click here to jump to the Remote Sensing Session on Oct 4
click here to jump to the Methane Session on Oct 5
click here to jump to the Addressing New Challenges Session on Oct 5
click here to jump to the General Flux Science Session on Oct 6
Looking for poster abstracts? –> Download them here.
Wednesday, October 4
Opening Session
Invited: Bill Munger[1], Lucy Hutyra[2]
“From Towers to Regions – Insights on Fluxes Across Spatial and Temporal Scales”
[1] Harvard University, [2] Boston University
The New England landscape has been shaped by ecological succession, climate, and land-use decisions over the past millennia. The forest we see today at Harvard Forest has regrown on lands that were cleared for cultivation or selectively logged by European settlers starting in the 1600s, and largely abandoned by the late 1800’s. This long-term legacy sets the context for how we interpret the carbon flux record over the past three decades to try and understand what drives interannual variability and trends. Since the flux tower observations began at Harvard Forest in 1991 the mean annual temperature has increased by over 1C and standing biomass increased by over 30%. On hourly to daily timescales the observed carbon and water fluxes largely respond to meteorological forcing within the constraints imposed by seasonality and the forest attributes within the flux footprint associated with the observed time interval. On longer time scales forest physiological responses have been changing. Although we can attribute some of the change to climate, concentrations of CO2 and air pollutants, increasing biomass and shifts in its species distribution, the observations point out the importance of disturbance events that can alter carbon dynamics for multi-year intervals, or permanently as we see with the death of hemlock trees from adelgid infestation and replacement by deciduous trees.
At regional scales, land use and disturbance dynamics shape landscapes both through the carbon that is lost and the changes in the ecosystems left behind or created. Forests edges are ubiquitous across the landscape and in temperate regions we have found dramatic alterations in carbon exchange, nutrient cycling, and microbial communities. Regional observations of CO2 mixing ratios also indicate altered carbon exchange in some of the most fragmented portions of our landscape. Finally, satellite-based observations of solar induced fluorescence similarly suggest vegetation productivity differences with land use and forest fragmentation. Taken as a whole, this talk will highlight key frontiers for carbon cycle science.
Invited: Martha Anderson, USDA-ARS
“On the synergy between flux towers and remote sensing in understanding landscape water use”
M.C. Anderson, Y. Yang, K. Knipper, J. Xue, B. Kustas, F. Gao, J. Alfieri, C.R. Hain
A variety of remote sensing approaches have been developed to map evapotranspiration (ET) and vegetation stress using thermal infrared (TIR) and/or visible-to-shortwave-infrared (VSWIR) imagery from satellites. Especially at pixel sizes resolving individual field and crops (100-m or finer), these data are valuable for informing a range of agricultural water management decisions, at the farm level to watershed and basin scales. This scale of ET retrieval is also useful for upscaling localized flux tower observations to larger regions of interest. In this presentation we discuss a multi-sensor/multi-scale approach to ET mapping utilizing a suite of TIR-VSWIR Earth-observing satellite sensors, fused to produce datacubes with both high spatial (30m or finer) and temporal (daily) resolution aimed at actionable agricultural decision making. Multi-year ET datacubes have been constructed over several field experiment sites and evaluated with tower observations collected in-field. Applications for these datasets will be described, including vineyard irrigation management and rangeland health monitoring. We demonstrate how the spatial information provided by the remote sensing, at resolutions sampling the tower footprint, provide valuable context for the tower fluxes and a metric of spatial representativity.
Short Talk: Lewis Kunik
“Satellite-based solar-induced fluorescence tracks seasonal and elevational patterns of photosynthesis in California’s Sierra Nevada mountains”
Lewis Kunik *[1], David R Bowling [1], Brett Raczka [2], Christian Frankenberg [3,4], Philipp Köhler [5], Rui Cheng [6], Kenneth R Smith [1], Martin Jung [7], and John C Lin [1]
[1] University of Utah, Salt Lake City, UT, USA
[2] National Center for Atmospheric Research, Boulder, CO, USA
[3] California Institute of Technology, Pasadena, CA, USA
[4] Jet Propulsion Laboratory, California Institute of Technology, Pasadena, USA
[5] EUMETSAT, Darmstadt, Germany
[6] Massachusetts Institute of Technology, Cambridge, MA, USA
[7] Max Planck Institute for Biogeochemistry, Jena, Germany
Robust carbon monitoring systems are needed for land managers to assess and mitigate the changing effects of ecosystem stress on Western U.S. forests, where most aboveground carbon is stored in mountainous areas. Atmospheric carbon uptake via gross primary productivity (GPP) is an important indicator of ecosystem well-being and is particularly relevant to such monitoring systems; however, limited ground-based observations in remote areas of complex topography represent a significant challenge for tracking regional-scale GPP. Satellite observations can help bridge these monitoring gaps, but the accuracy of remote sensing methods for inferring GPP is still limited in montane evergreen needleleaf biomes, where 1) photosynthetic activity is largely decoupled from canopy structure and chlorophyll content, and 2) strong heterogeneity in phenology and atmospheric conditions are difficult to resolve in space and time. Using monthly solar-induced chlorophyll fluorescence (SIF) sampled at ~4km from the TROPOMI satellite instrument, we show that high resolution, satellite-observed SIF follows ecological expectations of seasonal and elevational patterns of GPP across a 3000-meter elevation gradient in the Sierra Nevada mountains of California. After accounting for high snow cover-induced reflected radiance in TROPOMI SIF, we found strong agreement in the elevational patterns of SIF and a model-data GPP product (FLUXCOM-RS); however, peak-season GPP from FLUXCOM was sustained through late summer, revealing significant timing differences with TROPOMI SIF. We attribute these timing differences to overestimation of FLUXCOM GPP relative to eddy covariance (EC) observations in summer and fall. Comparisons with simulated GPP from a land surface model (CLM5.0) indicate that summer moisture limitation plays a crucial role in these timing differences at low-to-mid elevations, while TROPOMI SIF aligns closely with seasonal patterns of EC GPP. These results suggest that satellite-observed SIF can serve as a useful diagnostic and constraint to improve upon gridded estimates of GPP towards multi-scale carbon monitoring systems in montane, evergreen conifer biomes at regional scales.
Short Talk: Steve Kannenberg
“Multi-scale analysis reveals dominant role of soil moisture in mediating dryland ecosystem fluxes”
Steve Kannenberg *[1,2], Bill Anderegg [3,4], Mallory Barnes [5], Matt Dannenberg [6], Alan Knapp [2]
[1] Department of Biology, West Virginia University, Morgantown, WV, USA
[2] Department of Biology and Graduate Degree Program in Ecology, Colorado State University, Fort Collins, CO, USA
[3] School of Biological Sciences, University of Utah, Salt Lake City, UT, USA
[4] Wilkes Center for Climate Science and Policy, University of Utah, Salt Lake City, UT, USA
[5] O’Neill School of Public and Environmental Affairs, Indiana University, Bloomington, IN, USA
[6] Department of Geographical and Sustainability Sciences, University of Iowa, Iowa City, IA 52245, USA
Drylands exert a powerful influence over global interannual variability in carbon and water cycling due to their substantial heterogeneity over space and time, yet this variability in ecosystem fluxes presents a challenge for comprehensively understanding their primary drivers. Here, we quantified the sensitivity of dryland gross primary productivity (GPP) and evapotranspiration (ET) to various hydrometeorological drivers by synthesizing eddy covariance data, remote sensing products, and land surface model output. We found that daily GPP and ET derived from eddy covariance were predominantly sensitive to soil moisture fluctuations, with marginal sensitivity to vapor pressure deficit and little to no sensitivity to air temperature or light. Remotely-sensed GPP and ET products accurately captured the sensitivity of eddy covariance fluxes to soil moisture, but largely over-predicted their sensitivity to atmospheric drivers. In contrast, CMIP6 land surface models underestimated the sensitivity of GPP to soil moisture fluctuations by approximately 45%. Our findings underscore the critical importance of soil moisture for dryland carbon-water cycling amidst debates about the role of vapor pressure deficit in a changing climate. It is thus imperative to both improve model representation of vegetation water limitation and more realistically represent how atmospheric drivers affect dryland vegetation in global GPP and ET products.
Short Talk: Mostafa Javadian
Mostafa Javadian [1,2], Russel L. Scott [3], William Woodgate [4,5], Andrew D. Richardson [1,6], William K. Smith [2]
[1] Center for Ecosystem Science and Society (ECOSS), Northern Arizona University, Flagstaff, AZ, 86011, USA
[2] School of Natural Resources and the Environment, University of Arizona, Tucson, AZ, 85721, USA
[3] Southwest Watershed Research Center, USDA Agricultural Research Service, Tucson, AZ, 85719, USA
[4] School of the Environment, The University of Queensland, St Lucia, 4169, QLD, Aus
[5] CSIRO, Space and Astronomy, Kensington, 6151, WA, Aus
[6] School of Informatics, Computing, and Cyber Systems (SICCS), Northern Arizona University, Flagstaff, AZ, 86011, USA
Canopy temperature (Tc) is a crucial factor affecting plant productivity and water stress. A better understanding of the relationship between Tc and water stress is urgently needed to enable more accurate monitoring of ecosystem functioning in a changing climate. Here, we used high spatiotemporal resolution thermal infrared cameras deployed across a gradient of eddy covariance flux tower sites including a predominately water-limited mixed grassland/shrubland (US-WKG), a seasonally water-limited evergreen needleleaf forest (US-MtB), and a predominantly energy-limited deciduous broadleaf forest (AU-TUM) to investigate determinates of Tc seasonality and its relationship with gross primary productivity (GPP) and its environmental drivers. Consistent with previous studies, we found midday Tc was generally warmer than air temperature (Tair) during the growing season (Tc:Tair slope ranging from 1.14 to 1.27). Water-limited sites exhibited higher Tc deviations from Tair (2.30 ± 1.2°C) compared to the energy-limited site (1.29 ± 0.8°C) partly due to a shift towards more sensible and less latent heat flux. The Tc:Tair slope increased along an aridity gradient and with the transpiration potential of the vegetation, with slopes increasing from the most arid site to the predominantly energy-limited site. Peak GPP occurred when Tc was higher than Ta again following a gradient occurred from grassland (+7.5°C Tc-Tair) to broadleaf evergreen (+2.2°C Tc-Tair). Tc-Tair dynamics were mostly associated with soil water content in water-limited sites where canopies undergo a substantial cooling process during the transition from dormancy to the peak of season, while net radiation played a crucial role in energy-limited site where they heat up compared to Tair within the same time frame. Our findings provide novel insights into the relationship between Tc and ecosystem water stress, highlighting the drivers of Tc-Tair across diverse ecosystems in various phenological stages, which has implications for ecosystem management in a changing climate.
Thursday, October 5
Invited: Rodrigo Vargas, University of Delaware
“Methane Emissions from Tree Stems: A Science Frontier or a Research Curiosity”
Research on methane emissions from tree stems has gained attention from both scientists and the general public. Studies have shown that tree stems in wetlands are significant sources of methane emissions, and recent research has also reported on methane emissions from tree stems in upland forests. While most studies have focused on wetland ecosystems, publications on tree stem methane emissions from upland forests have increased in recent years. It is important to note that emissions from tree stems vary greatly in space and time, presenting technical challenges in measuring, scaling these fluxes, and estimating annual budgets. Additionally, there is currently no agreement on the biophysical mechanisms that drive stem methane production and emissions. This presentation will focus on addressing these challenges and exploring research opportunities in methane emissions from tree stems in upland forests.
Short Talk: Theresia Yazbeck
“Reducing uncertainty of wetland-greenhouse gas emissions in earth system models by including eco-hydrological patch types sub-grid representation coupled with Landsat Sentinel-2 derived patch distributions”
Theresia Yazbeck* [1], Gil Bohrer [1], Oleksandr Shchehlov [1], Yang Ju [1], Madeline Scyphers [1], Justine Missik [1], Eric J. Ward [2], Robert Bordelon [3], Diana Taj [3], Jorge Villa [3], Kelly Wrighton [4], Qing Zhu [5], and William J. Riley [5]
[1] Department of Civil, Environmental and Geodetic Engineering, Ohio State University, United States of America
[2] U.S. Geological Survey, Wetland and Aquatic Research Center, United States of America
[3] School of Geosciences, University of Louisiana at Lafayette, United States of America
[4] Soil and Crop Sciences, Colorado State University, United States of America
[5] Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, United States of America
Wetlands are considered to be the largest emitters of biogenic methane, yet they represent the highest source of uncertainty in global methane emission estimates in Earth System Models (ESMs). This uncertainty is partially attributed to the small-scale spatial and temporal heterogeneity of biogeochemical and hydrological processes driving methane production, oxidation, and transport. Due to their coarse scale (10s’ of Km), ESMs do not explicitly simulate within-wetland variability of ecosystem conditions and biogeochemical processes. In addition, these variabilities are usually under-represented in coarse spatial- and temporal-resolution remote sensing images. In this study, we apply the Energy Exascale Earth System Model (E3SM) Land Model (ELM), where we developed a separate wetland land-unit. Representing wetland land-units allows the model to simulate multiple eco-hydrological patches (i.e., different vegetation communities) within a wetland at the sub-grid level, with distinct ecological, microbial, and hydrological parameters representing each patch type. The patch cover distribution is input to ELM using global, high-resolution, Harmonized Landsat Sentinel-2 (HLS) multispectral products (30m × 30m). We use seasonal time-series of HLS-derived NDVI, which provide distinct seasonal temporal “fingerprints” to classify HLS pixels to specific patch types and infer the corresponding plant cover distribution within the wetland. Regular Eddy-Covariance, chamber flux, and pore-water concentrations from two study-sites in Louisiana were used to validate our models. Our results show correspondence between observed and modelled carbon and methane fluxes and soil methane concentrations after optimizing vegetation photosynthetic rates, respiration rates, and methane production and oxidation parameters using a Bayesian approach (BOA, Bayesian Optimization for Anything). Our findings also show a higher precision when simulating multiple patches compared to single patch representations, thus emphasizing the role of wetland sub-grid representation within-wetland patch distribution in reducing models’ uncertainty.
Short Talk: Sara Knox
“The importance of salinity in regulating greenhouse gas fluxes in wetlands of the Prairie Pothole Region”
Sara Knox* [1], Pascal Badiou [2], Nick Lee [3], Darian Ng [4], Matthew Bogard [5], Zoran Nesic [6]
[1] McGill University, Department of Geography, Montreal, QC, Canada
[2] Ducks Unlimited Canada, Winnipeg, MB, Canada
[3] Max Planck Institute for Biogeochemistry, Department of Biogeochemical Integration, Jena, Germany
[4] The University of British Columbia, Department of Geography, Vancouver, BC, Canada
[5] University of Lethbridge, Biological Sciences, Lethbridge, AB, Canada
[6] The University of British Columbia, Faculty of Land and Food Systems, Vancouver, BC, Canada
With growing interest in wetland management and restoration as a Natural Climate Solution (NCS) improved estimates of wetland carbon (C) sequestration and greenhouse gas (GHG) exchange across wetland types are strongly needed. In this study, we focus on small isolated wetlands in the Prairie Pothole Region (PPR) of North America since these ecosystems are understudied relative to other wetland types, yet play important roles in C cycling and climate regulation. Eddy covariance (EC) flux towers were installed at two small isolated wetlands embedded in grasslands and cropland ecosystems in the PPR of southwestern Manitoba. While both sites are freshwater marshes dominated by Typha, one site (Young) is spatially heterogenous, consisting of a mix of open water and vegetation patches, while the other (Hogg) is more homogenous and dominated entirely by emergent vegetation. While the more homogenous wetland (Hogg) was a net CO2 sink on an annual scale, sequestering ~34 gC m-2 y-1 in 2021, the more heterogeneous site (Young) was a net CO2 source, emitting ~57 gC m-2 y-1. Sites also differed in CH4 emissions, with the Young wetland emitting ~7 gC-CH4 m-2 yr-1, while the Hogg wetland had minimal CH4 emissions (emitting ~ 1 gC-CH4 m-2 yr-1). Given the low CH4 emissions at Hogg, this site was a net GHG sink in 2021, while Young was a net GHG source. Differences in CH4 emissions between these sites were driven by higher sulfate concentrations and salinity at Hogg relative to Young. EC measurements were combined with field surveys across the Canadian PPR, which showed that that salinity restricted wetland CH4 emissions across the Canadian Prairies. We found that excluding salinity from upscaling estimates, as is currently done, leads to overestimation of emissions from small Canadian Prairie waterbodies by at least 81% (~1 Tg yr-1 CO2 equivalent), a quantity comparable to other major national emissions sources.
Session: Addressing new challenges with Flux science
Invited: Sparkle Malone, Yale University
“The Path to Effective Methane Emission Management: Insights for Climate Action and Sustainability”
Understanding methane (CH4) emissions is an important aspect of environmental and climate policies. With a higher short-term warming potential and a shorter average residence time in the atmosphere compared to carbon dioxide (CO2), CH4 mitigation is ideal for near term sustainable climate solutions. Despite natural processes offsetting 60-90% of anthropogenic emissions, over the last decade atmospheric methane CH4 has been increasing at an alarming rate. This concerning trend has prompted the commitment of 150 countries to the global methane pledge, aiming to reduce global emissions by a minimum of 30% from 2020 levels by 2030. Managing atmospheric CH4 is crucial for reducing the overall impact of human activities on the climate. Even though CH4 mitigation will focus on anthropogenic emissions, understanding natural processes is essential for the development of effective targets to keep warming below 1.5℃.
Successful climate action needs to be based on empirical data. Surface–atmosphere exchange of CH4 from biogenic sources and sinks, the biological and environmental processes driving these fluxes (e.g., ebullition, aerenchyma pumping), and how CH4 sources and sinks change over space and time, including interannual variability, are not well constrained. Currently, the eddy covariance method is the best approach to making ecosystem level estimates of flux contributions from different ecosystem types. To understand landscape, regional, and global source potentials, we have to develop dynamic and effective ways to scale-up ecosystem estimates. This work will challenge the flux community to establish infrastructure, evaluate algorithms, estimate uncertainty, and to develop standards for data integration.
Short Talk: Stefan Metzger
“Carbon Dew: Anchoring Equitable Climate Solutions in Directly Measured Greenhouse Gas Exchange”
Stefan Metzger* [1], George Burba* [2], Kyle Hemes [3], Gerbrand Koren [4], Sung-Ching Lee [5], Benjamin Runkle [6]
[1] NEON Program, Battelle, Boulder, CO, USA
[2] LI-COR Biosciences, Water for Food Global Institute, Lincoln, NE, USA
[3] Stanford Woods Institute for the Environment, Stanford, CA, USA
[4] Utrecht University, Utrecht, The Netherlands
[5] Max Planck Institute for Biogeochemistry, Jena, Germany
[6] University of Arkansas, Fayetteville, AR, USA
Technological, nature-based, and demand-side solutions are envisioned to avert the most drastic consequences of climate change, connected via a greenhouse gas (GHG) economy and government incentives. However, limitations in current Measurement, Reporting and Verification (MRV) curtail consistent climate progress and equitable climate finance. A MRV benchmark that is directly measured, uniformly derived, universally applicable across climate solution pathways, and traceable in near-real time provides a promising avenue to address this challenge.
Basic science offers cutting-edge GHG quantification ripe for technology transfer to a climate solution benchmark. Next-generation GHG Flux Mapping can unlock eddy-covariance measurements to achieve unmatched source attribution, statistical power, and process insight in an economical and easy-to-use way (https://tinyurl.com/flux-tower-mapping). The result is an orders-of-magnitude improved stream of directly-measured emission and sequestration rates to anchor project-scale mitigation, or wall-to-wall remote sensing and modeling when applied to GHG flux networks such as NEON, AmeriFlux and FLUXNET. This approach can yield GHG flux maps at decameter- and sub-hourly resolution, with data locked in a secure vessel such as a blockchain to promote transparency and integrity.
Access to such localized GHG exchange information via mobile apps similar to weather data enables public awareness and confidence, and narrows the knowledge-action gap. APIs can power day-to-day GHG management, climate solution research and GHG certificate intercomparisons, and development of climate-smart technologies, multi-scale policies and regulations. Furthermore, the benchmark directly represents a financial commodity – the physical flux of GHGs. Monetization paths include the development of commercial products, tools and services, through connecting pixel-scale GHG exchange to voluntary and compliance practices across industries.
With this concept we invite all stakeholders to join Carbon Dew, the community anchoring equitable climate solutions in direct GHG measurements (https://www.carbondew.org/join). Together we can catalyze robust, transparent MRV to accelerate evidence-driven, equitable climate progress.
Short Talk: Erik Velasco
“Urban flux towers: applications & challenges ”
* Erik Velasco (Molina Center for Energy and the Environment)
Armando Retama (Independent)
Luisa T. Molina (Molina Center for Energy and the Environment)
Matthias Roth (National University of Singapore)
Tall towers equipped with fast and accurate instrumentation to measure fluxes of trace gases, aerosols and energy are increasingly used in urban areas. They provide a direct measure of the net surface-atmosphere exchange that includes all sources and sinks within a footprint that is usually similar to the size of a complete neighborhood. Their correct application makes it possible to monitor emissions of greenhouse gases, pollutant gases, and aerosols, and in turn validate the accuracy of gridded emission inventories used in air quality management and climate change mitigation. Along with other measurement and modeling tools, flux towers provide data to determine contributions from anthropogenic and natural sources and sinks. For instance, the potential of urban greenery to offset anthropogenic carbon emissions. They also provide micrometeorological data not measured by traditional weather stations, which is fundamental to address climatological phenomena such as the urban heat island, and to understand meteorological processes relevant for the dispersion of pollutants. This talk reviews the basics and technical requirements of the eddy covariance method used by these towers. Their use to improve the environmental management of cities is discussed through our experience in measuring fluxes in Mexico City and Singapore, and the progress in the method achieved by the scientific community. The talk closes with suggestions for expanding their use to assess other environmental issues and make the most of already existing urban flux towers.
Short Talk: Xiangmin Sun
“Long-term eddy covariance measurement of heat and carbon dioxide fluxes in a low-rise residential neighborhood of Phoenix, Arizona”
Xiangmin Sun [1,2], Enrique R. Vivoni [1,2], Eli R. Perez-Ruiz [3], Quincy Stewart [4], Stevan R. Earl [4]
[1] School of Sustainable Engineering and the Built Environment, Arizona State University, Tempe, AZ, 85287-8704, USA
[2] Center for Hydrologic Innovations, Arizona State University, Tempe, AZ, 85287-8704, USA
[3] Department of Civil and Environmental Engineering, Universidad Autonoma de Ciudad Juarez, Chihuahua, Mexico.
[4] Central Arizona–Phoenix Long-Term Ecological Research, Arizona State University, Tempe, AZ, USA
Urban regions face the dual challenges of rising greenhouse gas emissions and more frequent extreme climatic events, including heat waves. Accurate in situ measurements of energy exchanges and water and carbon dioxide (CO2) emissions across different types of urban landscapes are essential for verifying top-down emissions estimates and quantifying their temporal and spatial variability. In 2012, the Central Arizona Phoenix Long-term Ecological Research (CAP-LTER) program established a 26-m eddy covariance tower within a low-rise residential neighborhood of Phoenix, Arizona. Datasets from the tower in 2012 and 2015 have been used previously to study the seasonal variability of water, energy, and CO2 fluxes and identify the role of urban landscape properties, including irrigated areas and impervious cover, within the tower footprint. To date, however, the long-term variability and the effects of different seasonal conditions on the flux measurements have not been analyzed. Here, we utilize the available data from 2012 to 2021 to perform a first set of analysis on the long-term variability from intra-seasonal to interannual time scales of the water, energy, and CO2 fluxes. We focused on comparing wet versus dry seasons in the recent record, for instance the exceptionally dry and hot summer of 2020, and the role of heat waves on CO2 emissions and thermodynamic processes in the urban environment. We also test if the low-rise residential site shows the presence of the oasis effect identified at nearby irrigated park that occurs during prominently during excessive heat warning days. During oasis effect days, advected energy from the urban environment augments local fluxes, leading to high rates of evapotranspiration when water is present. As shown here, long-term in situ monitoring in urban areas reveals temporal patterns in water, energy, and CO2 fluxes that are critical for understanding how extreme events such as heat waves punctuate the summertime conditions.
Friday, October 6
Invited: Dan Ricciuto, Oak Ridge National Laboratory
“Improving E3SM Land Model Predictions using AmeriFlux Observations and Machine Learning”
Land-surface models like the E3SM land model (ELM) contain many uncertain parameters and algorithms, and model performance may be improved significantly by calibrating these parameters using flux observations. Calibration of model parameters requires ensembles, which may need to be quite large depending on the number of uncertain parameters. However, ELM is computationally expensive to perform even a forward simulation at a single grid cell. Here we apply surrogate modeling approaches, in which we can use a simple model to predict the response surface of ELM for specific quantities of interest. This reduces the number of ELM simulations needed and allows them to be performed in parallel. We train neural networks as surrogate models at 10 AmeriFlux sites using 1000-2000 ELM simulations at each site varying 10 uncertain parameters. We optimize these parameters by running Markov Chain Monte Carlo (MCMC) on the surrogate models using gap-filled monthly net ecosystem exchange (NEE), gross primary productivity (GPP), energy and methane fluxes as constraints. We present posterior probability density functions (PDFs) for model parameters and outputs for individual sites, and for sites grouped by plant functional types (PFT). We envision that this approach can be used to reduce biases in future simulations, and to pinpoint and reduce remaining uncertainties.
Short Talk: Daniel Beverly
“Using site-specific soil water retention curves to demonstrate the relevance of soil water potential to ecosystem flux”
Alex Crookshanks [1], Daniel Beverly [1, 2], Sebastien Biraud [3], Dennis Baldocchi [3], Gil Bohrer [4], Steve Kannenberg [5], Marcy Litvak [6], Russ Scott [7], Rich Phillips [2], Kim Novick [1]
[1] Indiana University O’Neill School of Public and Environmental Affairs, Bloomington, IN, USA
[2] Indiana University Biology Department, Bloomington, IN, USA
[3] Lawrence Berkeley National Laboratory, Berkeley, CA, USA
[4] Department of Civil, Environmental and Geodetic Engineering, The Ohio State University, Columbus, OH, USA
[5] Department of Biology and Graduate Degree Program in Ecology, Colorado State University, Fort Collins, CO, USA
[6] Department of Biology, University of New Mexico, Albuquerque, NM, USA
[7] Southwest Watershed Research Center, USDA-ARS, Tucson, AZ, USA
Soil water potential (ΨS) controls a large number of biophysical processes, including the function of leaves, roots, microbes, and the drive of water through the soil-plant-atmosphere continuum. It’s a notably important variable that is rarely measured in-situ. Volumetric water content (θ) is widely measured in situ and often used to model carbon and water fluxes where soil water retention is absent. The soil water retention curve connects θ and ΨS. This relationship is highly non-linear and dependent on the complex interactions between soil texture and structure. Pedotransfer functions (PTFs) provide soil water retention parameters with empirical equations founded upon the relationship between simple soil physical properties (usually only % sand, silt, clay) and soil water retention. While widely used, PTFs fail to account for the effect of a specific capillary/root matrix on soil hydraulics. Lab-derived water retention curves could increase the accuracy of ΨS estimates, thus enabling more robust linkages between hydroclimate and plant function. We explore the agreement between lab-derived and PTF soil water retention parameters and confront the ecological relevance of ΨS by linking lab-derived measurements to ecosystem flux. Gap-filled gross primary productivity (GPP) from AmeriFlux sites representing a wide range of soil textures and environments were used to explore the relationship of lab-derived ΨS to ecosystem flux. Results found that lab-derived water retention curves could reduce conceptual uncertainty about how ecosystem fluxes respond to soil water deficits. Lab-derived curves illustrated the effect of organic matter on soil retention capabilities better than PTFs. Lab-derived ΨS had a stronger correlation with GPP than θ, as indicated by R-squared values of 0.981 versus 0.752 at US-MMS and 0.781 versus 0.531 at US-UMB. These results develop a necessary understanding of site-specific soil water retention curves and stress the importance of soil water potential measurements moving forward.
Short Talk: Angela Lafuente
“Contrasting wet and dry season Carbon Dioxide and Methane Fluxes of an Amazonian Palm Peatland”
Angela Lafuente *[1], Daniel Tyler Roman * [2], Jhon Rengifo *[3], Fenghui Yuan [4], Jeffrey D Wood [5], Lizardo Fachin [3], Daniel M Ricciuto [6], Rodney Chimner [1], Hinsby Cadillo-Quiroz [7], Randall K Kolka [8], Craig Wayson [2],Timothy J Griffis [4], Erik Lilleskov [9] ,
[1] Michigan Technological University, Houghton, MI, United States,
[2] USDA Forest Service, International Programs, Washington DC, United States,
[3] Instituto de Investigaciones de la Amazonia Peruana, Iquitos, Peru,
[4] University of Minnesota, Saint Paul, MN, United States,
[5] University of Missouri Columbia, Department of Soil, Water, and Climate, Columbia, MN, United States,
[6] Oak Ridge National Laboratory, Environmental Sciences Division and Climate Change Science Institute, Oak Ridge, TN, United States,
[7] Arizona State University, Tempe, AZ, United States,
[8] USDA Forest Service, Grand Rapids, United States,
[9] USDA Forest Service, Northern Research Station, Houghton, MI, United States.
South America contains the largest area of tropical peatlands representing 46% of the total global tropical peatland area. Tropical peatlands are one of the largest natural sources of atmospheric methane (CH4) and are keystone ecosystems that play a significant role in regional and global carbon budgets. Here we aim to partition net ecosystem carbon fluxes and to quantify ecosystem and plot (soil and stem) carbon fluxes in a tropical palm swamp peatland located near the city of Iquitos, Peru. We established an Eddy Covariance tower and an automatic chamber-based system to measure carbon dioxide (CO2) and CH4 fluxes at ecosystem level and from soil and tree stems respectively. We established six collars on the soil covering the spatial variability of the site and installed stem chambers at 50cm from the soil surface on the three dominant species on the site: Mauritia flexuosa, Mauritiella armata and Tabebuia insignis. Our results cover a full year with two dryer periods (September and December) in contrast to a prolonged wet season. We observed a significant response between soil CO2 and CH4 fluxes and the water table. When soils were wet we found small CO2 fluxes and high CH4 fluxes. However, when the soils become dryer CO2 fluxes increase and CH4 fluxes decrease. Similar responses were observed in the tower data with decreased CH4 flux and more positive CO2 flux during the dry periods. The tree contribution to the total CH4 fluxes represented less than 1.5% of the total flux and the response to water level from trees was different depending on the species. Our results showed a diel pattern on CO2 fluxes from stems. However, we did not observe diel patterns on soil CO2 or on CH4 fluxes from either the soil or from stems. These novel results provide insight into the response of tropical palm peatlands to changes in hydrologic conditions and highlight the importance of observational data in these ecosystems at varying spatial and temporal scales.
Short Talk: David Reed
“Distance Decay of Carbon Fluxes Across Continental Scales”
David Reed [1,2,3]
Savannah Rivera [2,3]
Housen Chu [3]
[1] Yale School of the Environment
[2] University of Science and Arts of Oklahoma
[3] Lawrence Berkeley National Laboratory
While AmeriFlux is a network of single observation sites, each measuring land-atmosphere fluxes from ~1 km2 of the Earth’s surface, the data can be used to quantify processes that operate beyond the scale of single measurements. In order to quantify the spatial distance the ecosystem measurements are valid, we apply the ecological concept of distance-decay to the AmeriFlux network where site-similarity is quantified as a function of landscape distance. Using ~450 sites from the AmeriFlux network, focused on net ecosystem exchange (NEE) and where available photosynthesis, respirations, and methane carbon fluxes, we first aggregated data into multi-year composite diurnal-seasonal time-series, then calculated site-simility for each site-pair across the network by using the dynamic time warping. Site-simility results show clear differences in distance-decay across continental scales. For NEE, wetlands show a high degree of similarity out to ~800,000 km while agriculture sites become more unique relative to other agriculture sites at distances of ~ 300,000 km. Results show ecosystems processes that control carbon fluxes have a high degree of similarity between sites and that information from flux sites can be used to scale flux data across the landscape, adding to the strength of the AmeriFlux observation network for continental scale modeling or remote sensing work.
Invited: Andrew Richardson[1,2], Adam Ford[2,3]
“Tracking how vegetation phenology affects land-atmosphere fluxes using PhenoCam and near-surface remote sensing”
[1] School of Informatics, Computing, and Cyber Systems, Northern Arizona University, Flagstaff,
AZ 86011
[2] Center for Ecosystem Science and Society, Northern Arizona University, Flagstaff, AZ 86011
[3] National Ecological Observatory Network, Boulder, CO 80301
Part 1 (Richardson)
The PhenoCam Network uses repeat digital photography to track vegetation phenology in
ecosystems across the US and around the world. Many AmeriFlux sites also have collocated
PhenoCams. I will provide a brief overview of how the PhenoCam method “works”, and show
how indices derived from PhenoCam imagery tracks the phenology of canopy photosynthesis as
measured by eddy covariance. I will highlight the data and resources available through the
PhenoCam Network web page (https://phenocam.nau.edu). I will conclude by discussing some
of the potential research questions that can be addressed with PhenoCam data.
Part 2 (Young)
Vegetation phenology is a fundamental ecosystem process, influencing seasonal dynamics in
fluxes of carbon, water, heat, and momentum. Investigating how phenology acts as a driver of
land-atmosphere interactions, as well as the relative importance of phenology compared to
other environmental factors (e.g., temperature, precipitation), is a key objective for ultimately
understanding boundary layer dynamics in different terrestrial ecosystems. I will show several
examples of how PhenoCam data can be directly applied and leveraged with eddy covariance
datasets to explore seasonal dynamics of sensible and latent heat fluxes across a wide range of
sites from the AmeriFlux Network.
Short Talk: Yujie Liu
“Gap-filling extra-long gaps in eddy covariance measurements using extreme gradient boosting”
Yujie Liu (yujie.liu@nau.edu)[1], Darby D. Bergl[1], Benjamin Lucas[2], Andrew P. Ouimette[4], David Y. Hollinger[5], Andrew D. Richardson[1,3]
[1] Northern Arizona University, Center for Ecosystem Science and Society (ECOSS), AZ 86004, USA
[2] Northern Arizona University, Department of Mathematics and Statistics, AZ 86004, USA
[3] Northern Arizona University, School of Informatics, Computing & Cyber Systems (SICCS), AZ 86004, USA
[4] University of New Hampshire, Earth Systems Research Center, Durham, NH 03824, USA
[5] USDA Forest Service, Northern Research Station, 271 Mast Rd, Durham, NH 03824, USA
Filling extra-long gaps in eddy covariance data is challenging due to ecosystem property changes over time and uncertainties in gap-filling algorithms. Marginal distribution sampling (MDS) is commonly used but performs poorly for long gaps. It depends on similar meteorological conditions, limiting the number of predictors and introducing systematic bias.
In our Bartlett Experimental Forest analysis (US-Bar), addressing 54.5% of missing NEE data from 2004 to 2022 is crucial for reliable NEE annual sums and understanding carbon dynamics. Tree-based machine learning methods like random forest or XGBoost show remarkable performance in gap-filling flux data. GCC from PhenoCam imagery offers insights into vegetation greenness dynamics since 2010 and can be used to fill long gaps in NEE. Using GCC as a predictor for the first time, we assess its impact on NEE estimation.
Before hyperparameter tuning, we compared the performance of different regression models, with XGBoost showing superior performance compared to others. Then, a novel artificial gap generation procedure was employed to create training, validation, and test sets, mitigating bias towards models performing better on excessively long gaps by splitting data into train and test set directly.
After hyperparameter tuning using randomized cross-validation, the XGBoost model demonstrated powerful performance for data between 2010 and 2022 (RMSE of 2.37). In comparison, the model without GCC achieved lower performance (RMSE of 2.60), which was similar to that of MDS (RMSE of 2.60). GCC was not available for 2004-2009 but using Enhanced Vegetation Index (EVI) as a predictor showed better performance (RMSE of 2.40) than a model without EVI (RMSE of 2.64).
Next, we’ll use XGBoost to gap-fill NEE data from the NEON site US-xBR, located 90 m from US-Bar, to assess NEE consistency between paired eddy covariance measurements. In summary, XGBoost effectively fills extra-long gaps, with GCC enhancing its performance.