We used remotely-sensed and ground-based datasets and model results embedding snowmelt timing in phenology at seven tundra flux tower sites in Alaska during 2001–2018, showing that the carbon response to early greenup or delayed snowmelt varies greatly depending upon local climatic limits. Increases in net ecosystem productivity (NEP) due to early greenup were amplified at the higher latitudes where temperature and water strongly colimit vegetation growth, while NEP decreases due to delayed snowmelt were alleviated by a relief of water stress. Given the high likelihood of more frequent delayed snowmelt at higher latitudes, this study highlights the importance of understanding the role of snowmelt timing in vegetation growth and terrestrial carbon cycles across warming Arctic ecosystems.

We calibrated leaf phenology models based on two decades of data (1992–2011) and integrated them into a watershed ecohydrological model (RHESSys) to isolate the effects of changing vegetation timing on water cycles of a temperate mixed forest (Harvard Forest) in the northeastern United States. Over the past two decades, spring greenup has started 10.3 days earlier and autumn leaf-fall has been delayed by 6.0 days. We showed that the earlier greenup, in particular, caused a significant increase in annual carbon uptake and evapotranspiration. This, in turn, reduced the amount of water in streams during both the growing season and the following dormant season, as plants used up soil water earlier in the year. This study highlights the importance of understanding vegetation response to ongoing climate change in order to predict the future hydrological nonstationarity in this region.

We used an ecosystem demography model (ED2) to project ecological changes in a temperate-subtropical mixed forest on Jeju Island, South Korea, up to the year 2099. We analyzed how different climate change scenarios and model parameters contributed to uncertainty in the projections. The model predicts that as the climate warms, subtropical tree species will become more dominant, increasing from 30% to 41% of the forest by the end of the century. This shift leads to a longer growing season and greater leaf area, causing the forest to absorb more carbon (Net Ecosystem Productivity). Our result highlights that uncertainty in these projections comes from different sources: uncertainty in future leaf area was mostly due to the model's internal parameters, while uncertainty in carbon uptake was equally influenced by the choice of climate model and the emissions scenario.

This study was conducted in New England, U.S., in a forest area heavily infested by the hemlock woolly adelgid (HWA). We compared a hemlock-dominated catchment to a nearby catchment with less hemlock cover and used an ecohydrological model (RHESSys) to quantify the changes. The HWA infestation, which has been killing eastern hemlock trees, caused a 24–37% decrease in peak growing season evapotranspiration (water used by trees). As a result, the annual water yield (streamflow) from the affected area increased by as much as 15.6% by 2014, demonstrating that the insect-driven tree decline is significantly altering the region's water cycles. 

We analyzed a decade (2012–2021) of satellite data from the highly urbanized area of Seoul, South Korea, and its surroundings to see how urban conditions affect plant phenology. Urban factors (including urban fractions and artificial lights) are responsible for both advancing the start of the growing season (SOS) and delaying its end (EOS). We also found that changes in the SOS were more influenced by warmer daytime temperatures in more urbanized areas with higher artificial lights, while the delay in the EOS was more strongly linked to warmer nighttime temperatures in brighter areas during nighttime. Our findings indicate that vegetation’s response to rising temperatures will become more dynamic under future climate change. 

We developed a deep learning-based surrogate modeling framework to optimize urban roofing strategies for climate risk mitigation. We first implemented a process-based climate model (WRF-UCM) to generate the data regarding the effects of different roofing types (cool and green roofs) on heat stress, flash floods, and wind, over Seoul and its surroundings by the end of this century. We then used this data to train a high-performing deep learning algorithm (Multi-ResUNet) to find the most effective and cost-efficient options. We show that the best-case scenario involved covering 95.9% of the city with cool roofs, which reduced heat stress by over 50% in more than a third of the city's regions. A more balanced option with 60.2% cool roof coverage still offered significant heat reduction. Both scenarios had a minimal effect on flash floods but did alter wind circulation. The study emphasizes that the specific spatial arrangement of the roofs is critical to maximizing their climate benefits, and also demonstrates that the proposed surrogate modeling framework would be highly effective for use in a diverse range of urban contexts, advancing global efforts to mitigate urban climate risks.