SPIE Startup Challenge 2015 Founding Partner - JENOPTIK Get updates from SPIE Newsroom
  • Newsroom Home
  • Astronomy
  • Biomedical Optics & Medical Imaging
  • Defense & Security
  • Electronic Imaging & Signal Processing
  • Illumination & Displays
  • Lasers & Sources
  • Micro/Nano Lithography
  • Nanotechnology
  • Optical Design & Engineering
  • Optoelectronics & Communications
  • Remote Sensing
  • Sensing & Measurement
  • Solar & Alternative Energy
  • Sign up for Newsroom E-Alerts
  • Information for:

SPIE Photonics West 2017 | Register Today

SPIE Defense + Commercial Sensing 2017 | Call for Papers

Get Down (loaded) - SPIE Journals OPEN ACCESS


Print PageEmail PageView PDF

Remote Sensing

Monitoring temporal and spatial variations of vegetation phenology from space

Satellite-derived vegetation phenology indicates that earlier greenup and longer growing lengths occur at 45–50°N compared to other latitudes in deciduous broadleaf forest in the eastern US.
27 September 2010, SPIE Newsroom. DOI: 10.1117/2.1201009.003242

Vegetation phenology refers to seasonal plant cycle stages, i.e., greenup, maturity, senescence, and dormancy. Changes in vegetation phenology affect the carbon and water cycles and energy fluxes through photosynthesis and evapotranspiration.1 Vegetation phenology is, therefore, currently one of the main concerns in studies of climate change and carbon-balance estimation in ecosystems. Satellite remote-sensing techniques offer the potential to detect complete and continuous vegetation phenology on landscape scales, which is important for parametrizing ecological and climate models. Vegetation indices (VIs), which use reflective characteristics to differentiate and highlight ecological conditions, allow assessment of vegetation phenology.2

We have developed a new algorithm to detect every single growth or senescence period for landscape-vegetation phenology using VIs. Identification of the growing-season period is essential for studies of vegetation phenology using remote sensing. Many studies have used the moving-average-window method for this purpose. However, the method is sensitive to the moving average time interval. A large time interval may miss natural vegetation changes, while a small interval could result in extremely noisy curves.3 We used Fourier decomposition of VI time series, resulting in a sum of single ascending (growth) or descending (senescence) periods. Our approach enables us to avoid using time intervals and is independent of the temporal resolution of VI data sets. We used the VI curvature-change rate to identify the transition points of different phenological stages.4

We used Moderate Resolution Imaging Spectroradiometer (MODIS) nadir bidirectional-reflection-distribution function-adjusted reflectance data sets to derive time series of two VIs, i.e., the normalized-difference and enhanced VIs (NDVI, EVI) in the period 2001 to 2007. We used these data sets to estimate the greenup and dormancy onsets of vegetation in deciduous broadleaf forests (DBFs) in the eastern US. We compared the onset dates derived from VIs and ground observations of Harvard forest, which record the timing of vegetation development during the growing season and leaf-fall stages during senescence. We found that the greenup and dormancy onset dates were better estimated using NDVI and EVI, respectively. Figures 1 and 2 illustrate, for the DBF over the eastern US in 2003, the greenup and dormancy onsets estimated using NDVI and EVI, respectively. These figures show that the greenup wave progresses northward, while the dormancy wave moves southward. The rate of change is greater for greenup than for dormancy onset.

Figure 1. Greenup onset estimated using the normalized difference vegetation index (VI).

Figure 2. Dormancy onset estimated using the enhanced VI.

Interannual changes of vegetation phenology for DBF vary as a function of latitude. From 30 to 50°N, for every five degrees of latitude, we calculated the average day of greenup and dormancy onset and growing length (dormancy minus greenup onset day) for DBF over the eastern US (see Figures 3, 4, and 5). A slightly early greenup onset is evident at higher latitudes (45–50°N). Except for some interannual variability, no obvious late or early trend is apparent at latitudes between 30 and 45°N. Global warming can contribute to the earlier presence of greenup.5 Our results suggest that the impact of warming is more evident at latitudes higher than 45°N. The interannual variation of dormancy onset is larger than that of greenup onset, and variations are similar at different latitudes. For example, the dormancy onset was late in 2002. An early dormancy onset is obvious at latitudes between 35 and 50°N in 2006, while it varies only slightly between 2003 and 2004. The responses of dormancy onset to external drivers, such as environmental or climate impacts, are similar at different latitudes. Slightly longer growing lengths were measured from 2001 to 2007, which is especially obvious at higher latitudes (45–50°N). Considering that there is no evident late dormancy onset, this longer growing length at higher latitudes is mainly caused by the early greenup onset. Determination of the relative contributions from greenup and dormancy onset to growing length is difficult at other latitudes.

Figure 3. Greenup onset day for deciduous broadleaf forest (DBF) over the eastern US from 2001 to 2007 for different latitude ranges.

Figure 4. Dormancy onset day for DBF over the eastern US from 2001 to 2007.

Figure 5. Growing length for DBF over the eastern US from 2001 to 2007.

Our research offers an efficient approach to estimate vegetation phenology from VI time series. The interannual variation of satellite-derived vegetation phenology suggests that its response to climate warming is more evident at higher latitudes (45–50°N). In the future, we will apply our method to more land-cover types and investigate phenological changes in regional climate settings.

Min Li
George Mason University
Fairfax, VA

Min Li holds a PhD. Her research interests focus on remote sensing of vegetation dynamics, numerical modeling of land-scale vegetation phenology, and evaluating the impact of climate change on land-surface vegetation.

John J. Qu, Xianjun Hao
Environmental Science and Technology Center
George Mason University
Fairfax, VA