Jorge Mateu, Department of Mathematics of the University Jaume I of Castellon, Spain,Werner G. Müller, Department of Applied Statistics, Johannes Kepler University Linz, Austria.

Select Guide Rating

• Contributors xvForeword xix1 Collecting spatio-temporal data 1Jorge Mateu and Werner G. Müller1.1 Introduction 11.2 Paradigms in spatio-temporal design 21.3 Paradigms in spatio-temporal modeling 31.4 Geostatistics and spatio-temporal random functions 41.4.1 Relevant spatio-temporal concepts 41.4.2 Properties of the spatio-temporal covariance and variogram functions 61.4.3 Spatio-temporal kriging 81.4.4 Spatio-temporal covariance models 101.4.5 Parametric estimation of spatio-temporal covariograms 111.5 Types of design criteria and numerical optimization 131.6 The problem set: Upper Austria 171.6.1 Climatic data 171.6.2 Grassland usage 181.7 The chapters 23Acknowledgments 28References 282 Model-based frequentist design for univariate and multivariate geostatistics 37Dale L. Zimmerman and Jie li2.1 Introduction 372.2 Design for univariate geostatistics 382.2.1 Data-model framework 382.2.2 Design criteria 382.2.3 Algorithms 422.2.4 Toy example 422.3 Design for multivariate geostatistics 452.3.1 Data-model framework 452.3.2 Design criteria 472.3.3 Toy example 482.4 Application: Austrian precipitation data network 502.5 Conclusions 52References 533 Model-based criteria heuristics for second-phase spatial sampling 54Eric M. Delmelle3.1 Introduction 543.2 Geometric and geostatistical designs 563.2.1 Efficiency of spatial sampling designs 563.2.2 Sampling spatial variables in a geostatistical context 573.2.3 Sampling designs minimizing the kriging variance 583.3 Augmented designs: Second-phase sampling 593.3.1 Additional sampling schemes to maximize change in the kriging variance 593.3.2 A weighted kriging variance approach 603.4 A simulated annealing approach 633.5 Illustration 653.5.1 Initial sampling designs 663.5.2 Augmented designs 683.6 Discussion 68References 694 Spatial sampling design by means of spectral approximations to the error process 72Gunter Spöck and Jürgen Pilz4.1 Introduction 724.2 A brief review on spatial sampling design 754.3 The spatial mixed linear model 764.4 Classical Bayesian experimental design problem 774.5 The Smith and Zhu design criterion 794.6 Spatial sampling design for trans-Gaussian kriging 814.7 The spatDesign toolbox 824.7.1 Covariance estimation and variography software 834.7.2 Spatial interpolation and kriging software 844.7.3 Spatial sampling design software 854.8 An example session 894.8.1 Preparatory calculations 894.8.2 Optimal design for the BSLM 934.8.3 Design for the trans-Gaussian kriging 944.9 Conclusions 98References 995 Entropy-based network design using hierarchical Bayesian kriging 103Baisuo Jin, Yuehua Wu and Baiqi Miao5.1 Introduction 1035.2 Entropy-based network design using hierarchical Bayesian kriging 1055.3 The data 1075.4 Spatio-temporal modeling 1075.5 Obtaining a staircase data structure 1115.6 Estimating the hyperparameters H g and the spatial correlations between gauge stations 1135.7 Spatial predictive distribution over the 445 areas located in the 18 districts of Upper Austria 1175.8 Adding gauge stations over the 445 areas located in the 18 districts of Upper Austria 1205.9 Closing down an existing gauge station 1225.10 Model evaluation 124Appendix 5.1: Hierarchical Bayesian spatio-temporal modeling (or kriging) 124Appendix 5.2: Some estimated parameters 128Acknowledgments 129References 1296 Accounting for design in the analysis of spatial data 131Brian J. Reich and Montserrat Fuentes6.1 Introduction 1316.2 Modeling approaches 1346.2.1 Informative missingness 1346.2.2 Informative sampling 1356.2.3 A two-stage approach for informative sampling 1366.3 Analysis of the Austrian precipitation data 1376.4 Discussion 139References 1417 Spatial design for knot selection in knot-based dimension reduction models 142Alan E. Gelfand, Sudipto Banerjee and Andrew O. Finley7.1 Introduction 1427.2 Handling large spatial datasets 1457.3 Dimension reduction approaches 1467.3.1 Basic properties of low rank models 1467.3.2 Predictive process models: A brief review 1487.4 Some basic knot design ideas 1497.4.1 A brief review of spatial design 1497.4.2 A strategy for selecting knots 1517.5 Illustrations 1537.5.1 A simulation example 1537.5.2 A simulation example using the two-step analysis 1597.5.3 Tree height and diameter analysis 1607.5.4 Austria precipitation analysis 1627.6 Discussion and future work 165References 1668 Exploratory designs for assessing spatial dependence 170Agnes Fussl, Werner G. Müller and Juan Rodríguez-Díaz8.1 Introduction 1708.1.1 The dataset and its visualization 1728.2 Spatial links 1748.2.1 Spatial neighbors 1758.2.2 Spatial weights 1768.3 Measures of spatial dependence 1788.4 Models for areal data 1808.4.1 H0 : A spaceless regression model 1818.4.2 H0 : Spatial regression models 1858.5 Design considerations 1908.5.1 A design criterion 1928.5.2 Example 1948.6 Discussion 195Appendix 8.1: R code 198Acknowledgments 202References 2039 Sampling design optimization for space-time kriging 207Gerard B.M. Heuvelink, Daniel A. Griffith, Tomislav Hengl and Stephanie J. Melles9.1 Introduction 2079.2 Methodology 2099.2.1 Space-time universal kriging 2099.2.2 Sampling design optimization with spatial simulated annealing 2119.3 Upper Austria case study 2129.3.1 Descriptive statistics 2129.3.2 Estimation of the space-time model and universal kriging 2159.3.3 Optimal design scenario 1 2189.3.4 Optimal design scenario 2 2199.3.5 Optimal design scenario 3 2199.4 Discussion and conclusions 221Appendix 9.1: R code 222Acknowledgment 227References 22810 Space-time adaptive sampling and data transformations 231José M. Angulo, María C. Bueso and Francisco J. Alonso10.1 Introduction 23110.2 Adaptive sampling network design 23310.2.1 A simulated illustration 23510.3 Predictive information based on data transformations 23810.4 Application to Upper Austria temperature data 24210.5 Summary 246Acknowledgments 247References 24711 Adaptive sampling design for spatio-temporal prediction 249Thomas R. Fanshawe and Peter J. Diggle11.1 Introduction 24911.2 Review of spatial and spatio-temporal adaptive designs 25111.3 The stationary Gaussian model 25311.3.1 Model specification 25311.3.2 Theoretically optimal designs 25411.3.3 A comparison of design strategies 25411.4 The dynamic process convolution model 25711.4.1 Model specification 25711.4.2 A comparison of design strategies 25811.5 Upper Austria rainfall data example 26211.6 Discussion 264Appendix 11.1 266References 26712 Semiparametric dynamic design of monitoring networks for non-Gaussian spatio-temporal data 269Scott H. Holan and Christopher K. Wikle12.1 Introduction 26912.2 Semiparametric non-Gaussian space-time dynamic design 27112.2.1 Semiparametric spatio-temporal dynamic Gamma model 27112.2.2 Simulation-based dynamic design 27412.2.3 Extended Kalman filter for dynamic gamma models 27512.2.4 Extended Kalman filter design algorithm 27712.3 Application: Upper Austria precipitation 27812.4 Discussion 282Acknowledgments 282References 28313 Active learning for monitoring network optimization 285Devis Tuia, Alexei Pozdnoukhov, Loris Foresti and Mikhail Kanevski13.1 Introduction 28513.2 Statistical learning from data 28713.2.1 Algorithmic approaches to learning 28813.2.2 Over-fitting and model selection 28813.3 Support vector machines and kernel methods 28913.3.1 Classification: SVMs 29013.3.2 Density estimation: One-class SVM 29213.3.3 Regression: Kernel ridge regression 29313.3.4 Regression: SVR 29413.4 Active learning 29413.4.1 A general framework 29513.4.2 First steps in active learning: Reducing output variance 29613.4.3 Exploration–exploitation strategies: Towards mixed approaches 29713.5 Active learning with SVMs 29713.5.1 Margin sampling 29713.5.2 Diversity of batches of samples 29913.5.3 Committees of models 29913.6 Case studies 30013.6.1 Austrian climatological data 30013.6.2 Cesium-137 concentration after Chernobyl 30413.6.3 Wind power plants sites evaluation 30713.7 Conclusions 312Acknowledgments 314References 31414 Stationary sampling designs based on plume simulations 319Kristina B. Helle and Edzer Pebesma14.1 Introduction 31914.2 Plumes: From random fields to simulations 32014.3 Cost functions 32414.3.1 Detecting plumes 32414.3.2 Mapping and characterising plumes 32514.3.3 Combined cost functions 32514.4 Optimisation 32614.4.1 Greedy search 32614.4.2 Spatial simulated annealing 32814.4.3 Genetic algorithms 32914.4.4 Other methods 33114.4.5 Evaluation and sensitivity 33114.4.6 Use case: Combination and comparison of optimisation algorithms 33214.5 Results 33414.5.1 Simulations 33414.5.2 Greedy search 33514.5.3 Sensitivity of greedy search to the plume simulations 33614.5.4 Comparison of optimisation algorithms 33714.6 Discussion 340Acknowledgments 341References 341Index 345

Jorge Mateu, Department of Mathematics of the University Jaume I of Castellon, Spain,Werner G. Müller, Department of Applied Statistics, Johannes Kepler University Linz, Austria.

Select Guide Rating

• Contributors xvForeword xix1 Collecting spatio-temporal data 1Jorge Mateu and Werner G. Müller1.1 Introduction 11.2 Paradigms in spatio-temporal design 21.3 Paradigms in spatio-temporal modeling 31.4 Geostatistics and spatio-temporal random functions 41.4.1 Relevant spatio-temporal concepts 41.4.2 Properties of the spatio-temporal covariance and variogram functions 61.4.3 Spatio-temporal kriging 81.4.4 Spatio-temporal covariance models 101.4.5 Parametric estimation of spatio-temporal covariograms 111.5 Types of design criteria and numerical optimization 131.6 The problem set: Upper Austria 171.6.1 Climatic data 171.6.2 Grassland usage 181.7 The chapters 23Acknowledgments 28References 282 Model-based frequentist design for univariate and multivariate geostatistics 37Dale L. Zimmerman and Jie li2.1 Introduction 372.2 Design for univariate geostatistics 382.2.1 Data-model framework 382.2.2 Design criteria 382.2.3 Algorithms 422.2.4 Toy example 422.3 Design for multivariate geostatistics 452.3.1 Data-model framework 452.3.2 Design criteria 472.3.3 Toy example 482.4 Application: Austrian precipitation data network 502.5 Conclusions 52References 533 Model-based criteria heuristics for second-phase spatial sampling 54Eric M. Delmelle3.1 Introduction 543.2 Geometric and geostatistical designs 563.2.1 Efficiency of spatial sampling designs 563.2.2 Sampling spatial variables in a geostatistical context 573.2.3 Sampling designs minimizing the kriging variance 583.3 Augmented designs: Second-phase sampling 593.3.1 Additional sampling schemes to maximize change in the kriging variance 593.3.2 A weighted kriging variance approach 603.4 A simulated annealing approach 633.5 Illustration 653.5.1 Initial sampling designs 663.5.2 Augmented designs 683.6 Discussion 68References 694 Spatial sampling design by means of spectral approximations to the error process 72Gunter Spöck and Jürgen Pilz4.1 Introduction 724.2 A brief review on spatial sampling design 754.3 The spatial mixed linear model 764.4 Classical Bayesian experimental design problem 774.5 The Smith and Zhu design criterion 794.6 Spatial sampling design for trans-Gaussian kriging 814.7 The spatDesign toolbox 824.7.1 Covariance estimation and variography software 834.7.2 Spatial interpolation and kriging software 844.7.3 Spatial sampling design software 854.8 An example session 894.8.1 Preparatory calculations 894.8.2 Optimal design for the BSLM 934.8.3 Design for the trans-Gaussian kriging 944.9 Conclusions 98References 995 Entropy-based network design using hierarchical Bayesian kriging 103Baisuo Jin, Yuehua Wu and Baiqi Miao5.1 Introduction 1035.2 Entropy-based network design using hierarchical Bayesian kriging 1055.3 The data 1075.4 Spatio-temporal modeling 1075.5 Obtaining a staircase data structure 1115.6 Estimating the hyperparameters H g and the spatial correlations between gauge stations 1135.7 Spatial predictive distribution over the 445 areas located in the 18 districts of Upper Austria 1175.8 Adding gauge stations over the 445 areas located in the 18 districts of Upper Austria 1205.9 Closing down an existing gauge station 1225.10 Model evaluation 124Appendix 5.1: Hierarchical Bayesian spatio-temporal modeling (or kriging) 124Appendix 5.2: Some estimated parameters 128Acknowledgments 129References 1296 Accounting for design in the analysis of spatial data 131Brian J. Reich and Montserrat Fuentes6.1 Introduction 1316.2 Modeling approaches 1346.2.1 Informative missingness 1346.2.2 Informative sampling 1356.2.3 A two-stage approach for informative sampling 1366.3 Analysis of the Austrian precipitation data 1376.4 Discussion 139References 1417 Spatial design for knot selection in knot-based dimension reduction models 142Alan E. Gelfand, Sudipto Banerjee and Andrew O. Finley7.1 Introduction 1427.2 Handling large spatial datasets 1457.3 Dimension reduction approaches 1467.3.1 Basic properties of low rank models 1467.3.2 Predictive process models: A brief review 1487.4 Some basic knot design ideas 1497.4.1 A brief review of spatial design 1497.4.2 A strategy for selecting knots 1517.5 Illustrations 1537.5.1 A simulation example 1537.5.2 A simulation example using the two-step analysis 1597.5.3 Tree height and diameter analysis 1607.5.4 Austria precipitation analysis 1627.6 Discussion and future work 165References 1668 Exploratory designs for assessing spatial dependence 170Agnes Fussl, Werner G. Müller and Juan Rodríguez-Díaz8.1 Introduction 1708.1.1 The dataset and its visualization 1728.2 Spatial links 1748.2.1 Spatial neighbors 1758.2.2 Spatial weights 1768.3 Measures of spatial dependence 1788.4 Models for areal data 1808.4.1 H0 : A spaceless regression model 1818.4.2 H0 : Spatial regression models 1858.5 Design considerations 1908.5.1 A design criterion 1928.5.2 Example 1948.6 Discussion 195Appendix 8.1: R code 198Acknowledgments 202References 2039 Sampling design optimization for space-time kriging 207Gerard B.M. Heuvelink, Daniel A. Griffith, Tomislav Hengl and Stephanie J. Melles9.1 Introduction 2079.2 Methodology 2099.2.1 Space-time universal kriging 2099.2.2 Sampling design optimization with spatial simulated annealing 2119.3 Upper Austria case study 2129.3.1 Descriptive statistics 2129.3.2 Estimation of the space-time model and universal kriging 2159.3.3 Optimal design scenario 1 2189.3.4 Optimal design scenario 2 2199.3.5 Optimal design scenario 3 2199.4 Discussion and conclusions 221Appendix 9.1: R code 222Acknowledgment 227References 22810 Space-time adaptive sampling and data transformations 231José M. Angulo, María C. Bueso and Francisco J. Alonso10.1 Introduction 23110.2 Adaptive sampling network design 23310.2.1 A simulated illustration 23510.3 Predictive information based on data transformations 23810.4 Application to Upper Austria temperature data 24210.5 Summary 246Acknowledgments 247References 24711 Adaptive sampling design for spatio-temporal prediction 249Thomas R. Fanshawe and Peter J. Diggle11.1 Introduction 24911.2 Review of spatial and spatio-temporal adaptive designs 25111.3 The stationary Gaussian model 25311.3.1 Model specification 25311.3.2 Theoretically optimal designs 25411.3.3 A comparison of design strategies 25411.4 The dynamic process convolution model 25711.4.1 Model specification 25711.4.2 A comparison of design strategies 25811.5 Upper Austria rainfall data example 26211.6 Discussion 264Appendix 11.1 266References 26712 Semiparametric dynamic design of monitoring networks for non-Gaussian spatio-temporal data 269Scott H. Holan and Christopher K. Wikle12.1 Introduction 26912.2 Semiparametric non-Gaussian space-time dynamic design 27112.2.1 Semiparametric spatio-temporal dynamic Gamma model 27112.2.2 Simulation-based dynamic design 27412.2.3 Extended Kalman filter for dynamic gamma models 27512.2.4 Extended Kalman filter design algorithm 27712.3 Application: Upper Austria precipitation 27812.4 Discussion 282Acknowledgments 282References 28313 Active learning for monitoring network optimization 285Devis Tuia, Alexei Pozdnoukhov, Loris Foresti and Mikhail Kanevski13.1 Introduction 28513.2 Statistical learning from data 28713.2.1 Algorithmic approaches to learning 28813.2.2 Over-fitting and model selection 28813.3 Support vector machines and kernel methods 28913.3.1 Classification: SVMs 29013.3.2 Density estimation: One-class SVM 29213.3.3 Regression: Kernel ridge regression 29313.3.4 Regression: SVR 29413.4 Active learning 29413.4.1 A general framework 29513.4.2 First steps in active learning: Reducing output variance 29613.4.3 Exploration–exploitation strategies: Towards mixed approaches 29713.5 Active learning with SVMs 29713.5.1 Margin sampling 29713.5.2 Diversity of batches of samples 29913.5.3 Committees of models 29913.6 Case studies 30013.6.1 Austrian climatological data 30013.6.2 Cesium-137 concentration after Chernobyl 30413.6.3 Wind power plants sites evaluation 30713.7 Conclusions 312Acknowledgments 314References 31414 Stationary sampling designs based on plume simulations 319Kristina B. Helle and Edzer Pebesma14.1 Introduction 31914.2 Plumes: From random fields to simulations 32014.3 Cost functions 32414.3.1 Detecting plumes 32414.3.2 Mapping and characterising plumes 32514.3.3 Combined cost functions 32514.4 Optimisation 32614.4.1 Greedy search 32614.4.2 Spatial simulated annealing 32814.4.3 Genetic algorithms 32914.4.4 Other methods 33114.4.5 Evaluation and sensitivity 33114.4.6 Use case: Combination and comparison of optimisation algorithms 33214.5 Results 33414.5.1 Simulations 33414.5.2 Greedy search 33514.5.3 Sensitivity of greedy search to the plume simulations 33614.5.4 Comparison of optimisation algorithms 33714.6 Discussion 340Acknowledgments 341References 341Index 345

Jorge Mateu, Department of Mathematics of the University Jaume I of Castellon, Spain,Werner G. Müller, Department of Applied Statistics, Johannes Kepler University Linz, Austria.

Select Guide Rating

• Contributors xvForeword xix1 Collecting spatio-temporal data 1Jorge Mateu and Werner G. Müller1.1 Introduction 11.2 Paradigms in spatio-temporal design 21.3 Paradigms in spatio-temporal modeling 31.4 Geostatistics and spatio-temporal random functions 41.4.1 Relevant spatio-temporal concepts 41.4.2 Properties of the spatio-temporal covariance and variogram functions 61.4.3 Spatio-temporal kriging 81.4.4 Spatio-temporal covariance models 101.4.5 Parametric estimation of spatio-temporal covariograms 111.5 Types of design criteria and numerical optimization 131.6 The problem set: Upper Austria 171.6.1 Climatic data 171.6.2 Grassland usage 181.7 The chapters 23Acknowledgments 28References 282 Model-based frequentist design for univariate and multivariate geostatistics 37Dale L. Zimmerman and Jie li2.1 Introduction 372.2 Design for univariate geostatistics 382.2.1 Data-model framework 382.2.2 Design criteria 382.2.3 Algorithms 422.2.4 Toy example 422.3 Design for multivariate geostatistics 452.3.1 Data-model framework 452.3.2 Design criteria 472.3.3 Toy example 482.4 Application: Austrian precipitation data network 502.5 Conclusions 52References 533 Model-based criteria heuristics for second-phase spatial sampling 54Eric M. Delmelle3.1 Introduction 543.2 Geometric and geostatistical designs 563.2.1 Efficiency of spatial sampling designs 563.2.2 Sampling spatial variables in a geostatistical context 573.2.3 Sampling designs minimizing the kriging variance 583.3 Augmented designs: Second-phase sampling 593.3.1 Additional sampling schemes to maximize change in the kriging variance 593.3.2 A weighted kriging variance approach 603.4 A simulated annealing approach 633.5 Illustration 653.5.1 Initial sampling designs 663.5.2 Augmented designs 683.6 Discussion 68References 694 Spatial sampling design by means of spectral approximations to the error process 72Gunter Spöck and Jürgen Pilz4.1 Introduction 724.2 A brief review on spatial sampling design 754.3 The spatial mixed linear model 764.4 Classical Bayesian experimental design problem 774.5 The Smith and Zhu design criterion 794.6 Spatial sampling design for trans-Gaussian kriging 814.7 The spatDesign toolbox 824.7.1 Covariance estimation and variography software 834.7.2 Spatial interpolation and kriging software 844.7.3 Spatial sampling design software 854.8 An example session 894.8.1 Preparatory calculations 894.8.2 Optimal design for the BSLM 934.8.3 Design for the trans-Gaussian kriging 944.9 Conclusions 98References 995 Entropy-based network design using hierarchical Bayesian kriging 103Baisuo Jin, Yuehua Wu and Baiqi Miao5.1 Introduction 1035.2 Entropy-based network design using hierarchical Bayesian kriging 1055.3 The data 1075.4 Spatio-temporal modeling 1075.5 Obtaining a staircase data structure 1115.6 Estimating the hyperparameters H g and the spatial correlations between gauge stations 1135.7 Spatial predictive distribution over the 445 areas located in the 18 districts of Upper Austria 1175.8 Adding gauge stations over the 445 areas located in the 18 districts of Upper Austria 1205.9 Closing down an existing gauge station 1225.10 Model evaluation 124Appendix 5.1: Hierarchical Bayesian spatio-temporal modeling (or kriging) 124Appendix 5.2: Some estimated parameters 128Acknowledgments 129References 1296 Accounting for design in the analysis of spatial data 131Brian J. Reich and Montserrat Fuentes6.1 Introduction 1316.2 Modeling approaches 1346.2.1 Informative missingness 1346.2.2 Informative sampling 1356.2.3 A two-stage approach for informative sampling 1366.3 Analysis of the Austrian precipitation data 1376.4 Discussion 139References 1417 Spatial design for knot selection in knot-based dimension reduction models 142Alan E. Gelfand, Sudipto Banerjee and Andrew O. Finley7.1 Introduction 1427.2 Handling large spatial datasets 1457.3 Dimension reduction approaches 1467.3.1 Basic properties of low rank models 1467.3.2 Predictive process models: A brief review 1487.4 Some basic knot design ideas 1497.4.1 A brief review of spatial design 1497.4.2 A strategy for selecting knots 1517.5 Illustrations 1537.5.1 A simulation example 1537.5.2 A simulation example using the two-step analysis 1597.5.3 Tree height and diameter analysis 1607.5.4 Austria precipitation analysis 1627.6 Discussion and future work 165References 1668 Exploratory designs for assessing spatial dependence 170Agnes Fussl, Werner G. Müller and Juan Rodríguez-Díaz8.1 Introduction 1708.1.1 The dataset and its visualization 1728.2 Spatial links 1748.2.1 Spatial neighbors 1758.2.2 Spatial weights 1768.3 Measures of spatial dependence 1788.4 Models for areal data 1808.4.1 H0 : A spaceless regression model 1818.4.2 H0 : Spatial regression models 1858.5 Design considerations 1908.5.1 A design criterion 1928.5.2 Example 1948.6 Discussion 195Appendix 8.1: R code 198Acknowledgments 202References 2039 Sampling design optimization for space-time kriging 207Gerard B.M. Heuvelink, Daniel A. Griffith, Tomislav Hengl and Stephanie J. Melles9.1 Introduction 2079.2 Methodology 2099.2.1 Space-time universal kriging 2099.2.2 Sampling design optimization with spatial simulated annealing 2119.3 Upper Austria case study 2129.3.1 Descriptive statistics 2129.3.2 Estimation of the space-time model and universal kriging 2159.3.3 Optimal design scenario 1 2189.3.4 Optimal design scenario 2 2199.3.5 Optimal design scenario 3 2199.4 Discussion and conclusions 221Appendix 9.1: R code 222Acknowledgment 227References 22810 Space-time adaptive sampling and data transformations 231José M. Angulo, María C. Bueso and Francisco J. Alonso10.1 Introduction 23110.2 Adaptive sampling network design 23310.2.1 A simulated illustration 23510.3 Predictive information based on data transformations 23810.4 Application to Upper Austria temperature data 24210.5 Summary 246Acknowledgments 247References 24711 Adaptive sampling design for spatio-temporal prediction 249Thomas R. Fanshawe and Peter J. Diggle11.1 Introduction 24911.2 Review of spatial and spatio-temporal adaptive designs 25111.3 The stationary Gaussian model 25311.3.1 Model specification 25311.3.2 Theoretically optimal designs 25411.3.3 A comparison of design strategies 25411.4 The dynamic process convolution model 25711.4.1 Model specification 25711.4.2 A comparison of design strategies 25811.5 Upper Austria rainfall data example 26211.6 Discussion 264Appendix 11.1 266References 26712 Semiparametric dynamic design of monitoring networks for non-Gaussian spatio-temporal data 269Scott H. Holan and Christopher K. Wikle12.1 Introduction 26912.2 Semiparametric non-Gaussian space-time dynamic design 27112.2.1 Semiparametric spatio-temporal dynamic Gamma model 27112.2.2 Simulation-based dynamic design 27412.2.3 Extended Kalman filter for dynamic gamma models 27512.2.4 Extended Kalman filter design algorithm 27712.3 Application: Upper Austria precipitation 27812.4 Discussion 282Acknowledgments 282References 28313 Active learning for monitoring network optimization 285Devis Tuia, Alexei Pozdnoukhov, Loris Foresti and Mikhail Kanevski13.1 Introduction 28513.2 Statistical learning from data 28713.2.1 Algorithmic approaches to learning 28813.2.2 Over-fitting and model selection 28813.3 Support vector machines and kernel methods 28913.3.1 Classification: SVMs 29013.3.2 Density estimation: One-class SVM 29213.3.3 Regression: Kernel ridge regression 29313.3.4 Regression: SVR 29413.4 Active learning 29413.4.1 A general framework 29513.4.2 First steps in active learning: Reducing output variance 29613.4.3 Exploration–exploitation strategies: Towards mixed approaches 29713.5 Active learning with SVMs 29713.5.1 Margin sampling 29713.5.2 Diversity of batches of samples 29913.5.3 Committees of models 29913.6 Case studies 30013.6.1 Austrian climatological data 30013.6.2 Cesium-137 concentration after Chernobyl 30413.6.3 Wind power plants sites evaluation 30713.7 Conclusions 312Acknowledgments 314References 31414 Stationary sampling designs based on plume simulations 319Kristina B. Helle and Edzer Pebesma14.1 Introduction 31914.2 Plumes: From random fields to simulations 32014.3 Cost functions 32414.3.1 Detecting plumes 32414.3.2 Mapping and characterising plumes 32514.3.3 Combined cost functions 32514.4 Optimisation 32614.4.1 Greedy search 32614.4.2 Spatial simulated annealing 32814.4.3 Genetic algorithms 32914.4.4 Other methods 33114.4.5 Evaluation and sensitivity 33114.4.6 Use case: Combination and comparison of optimisation algorithms 33214.5 Results 33414.5.1 Simulations 33414.5.2 Greedy search 33514.5.3 Sensitivity of greedy search to the plume simulations 33614.5.4 Comparison of optimisation algorithms 33714.6 Discussion 340Acknowledgments 341References 341Index 345