Gap Filling and Quality Assessment of CO2 and Water Vapour Fluxes above an Urban Area with Radial Basis Function Neural Networks

Vertical turbulent fluxes of water vapour, carbon dioxide, and sensible heat were measured from 16 August to the 28 September 2006 near the city centre of Münster in north-west Germany. In comparison to results of measurements above homogeneous ecosystem sites, the CO2 fluxes above the urban investigation area showed more peaks and higher variances during the course of a day, probably caused by traffic and other varying, anthropogenic sources. The main goal of this study is the introduction and establishment of a new gap filling procedure using radial basis function (RBF) neural networks, which is also applicable under complex environmental conditions. We applied adapted RBF neural networks within a combined modular expert system of neural networks as an innovative approach to fill data gaps in micrometeorological flux time series. We found that RBF networks are superior to multi-layer perceptron (MLP) neural networks in the reproduction of the highly variable turbulent fluxes. In addition, we enhanced the methodology in the field of quality assessment for eddy covariance data. An RBF neural network mapping system was used to identify conditions of a turbulence regime that allows reliable quantification of turbulent fluxes through finding an acceptable minimum of the friction velocity. For the data analysed in this study, the minimum acceptable friction velocity was found to be 0.15  m s−1. The obtained CO2 fluxes, measured on a tower at 65 m a.g.l., reached average values of 12 μmol m−2 s−1 and fell to nighttime minimum values of 3 μmol m −2 s−1. Mean daily CO2 emissions of 21 g CO2 m−2d −1 were obtained during our 6-week experiment. Hence, the city centre of Münster appeared to be a significant source of CO2. The half-hourly average values of water vapour fluxes ranged between 0.062 and 0.989 mmol m−2 s−1and showed lower variances than the simultaneously measured fluxes of CO2.

[1]  Simon Haykin,et al.  Neural Networks: A Comprehensive Foundation , 1998 .

[2]  William J. Massman,et al.  Handbook of micrometeorology : a guide for surface flux measurement and analysis , 2004 .

[3]  T. Poggio,et al.  Networks and the best approximation property , 1990, Biological Cybernetics.

[4]  Leo Breiman,et al.  Bagging Predictors , 1996, Machine Learning.

[5]  Jean-Pierre Lagouarde,et al.  Monitoring the Sensible Heat Flux over Urban Areas using Large Aperture Scintillometry: Case Study of Marseille City During the Escompte Experiment , 2006 .

[6]  T. Vesala,et al.  On the separation of net ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm , 2005 .

[7]  Brian D. Ripley,et al.  Pattern Recognition and Neural Networks , 1996 .

[8]  Timothy R. Oke,et al.  Flux and turbulence measurements at a densely built‐up site in Marseille: Heat, mass (water and carbon dioxide), and momentum , 2004 .

[9]  L. Cooper,et al.  When Networks Disagree: Ensemble Methods for Hybrid Neural Networks , 1992 .

[10]  Jason Beringer,et al.  Characteristics influencing the variability of urban CO2 fluxes in Melbourne, Australia , 2007 .

[11]  Thomas Foken,et al.  On the Significance of the Webb Correction to Fluxes , 2003 .

[12]  R. Valentini,et al.  A new assessment of European forests carbon exchanges by eddy fluxes and artificial neural network spatialization , 2003 .

[13]  W. Massman,et al.  Eddy covariance flux corrections and uncertainties in long-term studies of carbon and energy exchanges , 2002 .

[14]  Jed O. Kaplan,et al.  Arctic climate change with a 2 ∘C global warming: Timing, climate patterns and vegetation change , 2006 .

[15]  G. Pyle,et al.  Toxicity of uranium mine receiving waters to early life stage fathead minnows (Pimephales promelas) in the laboratory. , 2002, Environmental pollution.

[16]  W. Light Some Aspects of Radial Basis Function Approximation , 1992 .

[17]  J. N. Armor,et al.  Addressing the CO2 dilemma , 2007 .

[18]  Matthias Mauder,et al.  Innovative gap-filling strategy for annual sums of CO2 net ecosystem exchange , 2006 .

[19]  Nikos A. Vlassis,et al.  The global k-means clustering algorithm , 2003, Pattern Recognit..

[20]  Roger Taesler,et al.  A discussion of statistical methods used to estimate extreme wind speeds , 2006 .

[21]  Geoffrey E. Hinton,et al.  Learning internal representations by error propagation , 1986 .

[22]  R. M. Redheffer,et al.  SOME APPLICATIONS OF MONOTONE OPERATORS IN MARKOV PROCESSES , 1965 .

[23]  Thomas Foken,et al.  Angewandte Meteorologie : mikrometeorologische Methoden , 2003 .

[24]  Tony R. Martinez,et al.  The general inefficiency of batch training for gradient descent learning , 2003, Neural Networks.

[25]  Christopher M. Bishop,et al.  Neural networks for pattern recognition , 1995 .

[26]  Nicholas J. Cook,et al.  “Discussion on modern estimation of the parameters of the Weibull wind speed distribution for wind speed energy analysis” by J.V. Seguro, T.W. Lambert , 2001 .

[27]  Ü. Rannik,et al.  Gap filling strategies for defensible annual sums of net ecosystem exchange , 2001 .

[28]  Dennis D. Baldocchi,et al.  Objective threshold determination for nighttime eddy flux filtering , 2005 .

[29]  J. Wilczak,et al.  Sonic Anemometer Tilt Correction Algorithms , 2001 .

[30]  B. Yegnanarayana,et al.  Artificial Neural Networks , 2004 .

[31]  Erik Lundtang Petersen,et al.  The European Wind Atlas , 1985 .

[32]  Yadvinder Malhi,et al.  A Re-Evaluation of Long-Term Flux Measurement Techniques Part I: Averaging and Coordinate Rotation , 2003 .

[33]  Mathias W. Rotach,et al.  Modification of an Operational Dispersion Model for Urban Applications , 2001 .

[34]  S. P. S. Arya,et al.  Introduction to micrometeorology , 1988 .

[35]  Jacek M. Zurada,et al.  Sensitivity analysis for minimization of input data dimension for feedforward neural network , 1994, Proceedings of IEEE International Symposium on Circuits and Systems - ISCAS '94.

[36]  D. Nowak,et al.  Local-scale fluxes of carbon dioxide in urban environments: methodological challenges and results from Chicago. , 2002, Environmental pollution.

[37]  A. Barron Approximation and Estimation Bounds for Artificial Neural Networks , 1991, COLT '91.

[38]  R. Eskridge,et al.  Improved Magnus form approximation of saturation vapor pressure , 1996 .

[39]  Lars Kai Hansen,et al.  Neural Network Ensembles , 1990, IEEE Trans. Pattern Anal. Mach. Intell..

[40]  D. Baldocchi Assessing the eddy covariance technique for evaluating carbon dioxide exchange rates of ecosystems: past, present and future , 2003 .

[41]  Gene H. Golub,et al.  Singular value decomposition and least squares solutions , 1970, Milestones in Matrix Computation.

[42]  Corinne Le Quéré,et al.  Climate Change 2013: The Physical Science Basis , 2013 .

[43]  Ü. Rannik,et al.  Estimates of the annual net carbon and water exchange of forests: the EUROFLUX methodology , 2000 .

[44]  Haimin Wang,et al.  Automatic Solar Flare Detection Using MLP, RBF, and SVM , 2003 .

[45]  John Moncrieff,et al.  Averaging, Detrending, and Filtering of Eddy Covariance Time Series , 2004 .

[46]  J. Canadell,et al.  Recent patterns and mechanisms of carbon exchange by terrestrial ecosystems , 2001, Nature.

[47]  O. Hartogensis,et al.  Variance Method to Determine Turbulent Fluxes of Momentum And Sensible Heat in The Stable Atmospheric Surface Layer , 2005 .

[48]  T. Foken,et al.  Tools for quality assessment of surface-based flux measurements , 1996 .

[49]  Brunello Tirozzi Neural Networks and Sea Time Series: Reconstruction and Extreme-Event Analysis , 2005 .

[50]  Andrew E. Suyker,et al.  Gap filling strategies for long term energy flux data sets , 2001 .

[51]  Üllar Rannik,et al.  Autoregressive filtering versus linear detrending in estimation of fluxes by the eddy covariance method , 1999 .

[52]  Thomas Foken,et al.  Post-Field Data Quality Control , 2004 .

[53]  D. Broomhead,et al.  Radial Basis Functions, Multi-Variable Functional Interpolation and Adaptive Networks , 1988 .

[54]  Dale Schuurmans,et al.  Automatic basis selection techniques for RBF networks , 2003, Neural Networks.

[56]  E. K. Webb,et al.  Correction of flux measurements for density effects due to heat and water vapour transfer , 1980 .

[57]  M. Reichstein,et al.  Net regional ecosystem CO2 exchange from airborne and ground‐based eddy covariance, land‐use maps and weather observations , 2007 .

[58]  Magnus R. Hestenes,et al.  Conjugate Direction Methods in Optimization , 1980 .

[59]  Gene H. Golub,et al.  Calculating the singular values and pseudo-inverse of a matrix , 2007, Milestones in Matrix Computation.

[60]  Y. Malhi,et al.  Gap-filling measurements of carbon dioxide storage in tropical rainforest canopy airspace , 2005 .

[61]  I. Jolliffe Principal Component Analysis , 2002 .

[62]  Zhou Zhi,et al.  Neural Network Ensemble , 2002 .

[63]  Matthias Roth,et al.  Temporal dynamics of CO2 fluxes and profiles over a Central European city , 2006 .

[64]  R. Monson,et al.  Gap-filling missing data in eddy covariance measurements using multiple imputation (MI) for annual estimations , 2004 .

[65]  Edwin Metcalfe,et al.  Optimisation of radial basis and backpropagation neural networks for modelling auto-ignition temperature by quantitative-structure property relationships , 1996 .

[66]  Man-Wai Mak,et al.  Application of MLP and RBF networks to cloud detection , 2001, Proceedings of 2001 International Symposium on Intelligent Multimedia, Video and Speech Processing. ISIMP 2001 (IEEE Cat. No.01EX489).

[67]  M. Bahn,et al.  Eddy covariance measurements of carbon dioxide, latent and sensible energy fluxes above a meadow on a mountain slope , 2007, Boundary-layer meteorology.

[68]  Hal Westberg,et al.  Measurements of CO2 fluxes from the Mexico City urban landscape , 2005 .

[69]  J. William Munger,et al.  Measurements of carbon sequestration by long‐term eddy covariance: methods and a critical evaluation of accuracy , 1996 .

[70]  Arden L. Buck,et al.  New Equations for Computing Vapor Pressure and Enhancement Factor , 1981 .

[71]  Andrew Hunter,et al.  Application of neural networks and sensitivity analysis to improved prediction of trauma survival , 2000, Comput. Methods Programs Biomed..

[72]  Eric A. Davidson,et al.  Spatial and temporal variability in forest–atmosphere CO2 exchange , 2004 .

[73]  Ryuichi Hirata,et al.  Comparisons of gap-filling methods for carbon flux dataset: A combination of a genetic algorithm and an artificial neural network , 2006 .

[74]  Steven R. Hanna,et al.  Observed winds, turbulence, and dispersion in built-up downtown areas of Oklahoma City and Manhattan , 2007 .

[75]  V. Stauch,et al.  A semi‐parametric gap‐filling model for eddy covariance CO2 flux time series data , 2006 .

[76]  Tom Heskes,et al.  Clustering ensembles of neural network models , 2003, Neural Networks.