2021 ET Workshop Lightning Talk Abstracts


A footprint-informed decomposition approach for deriving flux response functions at AmeriFlux sites

Housen Chu1, Patty Oikawa2, Thomas Fenster2, Camilo Rey-Sanchez3, Dennis Baldocchi4, Joe Verfaillie4, Stephen Chan1, Sigrid Dengel1, Sebastien C. Biraud1, Margaret S. Torn1

1Climate and Ecosystem Sciences Division, Lawrence Berkeley National Lab, Berkeley, CA, USA
2Department of Earth & Environmental Sciences, California State University – East Bay, Oakland, CA, USA
3Department of Environmental Science, Policy and Management, University of California, Berkeley, CA, USA
4Department of Marine, Earth, and Atmospheric Sciences, North Carolina State University, Raleigh, NC, USA

Eddy-covariance flux data are spatiotemporally dynamic. Global networks of eddy covariance towers provide the largest synthesized data sets of CO2, H2O, energy, and other GHGs fluxes. This includes FLUXNET and AmeriFlux data sets that have been widely used in many research and applications. While the eddy-covariance data are recognized for their rich temporal information, their spatially dynamic nature due to the varying source areas from time to time (i.e., so-called flux footprint) is often overlooked. As many flux tower sites are located in a more-or-less heterogeneous or patchy landscape, the spatial variations of land surface characteristics and the temporal dynamics of flux footprints jointly lead to the so-called representativeness issue, i.e., to what extent do the flux measurements taken at individual tower locations reflect the flux conditions of a specific land-cover or ecosystem type at all times. We developed a footprint-informed approach to decompose the response functions of CO2, H2O, and sensible heat fluxes at eddy-covariance sites. This approach incorporated the temporal dynamics of footprints and the spatial variations of land surface characteristics and was tested at selected AmeriFlux sites with different degrees of heterogeneity and patchiness. Our preliminary results showed that the approach was robust in decomposing flux response functions at sites with moderate heterogeneity and patchiness. And, the derived response functions can be used to inform fluxes at specific land-cover types within the flux footprints. The validity and uncertainty level of the flux decomposition approach depended on the degrees of heterogeneity and land-cover composition. Further tests and improvements will be conducted to better constrain the derivation of the response functions.



ET and Water Use Efficiency in Californian Dryland and Wetland Ecosystems

Siyan Ma1, Elke Eichelmann2, Dennis D. Baldocchi1

1Ecosystem Science Division, Department of Environmental Science, Policy and Management, University of California at Berkeley, Berkeley, CA, USA.
2School of Biology and Environmental Science, University College Dublin, Dublin, Ireland.

Evapotranspiration (ET) and water use efficiency (WUE) are critical concepts for ecosystem and water use management. However, ET and WUE are highly variable in space and time, and we need better understand their temporal and spatial variability in drier and wetter conditions. We measured and compared the CO2 and H2O vapor fluxes in California’s semi-arid oak-grass savanna and restored wetlands. ET and its partitioned terms, Transpiration (Tr) and Evaporation (Ev), were estimated.  Our data show that annual precipitation at the two study sites was 526±226 mm during the study period, but annual transpiration at the wetland site was up to 700-800 mm, two folds of that at the dry savanna. Photosynthesis and transpiration were coupling in wet conditions but decoupling in drier conditions. Water stress caused decreased GPP but increased WUE value since Tr dropped even faster. The analysis showed that transpiration was more sensitive to water deficit than photosynthesis.  This study provides insights into ecosystem carbon and water interactions, which will improve the understanding of ecosystem functions and their responses to changing climate.



Continuous observation of canopy water content changes with GPS sensors

Vincent Humphrey1,2, Brian L. Dorsey3, Christian Frankenberg1,4

1California Institute of Technology, Pasadena, CA, USA
2University of Zurich, Zurich, Switzerland
3The Huntington Library, Art Museum and Botanical Gardens, San Marino, CA, USA
4Jet Propulsion Laboratory, Pasadena, CA, USA

We present an experimental technique based on Global Navigation Satellite Systems (GNSS) to monitor in situ vegetation optical depth and canopy water content. Because GNSS microwave signals are obstructed and scattered by forest canopies, placing a GNSS sensor in a forest and measuring changes in signal quality can provide continuous information on canopy water content and forest structure. Using 8-months of measurements at our experimental site, we show how variations in GNSS signal attenuation reflect changes in both canopy structure and water content. Of particular interest, this technique appears sensitive enough to resolve the diurnal cycle of canopy water content. The rainfall and dew events captured during the observational record also show that canopy water interception (and dry-downs) can be monitored continuously. We discuss future strategies and requirements for deploying such cheap and practical systems at existing eddy-covariance sites and illustrate recent deployments at three U.S. and Swiss sites.



Soil moisture thresholds explain a shift from light-limited to water-limited sap velocity in the Central Amazon during the 2015-16 El Niño drought

Lin Meng1*, Jeffrey Chambers1,2, Charles Koven1, Gilberto Pastorello3, Bruno Gimenez4,8, Kolby Jardine1, Yao Tang1, Nate McDowell5, Robinson Negron-Juarez1, Marcos Longo1, Alessandro Araujo6, Clarissa Fontes7, Midhun Mohan1,2, Niro Higuchi4

1Climate and Ecosystem Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
2Department of Geography, University of California Berkeley, Berkeley, CA, USA
3Computational Research Division, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
4National Institute of Amazonian Research (INPA), Manaus, Brazil
5Earth Systems Science Division, Pacific Northwest National Laboratory, Richland, WA, USA
6Embrapa Amazônia Oriental, Tv. Dr. Enéas Piheiro, s/n, Marco, CEP 66095-903, Caixa postal 48, Belém, Pará, Brasil
7Department of Integrative Biology, University of California, Berkeley, Berkeley, CA, USA
8Center for Tropical Forest Science, Smithsonian Tropical Research Institute (STRI), Gamboa, Republic of Panama

Transpiration from rainforests was previously considered to be light- but not water-limited in the Central Amazon, due to abundant soil water even during the dry seasons. The 2015-16 El Niño drought provided a unique opportunity to examine whether transpiration is constrained by water under severe lack of rainfall. We examined the controls of net radiation and volumetric soil water content on sap velocity in the Central Amazon during 2015-2016, and used partial correlation analysis and multiple linear regression to identify the most likely drivers of sap velocity variability. We further identified critical thresholds of soil moisture limitation on sap velocity using a moving window approach. Our analysis showed that sap velocity and its variability experienced dramatic drops during the drought up to 56.8% and 87.8%, respectively, compared to during the wet season. Such changes were accompanied by a marked decline in soil moisture by 31.6% from 0.38 m3/m3 to 0.26 m3/m3. Sap velocity was largely limited by net radiation during wet and dry seasons, however, shifted to be limited only by soil moisture during the drought. We also found a significant interaction between net radiation and soil moisture on sap velocity during wet and early dry seasons in 2016 (P < 0.05), but not during the drought. Soil moisture thresholds on sap velocity were identified to be 0.31 m3/m3 ~ 0.33 m3/m3 (-41 ~ -73 kPa in soil matric potential), below which strong water stress starts to occur. Our results identified a strong soil moisture limitation on rainforest transpiration in the Central Amazon during extremely dry conditions, suggesting such limitation will likely become more frequent under future climate, due to increased surface temperatures combined with an increased frequency, intensity, duration and extent of extreme drought events.



Modeling Monthly ET by Ecosystem Type for Applications in Mapping Water Yield and GPP at Large Scales

Ge Sun and collaborators

USDA Forest Service

Water yield and GPP are key ecosystem services that of interest to scientists and land managers. Using data from the flux networks, we have developed a series of empirical ET models by ecosystem types require few input variables (PET, Precip, LAI) and the ET model is embedded in a watershed water balance model (WaSSI) to estimate water yield at the monthly time step. GPP is estimated as a linear function of ET. Similarly, ecosystem respiration and NEE are modeled as a linear function of GPP. We have validated and applied this simple model in several countries to map regional patterns of ET, water yield, and GPP. We will discuss success and lessons learned.



Resolve the continuous diurnal cycle of high-resolution ECOSTRESS Evapotranspiration and Land Surface Temperature

Jiaming Wen1*, Joshua B. Fisher2, Nicholas C. Parazoo3, Leiqiu Hu4, Marcy E. Litvak5, Ying Sun1

1School of Integrative Plant Science, Soil and Crop Sciences Section, Cornell University, Ithaca, NY, USA.
2Joint Institute for Regional Earth System Science and Engineering (JIFRESSE), University of California, Los Angeles (UCLA), Los Angeles, CA, USA.
3Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA.
4Department of Atmospheric and Earth Science, University of Alabama in Huntsville, AL, USA
5Department of Biology, University of New Mexico, Albuquerque, NM, USA

Remotely sensed evapotranspiration (ET) is well established, yet a diurnally-resolved product with high spatial resolution (<100 m) is still lacking, which is critically needed for agricultural and ecosystem monitoring but impossible with existing technology. The ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) provides, for the first time, land surface temperature (LST) and ET at 70m spatial resolution with diurnal samplings, enabling a promising opportunity to fill this data gap. This study developed the first framework to resolve the full diurnal cycle of LST and ET from temporally sporadic ECOSTRESS measurements. We first constructed the diurnally-resolved 70m ECOSTRESS LST using a diurnal temperature cycle (DTC) model in conjunction with Geostationary Operational Environmental Satellite (GOES) LST. Next, we derived the 70m ECOSTRESS diurnal ET from that of LST, along with ancillary meteorological and surface reflectance datasets, using the Priestley‐Taylor Jet Propulsion Laboratory (PT-JPL) algorithm. Our diurnally-resolved LST and ET successfully reproduced the spatial variation exhibited in the native ECOSTRESS measurements during overpasses (correlation coefficient r>0.96 for LST, r>0.99 for ET). Furthermore, the constructed time series captured the in situ diurnal variation in LST (r=0.91-0.99) and ET (r=0.47-0.70) measured at a semi-arid grassland flux tower (US-Seg) across distinct phenological stages. In addition, we tested our framework under different weather conditions, and found overall great agreement under clear sky, but degraded performance on cloudy/rainy days due to reduced data availability/quality as well as modelling bias. Finally, caveats and future refinement of the framework were discussed. This pilot study sets the stage for testing and applying our framework to broader climates and biome types towards eventually generating diurnally-resolved 70m global operational LST and ET products, holding great potential in enhancing ecological and agricultural applications.



Partitioning Evapotranspiration using satellite images and standard micrometeorological observations to quantify productive and non-productive water use in wheat spring crops

Orlando Ramírez-Valle1, Hugo A. Gutiérrez-Jurado1, Arturo Mendez-Barroso2, Enrico Yepez-Gonzalez2

1Earth, Environmental, and Resource Sciences Department, The University of Texas at El Paso, El Paso, TX, USA
2Department of Water Sciences and the Environment, Instituto Tecnológico de Sonora, Ciudad Obregón, México

Crop water use varies considerably in drylands, where limited water availability constrains farming activities and irrigation planning is essential for the efficient water use and long-term sustainability of agriculture. Currently, furrow or flood irrigation is the most widely used method of water delivery in agriculture. This is a highly inefficient irrigation practice in which more water is supplied than that used by the crops. To improve this irrigation practice, knowledge of the period of productive vs non-productive crop water use during the agricultural cycle can help determine irrigation amounts and timing to conserve water. In this work, we present a method to estimate Evapotranspiration (ET) and its partition into Evaporation (E) and Transpiration (Tr) as a function of the phenological evolution of the crop, taking into consideration, the prevailing meteorological conditions in the area and the energy balance of the parcel. We verify our ET estimates with independently measured ET using Eddy Covariance (ECET), satellite derived ET from the METRIC algorithm and point measurements of E and Tr using the Maximum Entropy Production method. The study site is located in the Experimental Center for Technology Transfer of the Sonoran Institute of Technology (ITSON), Mexico. Using images derived from Landsat 8 and the METRIC algorithm during the spring agricultural cycle (February to April 2019), the crop coefficient (Kc) and the actual evapotranspiration (ETa) was obtained. Daily ETa values between the 16-day Landsat image intervals, were estimated by interpolating Kc and daily reference Evapotranspiration (ET0) derived from in situ weather station data using Penman-Monteith. ET partition was achieved by separating daily Transpiration estimates from the daily ETa. Transpiration was obtained from daily ET0 and a dynamic basal crop coefficient (Kcb) calculated from Sentinel 2 vegetation indices and adjusting them by the vegetation fraction derived from pixel unmixing techniques. Our results showed good correlation between observed (ECET) and estimated ET from both METRIC (r2~0.87) and MEP (r2~0.85). The performance of our ET partitioning scheme was tested against these validated datasets yielding high r2 (>0.9) for both E and Tr, building confidence in this method. The new approach proposed here opens the possibility to estimate in near-real-time the productive (Transpiration) vs unproductive (Evaporation) use of water by crops in any location where there are micrometeorological datasets available.



Unvailing the performance of remote sensing-based ET models across uncharted South American ecoregions

Davi Melo1, Jamil Anache2, Valéria Borges3, Rodolfo Nóbrega4

1Federal University of Paraíba
2Federal University of Mato Grosso do Sul
3Federal University of Paraíba
4Imperial College London

Many satellite-based actual evapotranspiration (ET) algorithms have been proposed in the past decades and evaluated using flux tower data, mainly over North America and Europe. Model evaluation across South America has been done locally or using only a single algorithm at a time. Here, we provide the first evaluation of multiple remote sensing ET models, at a daily scale, across a wide variety of biomes, climate zones, and land uses in South America. We used meteorological data from 25 flux towers to force four remote sensing based ET models: Priestley–Taylor Jet Propulsion Laboratory (PT-JPL), Global Land Evaporation Amsterdam Model (GLEAM), Penman–Monteith Mu model (PM-MOD), and Penman–Monteith Nagler model (PM-VI). ET was predicted satisfactorily by all four models, with correlations consistently higher (R²>0.6) for GLEAM and PT-JPL, and PM-MOD and PM-VI presenting overall better responses in terms of PBIAS (-10<PBIAS<10%). As for PM-VI, this outcome is expected, given the model requires calibration with local data. Model skill seems to be unrelated to land-use type but instead presented a certain level of dependency on biome and climate, with the models producing the best results for wet to moderately wet environments. Our findings show the suitability of individual models for a number of combinations of land cover types, biomes, and climates. At the same time, no model outperformed the other for all conditions, and all models presented poor skills for sites in certain conditions (e.g. Polar Tundra climate), which emphasizes the need of adapting individual algorithms to take into account intrinsic characteristics of climates and ecosystems in South America.



A Comparison of Feature Sets and Machine Learning Models in Evapotranspiration Partitioning

Adam Stapleton1, Elke Eichelmann2, Mark Roantreek1

1School of Computing, Dublin City University, Collins Ave Ext, Whitehall, Dublin 9, Ireland
2School of Biology and Environmental Science, Science Centre West, University College Dublin, Belfield, Dublin 4, Ireland

A better understanding of the drivers of evapotranspiration and the modelling of each of its constituent parts, evaporation and transpiration, could be of significant importance to the monitoring and management of water resources globally over the coming decades. In this work, we build upon recent efforts in machine learning approaches to the problem of partitioning evapotranspiration by comparing feature selection strategies and machine learning models to deliver better predictive models for the partitioning process. Our experiments use 3 separate feature sets across 4 wetland sites as input into various machine learning models including parametric and non-parametric regressions, neural networks, support vector machines, decision trees and boosted ensembles.