Abstract:Estimating ecosystem evapotranspiration (ET) is important to understanding the global water cycle and to study land-atmosphere interactions. We developed a physics constrained machine learning (ML) model (hybrid model) to estimate latent heat flux (LE), which conserves the surface energy budget. By comparing model predictions with observations at 82 eddy covariance tower sites, our hybrid model shows similar performance to the pure ML model in terms of mean metrics (e.g., mean absolute percent errors) but, imp… Show more
“…The satellite-based model simulations showed overall higher accuracy, comparing with previous Tibetan Plateau permafrost maps generated using ground-based measurements (Figure 3, adapted from Zheng et al, 2020). In addition to the physics-based scheme, machine learning and artificial intelligence have also obtained reliable performance in solving the surface energy budget (Adnan et al, 2017;Zhao et al, 2019) as well as in regional permafrost mapping (Pastick et al, 2015;Aalto et al, 2018). A hybrid approach that combines machine learning-based surface energy balance and process-based heat-water coupled processes seems to be a potentially useful tool for regional mapping of Tibetan Plateau permafrost, while such studies are still lacking.…”
Section: Integrate Remote Sensing Data With Process-based Models To Imentioning
Recent climate change has induced widespread soil thawing and permafrost degradation in the Tibetan Plateau. Significant advances have been made in better characterizing Tibetan Plateau soil freeze/thaw dynamics, and their interaction with local-scale ecohydrological processes. However, factors such as sparse networks of in-situ sites and short observational period still limit our understanding of the Tibetan Plateau permafrost. Satellite-based optical and infrared remote sensing can provide information on land surface conditions at high spatial resolution, allowing for better representation of spatial heterogeneity in the Tibetan Plateau and further infer the related permafrost states. Being able to operate at “all-weather” conditions, microwave remote sensing has been widely used to retrieve surface soil moisture, freeze/thaw state, and surface deformation, that are critical to understand the Tibetan Plateau permafrost state and changes. However, coarse resolution (>10 km) of current passive microwave sensors can add large uncertainties to the above retrievals in the Tibetan Plateau area with high topographic relief. In addition, current microwave remote sensing methods are limited to detections in the upper soil layer within a few centimetres. On the other hand, algorithms that can link surface properties and soil freeze/thaw indices to permafrost properties at regional scale still need improvements. For example, most methods using InSAR (interferometric synthetic aperture radar) derived surface deformation to estimate active layer thickness either ignore the effects of vertical variability of soil water content and soil properties, or use site-specific soil moisture profiles. This can introduce non-negligible errors when upscaled to the broader Tibetan Plateau area. Integrating satellite remote sensing retrievals with process models will allow for more accurate representation of Tibetan Plateau permafrost conditions. However, such applications are still limiting due to a number of factors, including large uncertainties in current satellite products in the Tibetan Plateau area, and mismatch between model input data needs and information provided by current satellite sensors. Novel approaches to combine diverse datasets with models through model initialization, parameterization and data assimilation are needed to address the above challenges. Finally, we call for expansion of local-scale observational network, to obtain more information on deep soil temperature and moisture, soil organic carbon content, and ground ice content.
“…The satellite-based model simulations showed overall higher accuracy, comparing with previous Tibetan Plateau permafrost maps generated using ground-based measurements (Figure 3, adapted from Zheng et al, 2020). In addition to the physics-based scheme, machine learning and artificial intelligence have also obtained reliable performance in solving the surface energy budget (Adnan et al, 2017;Zhao et al, 2019) as well as in regional permafrost mapping (Pastick et al, 2015;Aalto et al, 2018). A hybrid approach that combines machine learning-based surface energy balance and process-based heat-water coupled processes seems to be a potentially useful tool for regional mapping of Tibetan Plateau permafrost, while such studies are still lacking.…”
Section: Integrate Remote Sensing Data With Process-based Models To Imentioning
Recent climate change has induced widespread soil thawing and permafrost degradation in the Tibetan Plateau. Significant advances have been made in better characterizing Tibetan Plateau soil freeze/thaw dynamics, and their interaction with local-scale ecohydrological processes. However, factors such as sparse networks of in-situ sites and short observational period still limit our understanding of the Tibetan Plateau permafrost. Satellite-based optical and infrared remote sensing can provide information on land surface conditions at high spatial resolution, allowing for better representation of spatial heterogeneity in the Tibetan Plateau and further infer the related permafrost states. Being able to operate at “all-weather” conditions, microwave remote sensing has been widely used to retrieve surface soil moisture, freeze/thaw state, and surface deformation, that are critical to understand the Tibetan Plateau permafrost state and changes. However, coarse resolution (>10 km) of current passive microwave sensors can add large uncertainties to the above retrievals in the Tibetan Plateau area with high topographic relief. In addition, current microwave remote sensing methods are limited to detections in the upper soil layer within a few centimetres. On the other hand, algorithms that can link surface properties and soil freeze/thaw indices to permafrost properties at regional scale still need improvements. For example, most methods using InSAR (interferometric synthetic aperture radar) derived surface deformation to estimate active layer thickness either ignore the effects of vertical variability of soil water content and soil properties, or use site-specific soil moisture profiles. This can introduce non-negligible errors when upscaled to the broader Tibetan Plateau area. Integrating satellite remote sensing retrievals with process models will allow for more accurate representation of Tibetan Plateau permafrost conditions. However, such applications are still limiting due to a number of factors, including large uncertainties in current satellite products in the Tibetan Plateau area, and mismatch between model input data needs and information provided by current satellite sensors. Novel approaches to combine diverse datasets with models through model initialization, parameterization and data assimilation are needed to address the above challenges. Finally, we call for expansion of local-scale observational network, to obtain more information on deep soil temperature and moisture, soil organic carbon content, and ground ice content.
“…Whitley et al, 2013). These approaches can also be used to identify the relative importance of different hydrometeorological drivers of transpiration (Zhao et al, 2019), or to produce global transpiration maps, by combining SAPFLUXNET with other data (Jung et al, 2019). This upscaling of stand transpiration to large areas will also 21 https://doi.org/10.5194/essd-2020-227 allow addressing broader questions at the regional and continental scale, such as the role of transpiration in moisture recycling (Staal et al, 2018).…”
Abstract. Plant transpiration links physiological responses of vegetation to water supply and demand with hydrological, energy and carbon budgets at the land-atmosphere interface. However, despite being the main land evaporative flux at the global scale, transpiration and its response to environmental drivers are currently not well constrained by observations. Here we introduce the first global compilation of whole-plant transpiration data from sap flow measurements (SAPFLUXNET, https://sapfluxnet.creaf.cat/). We harmonised and quality-controlled individual datasets supplied by contributors worldwide in a semi-automatic data workflow implemented in the R programming language. Datasets include sub-daily time series of sap flow and hydrometeorological drivers for one or more growing seasons, as well as metadata on the stand characteristics, plant attributes and technical details of the measurements. SAPFLUXNET contains 202 globally distributed datasets with sap flow time series for 2714 plants, mostly trees, of 174 species. SAPFLUXNET has a broad bioclimatic coverage, with woodland/shrubland and temperate forest biomes especially well-represented (80 % of the datasets). The measurements cover a wide variety of stand structural characteristics and plant sizes. The datasets encompass the period between 1995 and 2018, with 50 % of the datasets being at least 3 years long. Accompanying radiation and vapour pressure deficit data are available for most of the datasets, while on-site soil water content is available for 56 % of the datasets. Many datasets contain data for species that make up 90 % or more of the total stand basal area, allowing the estimation of stand transpiration in diverse ecological settings. SAPFLUXNET adds to existing plant trait datasets, ecosystem flux networks and remote sensing products to help increase our understanding of plant water use, plant responses to drought and ecohydrological processes. SAPFLUXNET version 0.1.5 is freely available from the Zenodo repository ( https://doi.org/10.5281/zenodo.3971689, Poyatos et al., 2020a). The sapfluxnetr R package, designed to access, visualise and process SAPFLUXNET data is available from CRAN.
“…A pure artificial neural network (ANN) was proven to have good performance in retrieving land surface fluxes, or in some cases, even better performance than that of hybrid models (Chen et al, 2020;Haughton et al, 2018;Zhao et al, 2019). In this study, we trained a multi-layer feedforward neural network model that consisted of an input layer, hidden layers, and an output layer to predict daily λE and H at the globally distributed weather stations.…”
Section: Artificial Neural Network Model Trainingmentioning
Abstract. Evapotranspiration (ET) accompanied by water and heat transport in the hydrological cycle is a key component in regulating surface aridity. Existing studies on changes in surface aridity have typically estimated ET using semi-empirical equations or parameterizations of land surface processes, which are based on the assumption that the parameters in the equation are stationary. However, plant physiological effects and its response to a changing environment are dynamically modifying ET, thereby challenging this assumption and limiting the estimation of long-term ET. In this study, the latent heat flux (ET in energy units) and sensible heat flux were retrieved for recent decades on a global scale using machine learning approach and driven by ground-based observations from flux towers and weather stations. The study resulted in several findings, namely that the evaporative fraction (EF) – the ratio of latent heat flux to available surface energy – exhibited a relatively decreasing trend on fractional land surfaces; In particular, the decrease in EF was accompanied by an increase in long-term runoff as assessed by precipitation (P) minus ET, accounting for 27.06 % of the global land areas. The signs were indicative of reduced surface conductance, which further emphasized that land-surface vegetation has major impacts on regulating the water and energy cycles, as well as aridity variability.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.