nmass, Cmassand LMA PLSR models improved more significantly with sampling frequency. Our results demonstrate that leaf spectra-trait relationships vary with time, and thus tracking the seasonality of leaf traits requires statistical models calibrated with data sampled throughout the growing season. Our results have broad implications for future research that use vegetation spectra to infer leaf traits at different growing stages.

Several studies have shown that satellite retrievals of solar-induced chlorophyll fluorescence (SIF) provide useful information on terrestrial photosynthesis or gross primary production (GPP). Here, we have incorporated equations coupling SIF to photosynthesis in a land surface model, the National Center for Atmospheric Research Community Land Model version 4 (NCAR CLM4), and have demonstrated its use as a diagnostic tool for evaluating the calculation of photosynthesis, a key process in a land surface model that strongly influences the carbon, water, and energy cycles. By comparing forward simulations of SIF, essentially as a byproduct of photosynthesis, in CLM4 with observations of actual SIF, it is possible to check whether the model is accurately representing photosynthesis and the processes cou- pled to it. We provide some background on how SIF is coupled to photosynthesis, describe how SIF was incorporated into CLM4, and demonstrate that our simulated relationship between SIF and GPP values are reasonable when com- pared with satellite (Greenhouse gases Observing SATellite; GOSAT) and in situ flux-tower measurements. CLM4 overestimates SIF in tropical forests, and we show that this error can be corrected by adjusting the maximum carboxyl- ation rate (Vmax) specified for tropical forests in CLM4. Our study confirms that SIF has the potential to improve photo- synthesis simulation and thereby can play a critical role in improving land surface and carbon cycle models.