We study terrestrial ecosystems and how they respond to climate variability and long-term change, as well as related feedbacks to the atmosphere through ecosystem carbon, water, and energy exchanges, with an emphasis on extreme climate events that have the largest ecological and societal impacts. Our group combines ecosystem and atmospheric observations, satellite remote sensing, and numerical models. We use methods from diverse disciplines, including atmosheric science, biogeochemistry, earth system models, and machine learning.
Theme 1: Carbon-Climate Feedbacks
Carbon dioxide and methane are the leading anthropogenic drivers of global climate change, yet significant uncertainties persist in their sources, sinks, and atmospheric lifetimes. Our group addresses these uncertainties through two complementary strategies: multi-tracer atmospheric inverse modeling that exploits chemical linkages among methane, formaldehyde, and carbon monoxide through OH and integration of "bottom-up" and "top-down" observational systems to reconcile carbon budgets from multiple angles.
Multi-tracer atmospheric inversions of methane and carbon monoxide
Most inversion frameworks treat methane in isolation; we developed a multi-species inversion framework that simultaneously assimilates CH₄, CO, and formaldehyde (CH₄-CO-HCHO) to constrain both sources and sinks of methane. Tropical wetlands and East Asian anthropogenic emissions are the leading contributors to the renewed rise in atmospheric methane after 2010 (Yin et al., 2021, ACP). The 2020 jump in methane growth rate was driven roughly equally by increased wetland emissions and a weakened OH sink due to reduced NOₓ emissions during COVID lockdowns (Peng et al., 2022, Nature). The sustained high methane growth rate in 2021 is dominated by emissions from tropical and boreal inundated areas, consistent with rising groundwater storage and regional warming (Lin et al., 2024, Nature Communications).
Our earlier work also investigated decadal trends in global carbon moxide and attributed CO decline from the early 2000s to the late 2010s to improved combustion efficiency and reduced wildfire emissions in the dry tropics (Yin et al., 2015, ACP; Zheng et al., 2019, ESSD).
Stratosphere-troposphere exchange and methane variability
Year-to-year fluctuations in the atmospheric methane growth rate are typically attributed to changes in surface emissions. We estimated that tratosphere-troposphere exchange accounts for approximately 40% of interannual variability in the methane growth rate, meaning a substantial portion of what has been attributed to changing emissions may instead reflect atmospheric dynamics (Yin et al., in submission).
Crop Yield and Climate Change
Projecting agricultural productivity under continued climate change requires disentangling three drivers that have co-varied for decades: rising atmospheric CO₂, a shifting climate, and accumulating agronomic improvements. This identification problem has been a persistent obstacle to constraining climate-impact projections and benchmarking process-based crop models, with downstream consequences for food security assessments and the evaluation of mitigation pathways. Our work addresses this identification problem using a natural experiment embedded in US agriculture. Soybean (C3) and maize (C4) are grown in rotation across the same counties, share equipment, inputs, and management improvements, and experience common environmental trends including air-quality changes, but they respond differently to rising CO₂ because of their contrasting photosynthetic pathways. Using long-term CO₂ reanalysis products and compared yield responses of C3 crops (soybean) and C4 crops (corn), we identified much larger CO₂ fertilization efffect than previous field experiments with step-wise change in CO₂ concentration and fixed cultivar, with sensitivity to moisture gradient consistent with theoretical model prediction (Wang et al., in review). 
However, extreme climate events may result in significant impacts on crop yields. In spring 2019, record flooding delayed planting across the U.S. Corn Belt. We combined satellite SIF with atmospheric CO₂ observations to quantify the resulting carbon and yield anomalies in near real time. Two independent observational systems (SIF and atmospheric CO₂) detected roughly 15% crop yield losses more than six months before USDA census data confirmed them, demonstrating that satellite carbon observations can provide near-real-time agricultural monitoring across continental scales (Yin et al., AGU Advances 2020).
Read more: | Caltech news | interactive map
Bottom-up and top-down integration via OCO
Atmospheric inversions and ecosystem flux measurements offer independent windows onto the same carbon cycle, yet they often disagree. Through two NASA Orbiting Carbon Observatory (OCO) Science Team projects, we are systematically reconciling these views by combining satellite SIF, column CO₂, and the global eddy covariance network. Quantifying how spatial sampling bias in flux tower networks affects upscaled global carbon estimates (He et al., in prep); assessing how anthropogenic respiration (human and livestock metabolism) reshapes regional carbon budgets in trade-connected systems. (Wang et al., in prep).
Theme 2: Wildfire Dynamics and Impacts
Wildfires are reshaping ecosystems, communities, and the global carbon cycle at scales unprecedented in the historical record. Our group investigates how climate, vegetation, and human activity interact to drive fire behavior, and how fires in turn cascade through air quality, infrastructure, and the carbon-climate system. We combine satellite fire and vegetation observations, atmospheric inverse modeling of fire emissions, and geospatial analysis of fire-society interactions across scales from individual events to global fire regimes.
Wildland-urban interface (WUI) wildfire risk: a scaling-law perspective
California wildfires have grown in size, intensity, and proximity to populated areas. Using four decades of fire records, we tested whether wildfires near urban areas follow the same size-scaling relationships as fires in remote landscapes. Urban proximity amplifies wildfire size scaling, with extreme wildfires disproportionately reach the wildland-urban interface, with implications for infrastructure vulnerability, fire suppression strategy, and where new development should and should not occur (Liu et al., in review).
Post-fire vegetation recovery in California
Whether and how landscapes recover after fire depends on land cover, fire intensity, and prior fire history, but these factors have rarely been disentangled at scale. Using satellite vegetation indices across thousands of California fire scars, we are characterizing recovery trajectories and their drivers. Undergraduate senior thesis project supported by the NYU Dean's Undergraduate Research Fund (Torres et al., in prep).
Fire emissions constrained by atmospheric inversions
Bottom-up fire emission inventories disagree substantially with each other. By inverting atmospheric carbon monoxide observations from satellites and surface networks, we can constrain fire emissions from the top down and identify which regions are over- or underestimated in existing inventories. Fire activity in dry tropical ecosystems declined over the early 21st century, a quiet but transient enhancement of the global land carbon sink due to not only direct emission reduction but also fire legacy effects.
Key publications. Yin et al., Nature Communications (2020); Yin et al., GRL (2016).
Tropical peatland fire-carbon dynamics with NISAR
Peat fires in Southeast Asia and the Amazon release vast amounts of carbon but are poorly quantified because the depth of burning is invisible to optical satellites. The upcoming NASA-ISRO Synthetic Aperture Radar (NISAR) mission offers unprecedented ability to measure surface elevation changes and detect smoldering combustion below the canopy. Our proposed project will use NISAR observations to quantify peatland burning depth, fuel consumption, and post-fire carbon trajectories.
Theme 3: Urban Climate Extremes
Cities concentrate both climate risk and the capacity to respond to it. Yet within any given city, heat exposure, green space access, and adaptation resources are distributed deeply unequally. Our group investigates how urban form and policy shape who bears the burden of rising temperatures — and how that information can guide more equitable adaptation. We combine high-resolution satellite thermal imagery (ECOSTRESS, Landsat), urban land cover data, census-tract demographics, and migration records to characterize intra-city climate disparities and their human consequences.
Unequal heatwave exposure in Los Angeles
During the 2020 Los Angeles heatwave, we used ECOSTRESS satellite thermal observations to map exposure at sub-neighborhood resolution and combined this with census data on race, income, and green space. Heatwave exposure varied dramatically across LA neighborhoods, and the disparity was driven primarily by unequal distribution of green space rather than other demographic or built-environment factors. Lower-income neighborhoods experienced temperatures up to 7°C higher than nearby wealthier neighborhoods during the same heatwave (Yin et al., 2023, Science Advances).
Cooling where it counts: a spatial framework for urban heat mitigation
Tree planting and other cooling investments are often distributed by ease of implementation rather than where they would do the most good. We developed a spatial framework that combines thermal observations, population vulnerability, and intervention cost to identify where each dollar of cooling investment yields the greatest reduction in human heat exposure. Manuscript in revision.
Intra-city heat disparities across major U.S. cities
The LA finding raised the question of whether similar disparities exist nationwide. Using satellite thermal observations and high-resolution urban land cover data across the largest U.S. cities, we are quantifying intra-city heat disparities and identifying the urban form factors that drive them. Impervious cover, more than green space deficit, explains intra-city heat disparities across U.S. cities — suggesting that reducing heat-retaining surfaces may be as important as adding vegetation. Older U.S. cities show larger thermal disparities, likely reflecting the legacy of historical land use and infrastructure decisions. Manuscript in submission.
Moving into harm's way: domestic migration and climate risk
Domestic U.S. migration patterns over the past two decades have moved millions of people toward Sun Belt cities, many of which are projected to experience the largest increases in heat, drought, and wildfire risk. We are mapping the mismatch between current migration patterns and future climate livability, asking where people are moving to versus where the climate is becoming most hostile. Supported by the NYU Climate Change Initiative Seed Grant and a forthcoming Burroughs Wellcome Fund Climate Change and Human Health Seed Grant.