Detecting Inter-Annual Variations in the Phenology of Evergreen Conifers Using Long-Term MODIS Vegetation Index Time Series

Laura Ulsig,David Landis,Ivan Mammarella,Caroline Nichol,Alexei Lyapustin,Janne Levula,Albert Porcar-Castell,Karl Huemmrich,Elizabeth Middleton

doi:10.3390/rs9010049

Abstract

Long-term observations of vegetation phenology can be used to monitor the response of terrestrial ecosystems to climate change. Satellite remote sensing provides the most efficient means to observe phenological events through time series analysis of vegetation indices such as the Normalized Difference Vegetation Index (NDVI). This study investigates the potential of a Photochemical Reflectance Index (PRI), which has been linked to vegetation light use efficiency, to improve the accuracy of MODIS-based estimates of phenology in an evergreen conifer forest. Timings of the start and end of the growing season (SGS and EGS) were derived from a 13-year-long time series of PRI and NDVI based on a MAIAC (multi-angle implementation of atmospheric correction) processed MODIS dataset and standard MODIS NDVI product data. The derived dates were validated with phenology estimates from ground-based flux tower measurements of ecosystem productivity. Significant correlations were found between the MAIAC time series and ground-estimated SGS (R2 = 0.36–0.8), which is remarkable since previous studies have found it difficult to observe inter-annual phenological variations in evergreen vegetation from satellite data. The considerably noisier NDVI product could not accurately predict SGS, and EGS could not be derived successfully from any of the time series. While the strongest relationship overall was found between SGS derived from the ground data and PRI, MAIAC NDVI exhibited high correlations with SGS more consistently (R2 > 0.6 in all cases). The results suggest that PRI can serve as an effective indicator of spring seasonal transitions, however, additional work is necessary to confirm the relationships observed and to further explore the usefulness of MODIS PRI for detecting phenology.

Highlights

Developing robust methods for monitoring terrestrial photosynthetic activity is key to improving our understanding of vegetation dynamics and the carbon cycle in the context of climate change.Temporal shifts in vegetation phenological events, such as green-up and senescence, are effective indicators of climate change [1] and can have a significant impact on the annual carbon uptake by forest ecosystems [2,3]
Despite the application of the best index slope extraction (BISE) algorithm to the product data, a considerable amount of noise was still present in the NDVIprod time series and there are more pronounced differences among the four smoothed curves, for fast Fourier transform (FFT) and Savitzky–Golay (Figure 2d)
By linking ground measurements of net ecosystem exchange (NEE) with MODIS data, it was demonstrated that vegetation indices (VIs) time series have significant potential for predicting the onset of the growing season in evergreen conifers

Summary

Introduction

Temporal shifts in vegetation phenological events, such as green-up and senescence, are effective indicators of climate change [1] and can have a significant impact on the annual carbon uptake by forest ecosystems [2,3]. At regional to global scales, satellite remote sensing provides the most efficient means to observe spatial and temporal variations in vegetation productivity and phenology. Phenological parameters can be extracted from per-pixel time series of spectral vegetation indices (VIs) that have been closely linked with vegetation productivity [4,5]. Long-term studies using VI time series have revealed widespread greening occurring in various regions and ecosystems [6,7,8]. Ongoing efforts are being made to establish the accuracy with which satellite imagery can be used to predict timings of seasonal transitions by validating the estimates with ground-based observations [7,9,10,11]

Objectives

Methods

Results

Discussion

Conclusion