Previous research projects
Land Surface Phenology
1. Identify Land Surface Phenology using change point estimation
Dramatic ecosystem responses to global climate change have been reported around the world. Land surface phenology (LSP) derived from satellite imagery provides a powerful tool to detect vegetation phenological responses to climate change at the landscape scale. However, most of current methods in determining LSP have a large and unquantified uncertainty in estimated phenological dates due to the relatively coarse temporal resolution and methodological limitations. It is essential to narrow the temporal uncertainty of estimated phenological transitions, an important bio-indicator of climate change.
In this study, we developed a new LSP estimation method using linear change point models to determine four phenological transitions using twice-daily Moderate Resolution Imaging Spectroradiometer (MODIS) Enhanced Vegetation Index (EVI) from 2000 to 2015. We evaluated the approach using long-term phenological ground observations and compare performance of four LSP estimations generated from two data sources (i.e. 8-day and twice daily EVI time series) and two methods (i.e. double logistic and change point estimation). We found that the LSP generated from change point estimation with twice daily EVI time series had the highest accuracy (i.e. lower Root Mean Square Error (RMSE), mean bias, and Mean Absolute Error (MAE)) for both spring and fall phenology evaluated by Harvard Forest phenology observations and a large citizen science database of phenological observations from the National Phenology Network.
2. Land surface phenology and climate variation - green-up of deciduous forest communities in northeastern North America
Regional mechanistic plant phenology models are needed for predictions of both temporal and spatial responses at large scales. Although different species are known to show different phenological responses spatially and temporally, community or landscape level phenological responses to environmental variation are rarely studied.
I investigated how variation in weather and climate relate to the known physiological chilling and heating requirements of deciduous forest tree species that result in spring green-up in New England, USA over nine years (2001-2009) (Xie et al. 2015 Landscape Ecology). I applied Bayesian survival models to remotely sensed phenology data from MODIS satellite imagery. Greater oak dominance had later green-up, while sites with more birch tended to have earlier green-up dates. Future predictions (2046-2065) based on climate change scenarios suggested that higher heating and chilling accumulations will lead to earlier green-up (8-48 days). However, in coastal areas green-up may be delayed due to reduced chilling accumulation. This study provides an innovative statistical modeling method combining plant physiological mechanisms, topographic spatial heterogeneity, and species composition to predict how land surface phenology responds to climate and weather variation and in turn allows future projections.
3. Land surface phenology and climate variation - dormancy of deciduous forest communities in northeastern North America
Currently our understanding of the mechanisms controlling plant fall phenology is surprisingly poor. Short day length and low temperature are commonly known as main triggers for leaf senescence and dormancy, but other factors such as frost, drought, heat and precipitation were also suggested to have effects on fall phenology, however were rarely examined or used in phenology models.
I used remotely sensed MODIS satellite phenology data from 2001 to 2012, and identified and quantified significant effects of a suite of environmental factors on the timing of fall dormancy of deciduous forest communities in New England, USA (Xie et al. 2015 PNAS). I also made future predictions of fall dormancy in two eco-regions for two periods (2041-2050 and 2090-2099) based on climate change scenario projections. The findings suggested that changes in frost and moisture conditions as well as extreme weather events (e.g. drought- and heat-stress, and flooding) should be considered in future predictions of autumn phenology for temperate deciduous forests. While climate change brings more frequent and intensity of extreme weather events, the phenological responses of deciduous forests to these events will be complicated.
Species-specific Autumn Phenology
Fall phenological response of deciduous trees to environmental changes
Shifts in autumnal phenology (leaf coloration and leaf drop) in temperate regions due to climate change bring substantial impacts on community and ecosystem processes (e.g. altered C and N cycling and biotic interactions) and the fall foliage ecotourism industry. However, our knowledge of the environmental control of autumn phenology has changed little over the past several decades. Using ground-based phenology observations in New England USA, we found several important weather/climatic factors to significantly affect autumn phenology of 12 dominant deciduous tree species. These patterns were revealed using linear mixed effects models. The significant weather/climate factors included not only autumn chill and frost, but also heat and drought stress plus precipitation. Species-specific sensitivities to those factors were also quantified for timing of leaf coloration and leaf drop. Positive relationship between spring and autumn phenology was confirmed by this study. Future changes in autumn phenology was predicted for all individual tree species from 2015 to 2099 based on future climate projections along with projected spring phenological changes. The findings suggested climatic stresses are critical factors affecting autumn phenology, and divergent phenological responses by different tree species are important in predicting future impacts from climate change on forest community and ecosystem patterns and processes.
Phenology from Near-surface Remote Sensing
Species-specific leaf phenology captured by digital cameras
Plant leaf phenology is typically observed either via ground-based visual observations on individuals or via remote sensing of land surface vegetation. But the challenge exists in interpreting phenological information from both data sources, collected at different spatial scales using different observational protocols. As an intermediate step, digital cameras are employed, that span an area with enough spatial resolution to identify temporal changes in individual deciduous tree species canopies with continuous observations. But it was unknown how these camera images relate to field observations, and how the metrics from those images provide comparable species-specific phenological responses to the environment. We set a suite of digital time-lapse cameras to take continuous photos of deciduous tree canopies in Connecticut from 2012 to 2014. Color indices were derived from three color channels from the images, focusing on green and red, and phenological transition dates were determined from time series of color indices for each tree canopy at each site. Comparisons between image derived dates and observed phenological dates showed that both green and red color indices could be matched to ground observations, and red color indices had better performance in match autumn phenology across our group of 8 dominant tree species.
This is a time-lapse video made from pictures took at one of my sites showing canopy change during a growing season.