Identification Of Plant Phenology And Physiology Traits Based On Phenology Cameras And Solar-induced Chlorophyll Fluorescence

Phenology is the study of the seasonal rhythms of biological phenomena,such as seasonal flowering and leaves color change of plants.These events are partially driven by the changes environmental conditions including temperature,photoperiod and precipitation which may reflect changes in climate.Therefore,plant phenology is a sensitive indicator of how the changing climate affects ecosystems.The shifting plant phenology,have been observed as the advance of spring onset and delay of autumn senescence in the temperate forest,is the common result of how ecosystems respond to climate change.Meanwhile,the shifting plant phenology has potential feedbacks to climate systems,for example,by prolonging the period of growing season and improving the magnitude of photosynthesis and respiration and thus altering the heat balance in climate.Observations of plant seasonality trajectories(i.e.,phenology and physiology)at the appropriate scales are thus important to help us to assess the impact of climate change.However,one of the main problems to study the impact of climate change on plant phenology is the lack of phenology observations at the appropriate spatial and temporal scale.Traditional manual phenological observations approach is time-consuming and only covers limited areas.The development of remote sensing technique and recent popularity of digital repeat photography provided another way to monitor vegetation activities at larger scales.In this study,we conducted field phenology and physiology experiments in temperate deciduous forests to evaluate the relationship between greenness index extracted from repeat canopy photography and physiological traits in the period of seasonal transform,assess the consistencies and discrepancies of phenology estimated from reflectance-based and solar-induced chlorophyll fluorescence observations(SIF,a by-product of photosynthesis),and explore multi-bands SIF in estimating canopy-level photosynthesis.The results indicated that:(1)The spring curves of greenness index were highly synchronous with the expansion of leaf size and the development of leaf area index(LAI)while there was a mismatch between the chlorophyll concentrations and greenness index(the peak of greenness index was 18 days earlier than maximum chlorophyll concentration),but the mismatch could be calibrated by using the Bayesian Change Point analysis(Chapter 3).During the autumn senescence,although it is difficult to detect the start of plant physiological senescence using greenness index because of the mismatch between canopy color and chlorophyll concentration in the initial stage of senescence,greenness index still could detect the timing of canopy phenological transition.The date of starting decline of LAI,defined as the start of senescence,could be mathematically identified from the autumn greenness index pattern by analyzing change points of the greenness index curve;and greenness index highly correlated with LAI after the first changing point when LAI was decreasing(R2=0.88,LAI<2.5 m2 m-2).The end of growing season,defined by the termination of photosynthetic activities and detected by the Bayesian Change Point analysis,was two weeks earlier than the end of greenness index curve decline(Chapter 4).(2)Based on the performance of optical images,reflectance-based vegetation indices and solar-induced chlorophyll fluorescence(SIF)datasets with various spatial and temporal resolutions in monitoring the GPP(gross primary production)-based phenology,all these data can serve as good indicators of phenological metrics in the spring that are derived from GPP time series.However,the autumn phenological metrics derived from all reflectance-based datasets are later than the tower-based GPP estimates.This is because the reflectance-based observations estimate phenology by tracking physiological properties including leaf area index(LAI)and leaf chlorophyll content(Chl),which does not reflect instantaneous changes in phenophase transitions,and thus the estimated fall phenological events may be later than GPP-based phenology.In contrast,we find that SIF has a good potential to track seasonal transition of photosynthetically activities in both spring and fall seasons.The advantage of SIF in estimating the GPP-based phenology lies in its inherent link to photosynthesis activities such that SIF can respond quickly to all factors regulating phenological events.Despite uncertainties in phenological metrics estimated from current spacebome SIF observations due to their coarse spatial and temporal resolutions,dates of the middle spring and autumn-the two most important metrics,can still be reasonably estimated from satellite SIF(Chapter 5).(3)Using high-frequency measurements of carbon flux and SIF emissions at the atmospheric absorption bands,we reconstructed full SIF spectrum from SIF signals at the absorption bands,and then analyzed correlations between the GPP and the selected single SIF bands and their combinations with using both the linear regression(LR)and Gaussian processes regression(GPR).The results indicate that the red SIF emission(640-700 nm)shows low correlation with GPP due to the strong reabsorption of red SIF emissions by leaf chlorophyll;the individual bands in near-infrared area(at 720 nm,740 nm,and 761 nm)can determine about 60%and 62%variance in GPP with hourly scale by LR and GPR,respectively,and the combination of those SIF bands can provide better predictive power by explaining 66%and 76%variance in GPP with hourly scale by LR and GPR,respectively;solar radiation saturation,fraction of direct solar radiation,air temperature and leaf area index may have negative impacts on the SIF-GPP correlations when they are beyond optimal thresholds;and the SIF-GPP correlations are not sensitive to the observation hour,but temporal aggregation(daily scale)tends to enhance these correlations:the daily combination of SIF bands(at 687 nm,720 nm,and 761 nm)can account for 80%and 93%of variance in daily GPP by LR and GPR,respectively(Chapter 6).In summary,we found that(1)camera-based phenology observation could be used as a proxy of the development of LAI and chlorophyll concentration based on the changing point analysis and can effectively quantify the dynamic pattern at the start of senescence(with declining LAI)and the end of senescence(when photosynthetic activities terminated)in the deciduous forest;(2)SIF provides a better way to monitor GPP-based phenological metrics,which suggests the need of establishing a SIF-based network;and(3)multi-bands SIF has a stronger capacity in predicting plant GPP than traditionally used signal band SIF data.