Projects
NASA ABoVE: The Arctic-Boreal Vulnerability Experiment
Over the past few years, I've been involved in the NASA ABoVE. This region is warming at an alarming rate, two to three times faster than the global average. ABoVE's goal is to understand how vulnerable and resilient these ecosystems are to environmental changes.
During my postdoctoral research at the University of Arizona and my current position at the University of Wisconsin-Madison, I've used satellite remote sensing and Earth system models to study the region. Here are some highlights of my research activities:
- Land use and Land-Cover Change (LULCC) and Surface Energy Balance: Using Landsat and MODIS satellite data, I studied how LULCC affects the surface energy balance (link). I calculated the sensitivity of albedo, land surface temperature, and evapotranspiration to various changes in land cover. Download the manuscript
- Disturbance and Recovery Processes: Using a novel counterfactual analysis in time series (link) and Landsat data, I showed that conventional definitions of recovery (bouncing back to pre-disturbance conditions) significantly underestimate recovery time because they don't consider the influence of climate change. This manuscript is under review.
- Data Assimilation to Improve Land Model Simulations: I contributed to assimilating satellite-derived Leaf Area Index (LAI) data from MODIS into the Community Land Model (CLM) using an Ensemble Adjustment Kalman Filter. This improved CLM's performance in simulating carbon and hydrological cycles. This manuscript is under review.
- Development of a New Vegetation Biomass Dataset: Our team developed a new vegetation biomass dataset for the region using Vegetation Optical Depth (VOD) microwave satellite data. This dataset captures interannual variability in biomass, which is often missing in other datasets. This manuscript is under review.
Skills: Time series analyses, Python (Xarray, Rbeast, Dask), R(CasusalImpact), HPC/HTC
Characterzing Drylands Using Satellite Remote Sensing and Earth System Modeling
The main goal of this project is to create a nitrogen map of shrublands in the western US using hyperspectral data. To achieve this, we use Partial Least Square Regression (PLSR), a machine learning technique. In this model, the spectral bands serve as predictors (features), and the measured percent nitrogen content is the target variable. Here is a brief rationale for choosing PLSR for this project. The nitrogen data were collected from multiple shrubs within 10x10 meter plots in Idaho and California.
Gradually updating projects as I find more time!
Meanwhile, you can check the publications page for more updates.
