Modeling urban evapotranspiration using remote sensing, flux footprints, and artificial intelligence

Vulova, Stenka; Meier, Fred; Rocha, Alby Duarte; Quanz, Justus; Nouri, Hamideh; Kleinschmit, Birgit

doi:10.1016/j.scitotenv.2021.147293

Cited by 46 publications

(17 citation statements)

References 62 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Vulova et al (2021) also mentioned that climate change has caused water scarcity and urban heat island problems in urban areas. In this regard, they focused on Evapotranspiration (ET) as a critical factor in urban greening projects, which is essential for alleviating these problems.…”

Section: Resultsmentioning

confidence: 99%

Applications of machine learning and deep learning methods for climate change mitigation and adaptation

Ladi

Jabalameli

Sharifi

2022

Environment and Planning B: Urban Analytics and City Science

View full text Add to dashboard Cite

Climate change is a global issue that must be considered and addressed immediately. Many articles have been published on climate change mitigation and adaptation. However, new methods are required to explore the complexities of climate change and provide more efficient and effective adaptation and mitigation policies. With the advancement of technology, machine learning (ML) and deep learning (DL) methods have gained considerable popularity in many fields, including climate change. This paper aims to explore the most popular ML and DL methods that have been applied for climate change mitigation and adaptation. Another aim is to determine the most common mitigation and adaptation measures/actions in general, and in urban areas in particular, that have been studied using ML and DL methods. For this purpose, word frequency analysis and topic modeling, specifically the Latent Dirichlet allocation (LDA) as a ML algorithm, are used in this study. The results indicate that the most popular ML technique in both climate change mitigation and adaptation is the Artificial Neural Network. Moreover, among different research areas related to climate change mitigation and adaptation, geoengineering, and land surface temperature are the ones that have used ML and DL algorithms the most.

show abstract

Section: Resultsmentioning

confidence: 99%

Applications of machine learning and deep learning methods for climate change mitigation and adaptation

Ladi

Jabalameli

Sharifi

2022

Environment and Planning B: Urban Analytics and City Science

View full text Add to dashboard Cite

show abstract

“…The two eddy covariance (EC) flux towers are operated by the Chair of Climatology at the Technische Universität Berlin (TUB) as part of the Urban Climate Observatory (UCO) Berlin (Scherer et al, 2019;Vulova et al, 2021). The EC measurement system is based on an open-path gas analyser and a three-dimensional sonic anemometerthermometer (IRGASON, Campbell Scientific).…”

Section: Eddy Covariance Flux Towersmentioning

confidence: 99%

“…As urban ET observations are rare, costly and available only for a few cities in the world, an alternative is to estimate ET using process-based or empirical models. Fitting classic empirical models or machine-learning algorithms is relatively common in natural landscapes but relatively scarce in an urban environment (Vulova et al, 2021;Wang et al, 2020). One reason is the necessity to train the model at representative locations and conditions, which is a challenge in urban areas due to the constantly changing land cover captured by the tower's footprint and lack of flux towers at different surfaces in a highly fragmented and heterogeneous environment (Feigenwinter et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

Modelling hourly evapotranspiration in urban environments with SCOPE using open remote sensing and meteorological data

Rocha¹,

Vulova²,

Tol

et al. 2022

Hydrol. Earth Syst. Sci.

Self Cite

View full text Add to dashboard Cite

Abstract. Evapotranspiration (ET) is a fundamental variable for assessing water balance and the urban heat island (UHI) effect. Terrestrial ET is deeply dependent on the land cover as it derives mainly from soil evaporation and plant transpiration. The majority of well-known process-based models based on the Penman–Monteith equation focus on the atmospheric interfaces (e.g. radiation, temperature and humidity), lacking explicit input parameters to precisely describe vegetation and soil properties. The model soil-canopy-observation of photosynthesis and energy fluxes (SCOPE) accounts for a broad range of surface–atmosphere interactions to predict ET. However, like most modelling approaches, SCOPE assumes a homogeneous vegetated landscape to estimate ET. As urban environments are highly fragmented, exhibiting a mix of vegetated and impervious surfaces, we propose a two-stage modelling approach to capture most of the spatiotemporal variability of ET without making the model overly complex. After predicting ET using the SCOPE model, the bias caused by the assumption of homogeneous vegetation is corrected using the vegetation fraction extracted by footprint modelling. Two urban sites equipped with eddy flux towers presenting different levels of vegetation fraction and imperviousness located in Berlin, Germany, were used as study cases. The correction factor for urban environments increased the model accuracy significantly, reducing the relative bias in ET predictions from 0.74 to 0.001 and 2.20 to −0.13 for the two sites considering the SCOPE model with remote sensing-derived inputs. Model errors (RMSE) were considerably reduced in both sites, from 0.061 to 0.026 and 0.100 to 0.021, while the coefficient of determination (R2) remained similar after correction, 0.82 and 0.47, respectively. The novelty of this study is to provide hourly ET predictions combining the temporal dynamics of ET in a natural environment with the spatially fragmented land cover in urban environments at a low computational cost. All model inputs are open data and available globally for most medium-sized and large cities. This approach can provide ET maps in different temporal resolutions to better manage vegetation in cities in order to mitigate the UHI effect and droughts.

show abstract

“…The two eddy covariance flux towers are operated by the Chair of Climatology at the Technische Universität Berlin (TUB) as part of the Urban Climate Observatory (UCO) (Scherer et al, 2019;Vulova et al, 2021). The flux towers measure turbulent fluxes of sensible and latent heat based on the EC method.…”

Section: Eddy Covariance Flux Towersmentioning

confidence: 99%

“…CC BY 4.0 License. rare in an urban environment (Vulova et al, 2021;Wang et al, 2020). One reason is the necessity to train the model at representative locations and conditions, which is a challenge in urban areas due the constantly changing land cover captured by the tower's footprint and lack of flux towers at different surfaces in a highly fragmented and heterogeneous environment (Feigenwinter et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

Modelling hourly evapotranspiration in urban environments with SCOPE using open remote sensing and meteorological data

Rocha¹,

Vulova²,

Tol

et al. 2021

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. Evapotranspiration (ET) is a fundamental variable to assess water balance and urban heat island effect. ET is deeply dependent on the land cover as it derives mainly from the processes of soil evaporation and plant transpiration. The majority of well-known process-based models based on the Penman-Monteith equation focus on the atmospheric interfaces (e.g. radiation, temperature and humidity), lacking explicit input parameters to describe the land surface. The model Soil-Canopy-Observation of Photosynthesis and Energy fluxes (SCOPE) accounts for a broad range of surface-atmosphere interactions to predict ET. However, like most modelling approaches, SCOPE assumes a homogeneous vegetated landscape to estimate ET. Urban environments are highly fragmented, exhibiting a blend of pervious and impervious anthropogenic surfaces. Whereas, high-resolution remote sensing (RS) and detailed GIS information to characterise land surfaces is usually available for major cities. Data describing land surface properties were used in this study to develop a method to correct bias in ET predictions caused by the assumption of homogeneous vegetation by process-based models. Two urban sites equipped with eddy flux towers presenting different levels of vegetation fraction and imperviousness located in Berlin, Germany, were used as study cases. The correction factor for urban environments has increased model accuracy significantly, reducing the relative bias in ET predictions from 0.74 to −0.001 and 2.20 to −0.13 for the two sites, respectively, considering the SCOPE model using RS data. Model errors (i.e. RMSE) were also considerably reduced in both sites, from 0.061 to 0.026 and 0.100 to 0.021, while the coefficient of determination (R2) remained similar after the correction, 0.82 and 0.47, respectively. This study presents a novel method to predict hourly urban ET using freely available RS and meteorological data, independently from the flux tower measurements. The presented method can support actions to mitigate climate change in urban areas, where most the world population lives.

show abstract

Modeling urban evapotranspiration using remote sensing, flux footprints, and artificial intelligence

Cited by 46 publications

References 62 publications

Applications of machine learning and deep learning methods for climate change mitigation and adaptation

Applications of machine learning and deep learning methods for climate change mitigation and adaptation

Modelling hourly evapotranspiration in urban environments with SCOPE using open remote sensing and meteorological data

Modelling hourly evapotranspiration in urban environments with SCOPE using open remote sensing and meteorological data

Contact Info

Product

Resources

About