Click here to jump to the General Flux Science Session on October 22
Click here to jump to the Data Processing, Modeling and Upscaling Session on October 23
Click here to jump to the Remote Sensing Session on October 23
Click here to jump to the MexFlux talk on October 24
Click here to jump to the Urban and Managed Lands Session on October 24
Download the booklet with the poster abstracts here.
Wednesday, October 22
Opening Session: General Flux Science
Invited Speaker: Lucas Zepetello, UC Berkeley
“Water stress causes emergent relationships between surface fluxes, temperature, and VPD across ecosystems”
Short Talk: Julia Green, University of Arizona
“Underestimation of Time Integrated Water Limitation Impacts on Terrestrial Ecosystems”
Since the 1960s, drought indices have been used to study the effects of water limitation on ecosystems. However, with anthropogenic warming and changes in global precipitation patterns, prevalence of high-intensity rainfall events and flash droughts at local scales raise uncertainties about whether monthly drought indices can be reliably used to capture ecosystems in water-limited states. While most drought indices provide monthly values integrating water deficits over a defined time window, plants respond to water availability on much shorter time scales, and monthly integration may mask these responses. Here, we systematically analyzed 12,447 drought months based on monthly drought indices for 408 eddy covariance sites located across biomes and climate zones. Daily data revealed that in months classified as climate anomaly drought, soil moisture conditions were not limiting to photosynthesis on roughly half of those days, leading monthly drought indices to misrepresent the effects of water limitation on ecosystem carbon, water and energy fluxes. These findings highlight the need for higher-resolution approaches to quantify the actual impacts of water limitation on ecosystem function. More accurate representations are essential for improving predictions of biosphere–climate interactions, particularly in an era of increasing drought frequency and intensity.
Short Talk: Anam Khan, Northern Arizona University
“The role of hydrologic intensification in ecosystem recovery from flash drought induced acute water stress across the western United States”
Ecosystems experience rapid declines in soil moisture, flash droughts, which can create acute water stress. Prolonged drought can leave ‘legacies’, delayed recovery of plant growth, but less is known about the recovery dynamics of ecosystem fluxes following flash droughts. Many ecosystem processes also exhibit memory of past climatic phenomena (e.g., pluvials), complicating attribution of flux anomalies to specific events like a single drought. Here, we ask: do flash droughts leave a legacy on the net ecosystem exchange of CO2 (NEE) and evapotranspiration (ET) that cannot be explained by climatic memory? To address this, we first identified flash droughts based on rapid soil moisture declines across the western USA at 12 AmeriFlux sites. Then, we fit memory-aware reference models of NEE and ET that include current and antecedent covariates. Finally, we used prediction errors during the recovery phase of flash droughts to quantify NEE and ET legacies. About 50% of flash droughts developed within 6 – 23 days and persisted for 30 – 93 days. 46% of NEE legacies showed the ecosystem was a higher net source of CO2 compared to the reference model and 9% showed that the ecosystem was a net source of CO2 while the reference model suggested a net sink. 43% of ET legacies showed that ET was reduced compared to the reference model. Many flash droughts ended because of large rain events and an unusually large soil moisture recharge that exceeded the 0.8 quantile for the given time of year. This resulted in ecosystems experiencing unusually wet conditions immediately following dry conditions, often within 30 days, reflective of recently reported “hydroclimate volatility”. The highest absolute NEE and ET legacies occurred during these whiplash dynamics in soil moisture. Our findings show that rapidly evolving water stress conditions and increasing hydrologic intensity increase variability and decrease predictability of ecosystem fluxes.
Short Talk: Alejandro Castellanos, Universidad de Sonora
“Upscaling ecophysiological processes in natural and managed Sonoran Desert Drylands: Multi-scale Observations at La Colorada MexFlux Sites (MXCH)”
Fifteen years ago, we initiated a comparative study of natural and managed dryland ecosystems at El Churi experimental Ranch in La Colorada, Sonora, aiming to understand ecosystem function better. At the site, the longest eddy-flux ecosystem comparison in Mexico has been running uninterrupted (except for technical issues). Here also, the first Phenocam type I site in Mexico has been instrumented. Our aim with this presentation is to foster multi-institutional and multi-national exchanges and collaborations to potentiate and ensure monitoring in this long-term observational site.
El Churi is a dryland Sonoran Desert site located within the neotropical and neartic biogeographical ecotone in Central Sonora, dominated by desert and subtropical plant species. It has a short growing season (~90 days) that could extend because of a biseasonal winter period of rains. Changes in dominance in its managed ecosystem are because of the transformation of shrubland to a savanna-like ecosystem, with exotic buffegrass.
Over the years, the dominant species at the site have been measured for ecophysiologic, hydraulic, and stoichiometric traits, as well as community dominance to complement eddy-flux monitoring. Phenological monitoring, along with unmanaged aerial and satellite observational data, has also been gathered with the long-term objective of up- and down-scaling ecosystem processes.
Trait-based measurements and abundance metrics using aerial imagery are a key approach to community-weighted measurements, still under-explored in drylands. This line of research is particularly relevant given that climate change can modify resource availability and induce the loss of susceptible species, with changes in ecosystem productivity and resilience.
A synthesis of the research at El Churi is underway, with numerous articles and dissertations produced over the years from the site, providing a solid context for future ecophysiological, stoichiometric, phenological, and eddy fluxes biogeochemical research.
Thursday, October 23
Session: Approaches in data processing, modeling and upscaling
Invited Speaker: Yujie Liu, Northern Arizona University
“Evaluating the continuity of NEON and AmeriFlux data streams recorded at collocated sites from tundra to subtropics”
Long-term time series are essential for detecting ecological changes and predicting future trends. However, NEON datasets are < 10 y in length. Integrating new NEON data with pre-existing AmeriFlux data would provide a robust understanding of ecosystem changes over 15-25 year time scales. We evaluated how comparable NEON’s measurements are with data from pre-existing, nearby AmeriFlux sites. A case study at Bartlett, NH suggests strong agreement in meteorology and phenology and moderate agreement in half-hourly CO₂ and energy fluxes with substantial divergence in annual carbon and water flux estimates.
To evaluate whether similar patterns emerge at other sites, we expanded our analysis to a dozen paired sites from tundra to subtropics. We developed a “continuity scorecard” to quantitatively assess the level of agreement between NEON and AmeriFlux measurements for meteorological, flux, and phenology variables. We evaluated 5 metrics: slope deviation, mean distance from 1:1 line, RMSE, skewness and concordance correlation coefficient.
Meteorology shows good or fair agreement between NEON and AmeriFlux, with most sites achieving “passing” grades. The Bray–Curtis dissimilarity in footprint-weighted EVI and percentage of land cover types varied considerably for different paired sites. However, agreement of fluxes was relatively poor even at sites where land cover and the EVI were highly similar across the flux footprints. These results suggest that – in addition to distance between paired sites – seemingly minor differences in infrastructure (tower design and height) and instrument setup (particularly for flux measurements) may result in substantial differences in measured meteorological and flux data, and point to large and under-appreciated uncertainties associated with the representativeness of any one tower relative to the broader landscape.
Short Talk: Gil Bohrer, Ohio State University
“Observations and modeling of species-specific hydraulic response traits – using long-term tree-level measurements in US-UMB”
Species‐specific hydraulic traits play an important role in scaled ecosystem responses to water stress. These responses shape the instantaneous and short-term (seconds-days) evaporation and primary production of forest plots and scale to impact annual carbon sequestration. However, representation of biodiverse forest composition and its effects on water stress response remains a challenge in land surface models. We used a 21-year dataset of annual stem growth increments in the footprint of the US-UMB flux site in Northern Michigan to analyze the sensitivity of responses of individual species to environmental characteristic of water limitation. We found that annual or seasonal gross primary productivity (GPP), instantaneous or lagged, is poorly correlated with annual increment of stem diameter for almost all species. We show that annual stem growth is influenced by VPD and atmospheric water demand or light availability in some species, by precipitation (or drought indices) in others, and by soil moisture limitations in some deep-rooted species. These indicate differences in species-specific water relations and carbon allocation strategies that derive from the trees’ variable “perception” of water stress. We introduce FETCH4, a multispecies canopy‐level version of the tree hydrodynamic model FETCH. FETCH4 simulates water transport through the soil, roots, and stem as porous media flow. Stomatal conductance is controlled by xylem water potential, which is resolved along the vertical dimension. We demonstrate the model’s performance in US-UMB with species of contrasting hydraulic strategies. We optimize the model’s species‐specific hydraulic parameters using a Bayesian optimization framework incorporating long-term sapflow measurements. The model results were able to capture emergent hydraulic traits, which characterize differences between species traits of drought sensitivity. Using FETCH4 in combination with available observations can provide unique insights about difficult-to-measure hydraulic traits and plant hydrodynamic behaviors.
Short Talk: Yang Gu, Boston University
“Improving NEE and LE Estimates Across North America: A Hybrid Data Assimilation + Machine Learning Approach to Gap-Filling, Model Bias Correction, and Downscaling”
Accurately estimating C fluxes at continental scales is essential for understanding and monitoring terrestrial C dynamics. Terrestrial Biosphere Models (TBMs) are widely used for estimating C fluxes from local to global scales, but their predictions are limited by biases in the modeling process, observations, and unknown ecological mechanisms. Existing multi-model ensembles remain highly uncertain and show large discrepancies among models. To improve C estimates, a hybrid data assimilation framework was developed to address the uncertainties in Net Ecosystem Exchange (NEE) and Latent Heat (LE) across North America by combining remote sensing and field measured data (ERA5, MODIS, Landsat, GEDI, SMAP, SoilGrids, EC towers), a process-based land model (SIPNET), a Bayesian state data assimilation (SDA) system (PEcAn), and multiple machine learning models. An XGBoost-based model was applied to gap-fill missing flux data for 193 AmeriFlux eddy covariance sites across North America, reducing interpolation errors over the traditional MDS method. Using the gap-filled observations, we then built an XGBoost bias correction model to predict residuals between SDA ensemble outputs and observed NEE and LE. The bias correction model significantly improved accuracy and reduced uncertainty of the original output. After applying this bias correction from 2012-2024 across all 8000 locations in the PEcAn SDA, a Random Forest–based emulator approach was then used to generate 1km maps of monthly NEE and LE dynamics. The results narrow flux uncertainties while providing insights into the spatial and temporal patterns of NEE and LE at finer spatial resolutions than existing multi-model ensembles, while simultaneously reconciling C pool and flux information at a continental scale.
Short Talk: Qing Zhu, Lawrence Berkeley National Laboratory
“Reconstruct multi-decadal wetland CH4 fluxes with knowledge-guided machine learning”
Understanding long-term wetland methane (CH4) dynamics is critical for constraining the global CH4 budget and its environmental feedbacks. In this study, we reconstruct multi-decadal (2000–2024) daily CH4 fluxes from wetlands using a novel knowledge-guided machine learning framework. Leveraging eddy covariance (EC) measurements from AmeriFlux sites with over eight years of continuous CH4 observations, our approach robustly extrapolates site-level fluxes beyond the measurement period by embedding domain knowledge into model design and training. The reconstructed time series reveal key temporal patterns and interannual variability in wetland CH4 emissions over the past two decades, providing critical insights into long-term biogeochemical processes.
Short Talk: Oscar Zimmerman, Northern Arizona University
“Seasonal variation in canopy greenness and ecosystem photosynthesis in an arid evergreen conifer woodland”
Temperate evergreen conifers display distinct periods of productivity and dormancy during the year that are associated with changes in photosynthetic capacity. A number of studies have shown that these patterns are correlated with seasonal variation in leaf pigments, especially photoprotective carotenoids that increase in content during winter, causing needle ‘yellowing’. These changes form the basis for tracking the seasonality of evergreen conifer forests using visible-wavelength remote sensing indices. Prior studies testing this approach focused on forests in relatively cool and/or mesic climates, though evergreen conifers cover large semi-arid and arid regions of the western USA. To fill this knowledge gap, we are using near-surface remote sensing data collected at Cedar Mesa (US-CdM) in southern Utah to investigate: (1) seasonal changes in canopy color and spectral reflectance; (2) how temperature and water availability drive those changes; and (3) its correlation with the ecosystem-level CO2 exchange.
So far, we have found that the evergreen canopy dominated by juniper displays large seasonal variation in greenness calculated from phenocam imagery. Canopy greenness was highly correlated with the chlorophyll/carotenoid index calculated from narrowband radiometer data (r = 0.97), suggesting that changes in greenness were driven by photosynthetic pigments. A temperature sum model did a good job at explaining changes in greenness across two years during both spring and autumn (r-squared = 0.94 and 0.97, respectively). However, the magnitude of spring green-up was about 33% higher in the year with greater winter soil water content, a pattern not captured by the model. Analyses are underway to incorporate water availability into the model and compare changes in greenness with the CO2 flux data.
Our study advances the application of tools for investigating how climate warming might impact woodland ecosystems of the western USA through shifts in growing season timing and length.
Short Talk: Xian Wang, Indiana University
“Estimating Cropland Carbon Fluxes and the Role of Agricultural Conservation Practices Using Remote Sensing and Machine Learning”
Accurately estimating Net Ecosystem Exchange (NEE) at broad spatial and temporal scales is critical for understanding carbon dynamics in agricultural landscapes. Traditional methods relying on direct soil organic carbon measurements are labor-intensive and lack scalability, limiting their utility for regional assessments. To address this gap, we developed a framework that leverages eddy covariance data to model NEE and generate early indicators of soil carbon trends without depending solely on long-term, in-situ observations. Our approach integrates Harmonized Landsat and Sentinel-2 satellite imagery with the Daymet meteorological dataset to estimate NEE at 30-meter resolution across 16 Midwestern and Southern U.S. states. Using data from 57 eddy covariance towers, we trained a random forest model to estimate monthly NEE, achieving an R² of approximately 0.71. When compared to independent benchmarks, including the FLUXCOM product, the model’s predictions showed strong agreement, with an R² of 0.85, reinforcing its robustness and reliability.
To demonstrate the practical utility of the predicted NEE estimates, we conducted a case study to assess the impact of cover cropping on carbon dynamics. Specifically, we used the model to compare predicted NEE values between fields with and without cover crops. These fields were identified using a comprehensive transect survey dataset covering over 14,000 fields annually from 2014 to 2022. The initial analysis revealed a modest annual NEE difference between cover-cropped and non-cover-cropped fields—smaller than soil carbon gains often reported at the field scale. This value may evolve as the analysis is refined but also reflect a temporal lag between NEE and observable changes in soil organic carbon following the implementation of cover cropping, highlighting the importance of accounting for time dynamics in carbon sequestration assessments.
Overall, this study demonstrates the value of combining remote sensing, eddy covariance, and machine learning to produce high-resolution, scalable carbon assessments in agricultural systems and inform conservation practice evaluation at regional scales.
Friday, October 24
Invited Speaker: M. Susana Alvarado-Barrientos, INECOL
“Operational advancements of the MexFlux network on its fifteenth anniversary”
Author List: Enrico A. Yepez[1], Tonantzin Tarin[2], Zulia M. Sanchez-Mejia[1], M. Susana Alvarado-Barrientos*[3], Stephen H. Bullock [4], Alejandro E. Castellanos [5], Mónica Cervantes-Jiménez [6], Bruno M. Chavez-Vergara [2], Alejandro Cueva [7], Josue Delgado-Balbuena [1,8], Bernardo Figueroa-Espinoza [2], Dulce Flores [9], Jaime Garatuza-Payán[1], Eugenia Gonzalez del Castillo [2], Cesar Hinojo-Hinojo [5], Friso Holwerda [2], Eli R. Pérez-Ruiz [10], Julio C. Rodríguez[5], Nidia E. Rojas-Robles [11], Jorge M. Uuh-Sonda[1], Rodrigo Vargas [11], Martha L. Vargas-Terminel [1], Erik Velasco[12], Samuel Villareal [13]
[1] Instituto Tecnologico de Sonora (ITSON), Mexico
[2] Universidad Nacional Autonoma de Mexico (UNAM), Mexico
[3] Instituto de Ecologia A.C. (INECOL), Mexico
[4] Centro de Investigación Científica y de Educación Superior de Ensenada (CICESE), Mexico
[5] Universidad de Sonora (UNISON), Mexico
[6] Universidad Autonoma de Queretaro (UAQ), Mexico
[7] El Colegio de la Frontera Sur (ECOSUR), Mexico
[8] Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP), Mexico
[9] Centro de Investigación y de Estudios Avanzados (CINVESTAV), Mexico
[10] Universidad Autónoma de Ciudad Juárez (UACJ), Mexico
[11] Arizona State University (ASU), USA
[12] Molina Center for Energy and the Environment (MCE2), USA
[13] Centro de Investigación en Materiales Avanzados S.C. (CIMAV), Mexico
The MexFlux network, a conglomerate of 13 public Mexican universities and research institutions with 10 active flux sites and historical flux data spread across Mexico, is celebrating its fifteenth anniversary. In 2010 we committed to exchange the best available science on carbon and water fluxes to better understand the function of neotropical ecosystems in Mexico. Slowly but steadily, after navigating three lustrums of multiple challenges, today we present three cornerstones of community consolidation that will guide the next decade of MexFlux’ goals. 1) Generational change and commitment to train and engage new generations. Today, under the leadership of an empowered new generation of MexFlux scientists -predominantly women- we have constructed a space for dialogue among ourselves, and with our peers in North and Latin America, as well as an agenda to train a new generation of Spanish-speaking flux scientists. We will present the results of TAFE, “Taller de Aprendizaje de Flujos Ecosistemicos”, a workshop funded by the US Department of Energy in collaboration with MexFlux which represents the culmination of efforts of the MexFlux-AmeriFlux alliance to foster the engagement of flux science instructors and students from 10 countries of the Americas. 2) Commitment to sustain flux measurements in Mexico. MexFlux gathers soil, plant and atmospheric scientists that share common curiosity and goals to research both, the magnitude and the overarching controls of water and carbon fluxes in ecosystems across Mexico´s diverse and complex landscape. The number of active flux sites is small but the effort to add sites continues. Four new observatories will be operational in 2025-2026, including two mangrove forests; one in the Pacific and one in the Atlantic, and two distinct forest types previously unmonitored in Mexico (a tropical rainforest in the Yucatan peninsula and a managed conifer forest in Durango). 3) Organizational milestones and governmental recognition. Today, MexFlux is recognized by our federal government with the distinction of: “Laboratorio Nacional” (a national governmental strategy to support networks´ capabilities poised to solve national problems), which in addition to recognizing the potential of MexFlux to deliver science-based solutions to some our national environmental problems, the network will, for the first time, receive financial resources from the Mexican government to advance our goals in the next three years.
Session: Urban and Managed Lands
Short Talk: Ankur Desai, University of Wisconsin
“Here comes the sun: Eddy fluxes over agrivoltaic solar fields”
A thriving planet needs to sustain its supply of food and energy. For electricity generation, solar photovoltaics (PVs) is the most viable, fastest growing, and cheapest option. However, demand for land to grow food can conflict for space for solar PV. Can we transition to solar while also protecting arable land? The burgeoning development of agrivoltaics (AV) is now seen as one solution being rapidly implemented in many parts of the world. However, direct evaluation of year-round impacts on local microclimates, evapotranspiration, and net carbon productivity have been limited, especially in temperate mid-latitude locations. Here, we present the first North American eddy flux towers at an AV array both before (US-KSP) and after array installation (US-KSA), and at an adjacent control site (US-KSC), in southern Wisconsin USA. These towers were placed in or near a research-focused AV facility co-developed with a local energy company and university, providing 2.25 MW AC / 2.87 MW DC of power from solar PV. Initial results from 2024-2025 demonstrate that tower turbulence metrics generally denote high-quality fluxes despite the challenging surface heterogeneity and topography. Compared to the control tower, the array tends to dampen evapotranspiration fluxes and carbon uptake especially at high VPD. Additional soil moisture, thermal, and radiation sensors, drone flights, and photosynthesis chamber measurements provide additional context for these patterns. We look forward to sharing these new site data in the Ameriflux repository and finding collaboration with other renewable energy land use studies.
Short Talk: Kyle Delwiche, UC Berkeley
“Scalable Annual Carbon Flux Estimates in Restored Wetlands Using Data–Model Fusion”
Wetland restoration is increasingly recognized as a natural climate solution with the potential to sequester substantial amounts of carbon and deliver multiple co-benefits. However, quantifying greenhouse gas (GHG) emissions from restored wetlands in a timely and cost-effective manner remains a major challenge. Continuous ecosystem-scale flux measurements using eddy covariance systems offer high-resolution data but are prohibitively expensive for widespread or long-term deployment across all restoration sites.
To address this gap, we present a data–model fusion approach that integrates process-based modeling with short-term eddy covariance measurements to generate accurate annual estimates of carbon fluxes. We demonstrate that just a few weeks of flux data can be used to constrain model parameters sufficiently to produce robust annual estimates of net carbon exchange. This approach offers a pragmatic alternative to relying solely on permanent flux towers or on modeled estimates without site-specific calibration. By balancing accuracy with practicality, this method supports the development of economically viable carbon accounting strategies for wetland restoration projects seeking to generate carbon credits.