KHKatherine D. Haynes

Katherine D. Haynes

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Katherine D. Haynes has a total of 77 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Alaskan regional-scale measurements show late season ecosystem carbonyl sulfide uptake decoupled from gross photosynthesis

    by Aleya Kaushik, John Miller, Stephen Montzka, Lei Hu, Colm Sweeney, Kathryn McKain, Ian Baker, Katherine Haynes, Scott Denning, Arlyn Andrews

    Increasing atmospheric CO 2 seasonal cycle amplitudes in boreal regions have been attributed to climate-driven changes in land ecosystems, but terrestrial biosphere models (TBMs) are unable to replicate observations, leading to large uncertainties in future predictions of carbon cycle changes. Accurately partitioning net ecosystem exchange into its component fluxes—gross primary production (GPP) and respiration—is essential for understanding impacts of changing climate on the Arctic and boreal carbon balance, yet these component fluxes cannot be measured directly. Carbonyl sulfide (OCS) has been used to infer GPP at site to global scales, because its one-way uptake by plants is an analog for photosynthesis. However, expanding site-level process understanding to regional scales remains challenging. Here, we use atmospheric OCS mole fraction observations representative of Alaskan boreal forests to evaluate simulations of OCS fluxes in a state-of-the-science TBM. We use TBM-estimated OCS fluxes and surface influence functions on the order of 100–1000 km to simulate OCS mole fractions at the NOAA Global Monitoring Laboratory’s CRV tower site in central Alaska. By comparing with atmospheric observations, we can evaluate the TBM over much larger scales than is possible using eddy covariance data while still providing valuable information about underlying mechanisms. Comparisons reveal a missing ecosystem sink corresponding to a concentration difference of 22.3 ± 9.1 ppt OCS for July–November relative to observed concentrations of 433 ± 26 ppt at CRV. Solely improving the temperature sensitivity of apparent mesophyll conductance reduces the July–November mismatch between modeled and observed OCS concentration data by ∼5.3 ± 2.7 ppt. Consideration of alternate land cover maps provides an additional ∼5.3 ± 3.0 ppt towards the mismatch. These results demonstrate a strong decoupling of OCS and GPP, especially after the end of the growing season. Our analyzes demonstrate the limitations of using OCS as a proxy for GPP and highlight potential missing processes that need to be incorporated into future OCS modeling efforts to maximize its potential as a photosynthetic tracer.


  • Exploring the Use of Machine Learning to Improve Vertical Profiles of Temperature and Moisture

    by Katherine Haynes, Jason Stock, Jack Dostalek, Charles Anderson, Imme Ebert-Uphoff

    Vertical profiles of temperature and dewpoint are useful in predicting deep convection that leads to severe weather, which threatens property and lives. Currently, forecasters rely on observations from radiosonde launches and numerical weather prediction (NWP) models. Radiosonde observations are, however, temporally and spatially sparse, and NWP models contain inherent errors that influence short-term predictions of high impact events. This work explores using machine learning (ML) to postprocess NWP model forecasts, combining them with satellite data to improve vertical profiles of temperature and dewpoint. We focus on different ML architectures, loss functions, and input features to optimize predictions. Because we are predicting vertical profiles at 256 levels in the atmosphere, this work provides a unique perspective at using ML for 1D tasks. Compared to baseline profiles from the Rapid Refresh (RAP), ML predictions offer the largest improvement for dewpoint, particularly in the middle and upper atmosphere. Temperature improvements are modest, but CAPE values are improved by up to 40%. Feature importance analyses indicate that the ML models are primarily improving incoming RAP biases. While additional model and satellite data offer some improvement to the predictions, architecture choice is more important than feature selection in fine-tuning the results. Our proposed deep residual U-Net performs the best by leveraging spatial context from the input RAP profiles; however, the results are remarkably robust across model architecture. Further, uncertainty estimates for every level are well calibrated and can provide useful information to forecasters.


  • Creating and Evaluating Uncertainty Estimates with Neural Networks for Environmental-Science Applications

    by Katherine Haynes, Ryan Lagerquist, Marie McGraw, Kate Musgrave, Imme Ebert-Uphoff

    Neural networks (NN) have become an important tool for prediction tasks—both regression and classification—in environmental science. Since many environmental-science problems involve life-or-death decisions and policy making, it is crucial to provide not only predictions but also an estimate of the uncertainty in the predictions. Until recently, very few tools were available to provide uncertainty quantification (UQ) for NN predictions. However, in recent years the computer-science field has developed numerous UQ approaches, and several research groups are exploring how to apply these approaches in environmental science. We provide an accessible introduction to six of these UQ approaches, then focus on tools for the next step, namely, to answer the question: Once we obtain an uncertainty estimate (using any approach), how do we know whether it is good or bad? To answer this question, we highlight four evaluation graphics and eight evaluation scores that are well suited for evaluating and comparing uncertainty estimates (NN based or otherwise) for environmental-science applications. We demonstrate the UQ approaches and UQ-evaluation methods for two real-world problems: 1) estimating vertical profiles of atmospheric dewpoint (a regression task) and 2) predicting convection over Taiwan based on Himawari-8 satellite imagery (a classification task). We also provide Jupyter notebooks with Python code for implementing the UQ approaches and UQ-evaluation methods discussed herein. This article provides the environmental-science community with the knowledge and tools to start incorporating the large number of emerging UQ methods into their research. Significance Statement: Neural networks are used for many environmental-science applications, some involving life-or-death decision-making. In recent years new methods have been developed to provide much-needed uncertainty estimates for NN predictions. We seek to accelerate the adoption of these methods in the environmental-science community with an accessible introduction to 1) methods for computing uncertainty estimates in NN predictions and 2) methods for evaluating such estimates.


  • Open Access

    A Modified Vegetation Photosynthesis and Respiration Model (VPRM) for the Eastern USA and Canada, Evaluated With Comparison to Atmospheric Observations and Other Biospheric Models

    by Sharon M. Gourdji, Anna Karion, Israel Lopez‐Coto, Subhomoy Ghosh, Kimberly L. Mueller, Yu Zhou, Christopher A. Williams, Ian T. Baker, Katharine D. Haynes, James R. Whetstone

    Atmospheric CO2 measurements from a dense surface network can help to evaluate terrestrial biosphere model (TBM) simulations of Net Ecosystem Exchange (NEE) with two key benefits. First, gridded CO2 flux estimates can be evaluated over regional scales, not possible using flux tower observations at discrete locations for model evaluation. Second, TBM ability to explain atmospheric CO2 fluctuations due to the biosphere can be directly tested, an important objective for anthropogenic emissions monitoring using atmospheric observations. Here, we customize the Vegetation Photosynthesis and Respiration Model (VPRM) for an eastern North American domain with strong biological activity upwind of urban areas. Parameters are optimized using flux tower observations from a historical database with sites in (and near) the domain. In addition, the respiration model (originally a linear function of temperature) is modified to account for impacts of changing foliage, non‐linear temperature, and water stress. Flux estimates from VPRM, the Carnegie‐Ames‐Stanford Approach (CASA) model and the Simple Biosphere Model v4 (SiB4), are convolved with footprints from atmospheric transport models for evaluation with CO2 observations at 21 towers in the domain, with roughly half of the towers used here for the first time. Results show that the new respiration model in VPRM helps to correct a growing season sink bias in the atmosphere associated with underestimated summertime respiration using the original model with annual parameters. The new VPRM also better explains fine‐scale atmospheric CO2 variability compared to other TBMs, due to higher resolution diagnostic phenology, the new respiration model, domain‐specific parameters, and high‐quality input data sets.


  • Open Access

    Accurate Simulation of Both Sensitivity and Variability for Amazonian Photosynthesis: Is It Too Much to Ask?

    by Sarah M. Gallup, Ian T. Baker, John L. Gallup, Natalia Restrepo‐Coupe, Katherine D. Haynes, Nicholas M. Geyer, A. Scott Denning

    Estimates of Amazon rainforest gross primary productivity (GPP) differ by a factor of 2 across a suite of three statistical and 18 process models. This wide spread contributes uncertainty to predictions of future climate. We compare the mean and variance of GPP from these models to that of GPP at six eddy covariance (EC) towers. Only one model's mean GPP across all sites falls within a 99% confidence interval for EC GPP, and only one model matches EC variance. The strength of model response to climate drivers is related to model ability to match the seasonal pattern of the EC GPP. Models with stronger seasonal swings in GPP have stronger responses to rain, light, and temperature than does EC GPP. The model to data comparison illustrates a trade‐off inherent to deterministic models between accurate simulation of a mean (average) and accurate responsiveness to drivers. The trade‐off exists because all deterministic models simplify processes and lack at least some consequential driver or interaction. If a model's sensitivities to included drivers and their interactions are accurate, then deterministically predicted outcomes have less variability than is realistic. If a GPP model has stronger responses to climate drivers than found in data, model predictions may match the observed variance and seasonal pattern but are likely to overpredict GPP response to climate change. High or realistic variability of model estimates relative to reference data indicate that the model is hypersensitive to one or more drivers.


  • Open Access

    A modified Vegetation Photosynthesis and Respiration Model (VPRM) for the eastern USA and Canada, evaluated with comparison to atmospheric observations and other biospheric models

    by Sharon Gourdji, Anna Karion, Israel Lopez-Coto, Subhomoy Ghosh, Kim Lynn Mueller, Yu Zhou, Christopher A. Williams, Ian T. Baker, Katherine Haynes, James Whetstone

    Increasing atmospheric CO 2 measurements in North America, especially in urban areas, may help enable the development of an operational CO 2 emission monitoring system. However, isolating the fossil fuel emission signal in the atmosphere requires factoring out CO 2 fluctuations due to the biosphere, especially during the growing season. To help improve simulations of the biosphere, here we customize the Vegetation Photosynthesis and Respiration Model (VPRM) at high-resolution for an eastern North American domain, upwind of coastal cities from Washington D.C. to Boston, MA, optimizing parameters using domain-specific flux tower data from 2001 to the present. We run three versions of VPRM from November 2016 to October 2017 using i) annual (VPRM ann ) and ii) seasonal parameters (VPRM seas ), and then iii) modifying the respiration equation to include the Enhanced Vegetation Index (EVI), a squared temperature term and interactions between temperature and water stress (VPRM new ). VPRM flux estimates are evaluated by comparison with other models (the Carnegie-Ames-Stanford Approach model, or CASA, and the Simple Biosphere Model v4), and with comparison to atmospheric CO 2 mole fraction data at 21 surface towers. Results show that VPRM new is relatively unbiased and outperforms all other models in explaining CO 2 variability from April to October, while VPRM ann overestimates growing season sinks by underestimating summertime respiration. Despite unknown remaining errors in VPRM new , and uncertainties associated with other components of the atmospheric CO 2 comparisons, VPRM new appears to hold promise for more effectively separating anthropogenic and biospheric signals in atmospheric inversion systems in eastern North America.


  • Open Access

    Beyond ecosystem modeling: A roadmap to community cyberinfrastructure for ecological data‐model integration

    by Istem Fer, Anthony K. Gardella, Alexey N. Shiklomanov, Eleanor E. Campbell, Elizabeth M. Cowdery, Martin G. De Kauwe, Ankur Desai, Matthew J. Duveneck, Joshua B. Fisher, Katherine D. Haynes, et al.

    In an era of rapid global change, our ability to understand and predict Earth's natural systems is lagging behind our ability to monitor and measure changes in the biosphere. Bottlenecks to informing models with observations have reduced our capacity to fully exploit the growing volume and variety of available data. Here, we take a critical look at the information infrastructure that connects ecosystem modeling and measurement efforts, and propose a roadmap to community cyberinfrastructure development that can reduce the divisions between empirical research and modeling and accelerate the pace of discovery. A new era of data‐model integration requires investment in accessible, scalable, transparent tools that integrate the expertise of the whole community, including both modelers and empiricists. This roadmap focuses on five key opportunities for community tools: the underlying foundationsof community cyberinfrastructure; data ingest; calibration of models to data; model‐data benchmarking; and data assimilation and ecological forecasting. This community‐driven approach is key to meeting the pressing needs of science and society in the 21st century.


  • Open Access

    Accurate Simulation of Both Sensitivity and Variability for Amazonian Photosynthesis: Too Much to Ask

    by Sarah Gallup, Ian T. Baker, John L Gallup, Natalia Restrepo-Coupe, Katherine Haynes, Nicholas Geyer, A Scott Dennning

    Causes of climate predictions’ uncertainty include wide spread in modeled gross primary productivity (GPP) for evergreen broadleaf forests. Deterministic predictions inherently lack the portion of variability that a regression’s error term summarizes. Omitted predictors’ contribution to error represent simulations’ necessary underestimation of real variability. Earth system model outputs with high variability relative to reference data warrant skeptical examination. We compare three statistical and 15 process models to site-level means, seasonal amplitude and driver responsiveness of GPP as calculated at six Amazon eddy covariance (EC) towers. Current month’s weather determines only 12% of the variability in EC GPP, implying that models whose predicted GPP’s variability approaches that of EC GPP probably are substantially hypersensitive to weather drivers. Roughly half the models have stronger seasonal GPP variability than ECs show, and inaccurately identify the timing of annual minimum GPP. Responses to temperature and light for some highly seasonal models are of the opposite sign as EC GPP’s. Strongly seasonal models’ deepest dip in photosynthesis both occurs later in the dry season and is more severe than EC estimates. Excessive reactivity to drivers appears to cause the high simulated variability of the strongly seasonal models.

  • Open Access

    Representing Grasslands Using Dynamic Prognostic Phenology Based on Biological Growth Stages: Part 2. Carbon Cycling

    by Katherine D. Haynes, Ian T. Baker, A. Scott Denning, Sebastian Wolf, Georg Wohlfahrt, Gerard Kiely, Renee C. Minaya, John M. Haynes

    Grasslands are one of the most widely distributed and abundant vegetation types globally, and land surface models struggle to accurately simulate grassland carbon dioxide, energy, and water fluxes. Here we hypothesize that this is due to land surface models having difficulties in reproducing grassland phenology, in particular in response to the seasonal and interannual variability of precipitation. Using leaf area index (LAI), net primary productivity, and flux data at 55 sites spanning climate zones, the aim of this study is to evaluate a novel prognostic phenology model (Simple Biosphere Model, SiB4) while simultaneously illustrating grassland relationships across precipitation gradients. Evaluating from 2000 to 2014, SiB4 predicts daily LAI, carbon, and energy fluxes with root‐mean‐square errors < 15% and individual biases −2per day, with SiB4 and Fluxnet estimates within 18%. Despite differences in mean annual precipitation and growing season length, all grassland sites shift to seasonal carbon sinks one month prior to peak uptake. During a U.S. drought, MODIS and SiB4 had nearly identical LAI responses, and the LAI change due to drought was less than the LAI change across the precipitation gradient, indicating that grassland drought response is not as strong as the overlying climate response.


  • Open Access

    Representing Grasslands Using Dynamic Prognostic Phenology Based on Biological Growth Stages: 1. Implementation in the Simple Biosphere Model (SiB4)

    by K. D. Haynes, I. T. Baker, A. S. Denning, R. Stöckli, K. Schaefer, E. Y. Lokupitiya, J. M. Haynes

    Grasslands grow in a sequence of seasonal growth stages that respond to both climate and weather, and these relationships can be used to establish a strategy for predicting plant phenology. Current plant states (phenophase) can be represented as one of established growth stages that dictate carbon allocation and leaf photosynthetic capacity. Calculating daily phenophases from climate and environmental relationships allows for sequential growth stages (i.e., well‐defined seasonal cycles with a single growth period) or dynamic growth stages (i.e., multiple growth periods during a growing season). Senescence results from biomass mortality in response to environmental conditions. This approach uses a single mechanistic framework to represent grassland ecology, removing the dependence on satellite‐based vegetation indices and individual site tuning of parameters. Rather than being specified, a variety of properties emerge, from allometric relationships such as root‐shoot ratios, to behavior across moisture gradients, to interannual variability in growing season lengths, carbon stores, and land surface fluxes. Using dynamic phenology stages to link biophysical and biogeochemical processes provides a mechanism to predict self‐consistent land‐atmosphere exchanges of carbon, water, energy, radiation, and momentum, as well as carbon storage in cascading pools of biomass; and describing these processes in a mathematically determinate model makes them clear, testable, and usable for predictions. This paper describes this new phenology method as it is implemented in the Simple Biosphere Model Version 4 (SiB4), and a companion paper evaluates this method at grassland sites worldwide.