Advanced Mapping of Environmental Data

Mikhail Kanevski, Institute of Geomatics and Analysis of Risk, University of Lausanne, Switzerland.

• Preface xiChapter 1. Advanced Mapping of Environmental Data: Introduction 1M. KANEVSKI1.1. Introduction 11.2. Environmental data analysis: problems and methodology 31.2.1. Spatial data analysis: typical problems 31.2.2. Spatial data analysis: methodology 51.2.3. Model assessment and model selection 81.3. Resources 121.3.1. Books, tutorials 121.3.2. Software 121.4. Conclusion 141.5. References 15Chapter 2. Environmental Monitoring Network Characterization and Clustering 19D. TUIA and M. KANEVSKI2.1. Introduction 192.2. Spatial clustering and its consequences 202.2.1. Global parameters 212.2.2. Spatial predictions 222.3. Monitoring network quantification 232.3.1. Topological quantification 232.3.2. Global measures of clustering 232.3.2.1. Topological indices 232.3.2.2. Statistical indices 242.3.3. Dimensional resolution: fractal measures of clustering 262.3.3.1. Sandbox method 272.3.3.2. Box-counting method 302.3.3.3. Lacunarity 332.4. Validity domains 342.5. Indoor radon in Switzerland: an example of a real monitoring network 362.5.1. Validity domains 372.5.2. Topological index 372.5.3. Statistical indices 382.5.3.1. Morisita index 382.5.3.2. K-function 392.5.4. Fractal dimension 402.5.4.1. Sandbox and box-counting fractal dimension 402.5.4.2. Lacunarity 422.6. Conclusion 432.7. References 44Chapter 3. Geostatistics: Spatial Predictions and Simulations 47E. SAVELIEVA, V. DEMYANOV and M. MAIGNAN3.1. Assumptions of geostatistics 473.2. Family of kriging models 493.2.1. Simple kriging 503.2.2. Ordinary kriging 503.2.3. Basic features of kriging estimation 513.2.4. Universal kriging (kriging with trend) 563.2.5. Lognormal kriging 563.3. Family of co-kriging models 583.3.1. Kriging with linear regression 583.3.2. Kriging with external drift 583.3.3. Co-kriging 593.3.4. Collocated co-kriging 603.3.5. Co-kriging application example 613.4. Probability mapping with indicator kriging 643.4.1. Indicator coding 643.4.2. Indicator kriging 663.4.3. Indicator kriging applications 693.4.3.1. Indicator kriging for 241Am analysis 693.4.3.2. Indicator kriging for aquifer layer zonation 713.4.3.3. Indicator kriging for localization of crab crowds 743.5. Description of spatial uncertainty with conditional stochastic simulations 763.5.1. Simulation vs. estimation 763.5.2. Stochastic simulation algorithms 773.5.3. Sequential Gaussian simulation 813.5.4. Sequential indicator simulations 843.5.5. Co-simulations of correlated variables 883.6. References 92Chapter 4. Spatial Data Analysis and Mapping Using Machine Learning Algorithms 95F. RATLE, A. POZDNOUKHOV, V. DEMYANOV, V. TIMONIN and E. SAVELIEVA4.1. Introduction 954.2. Machine learning: an overview 964.2.1. The three learning problems 964.2.2. Approaches to learning from data 1004.2.3. Feature selection 1014.2.4. Model selection 1034.2.5. Dealing with uncertainties 1074.3. Nearest neighbor methods 1084.4. Artificial neural network algorithms 1094.4.1. Multi-layer perceptron neural network 1094.4.2. General Regression Neural Networks 1194.4.3. Probabilistic Neural Networks 1224.4.4. Self-organizing (Kohonen) maps 1244.5. Statistical learning theory for spatial data: concepts and examples 1314.5.1. VC dimension and structural risk minimization 1314.5.2. Kernels 1324.5.3. Support vector machines 1334.5.4. Support vector regression 1374.5.5. Unsupervised techniques 1414.5.5.1. Clustering 1424.5.5.2. Nonlinear dimensionality reduction 1444.6. Conclusion 1464.7. References 146Chapter 5. Advanced Mapping of Environmental Spatial Data: Case Studies 149L. FORESTI, A. POZDNOUKHOV, M. KANEVSKI, V. TIMONIN, E. SAVELIEVA, C. KAISER, R. TAPIA and R. PURVES5.1. Introduction 1495.2. Air temperature modeling with machine learning algorithms and geostatistics 1505.2.1. Mean monthly temperature 1515.2.1.1. Data description 1515.2.1.2. Variography 1525.2.1.3. Step-by-step modeling using a neural network 1535.2.1.4. Overfitting and undertraining 1545.2.1.5. Mean monthly air temperature prediction mapping 1565.2.2. Instant temperatures with regionalized linear dependencies 1595.2.2.1. The Föhn phenomenon 1595.2.2.2. Modeling of instant air temperature influenced by Föhn 1605.2.3. Instant temperatures with nonlinear dependencies 1635.2.3.1. Temperature inversion phenomenon 1635.2.3.2. Terrain feature extraction using Support Vector Machines 1645.2.3.3. Temperature inversion modeling with MLP 1655.3. Modeling of precipitation with machine learning and geostatistics 1685.3.1. Mean monthly precipitation 1695.3.1.1. Data description 1695.3.1.2. Precipitation modeling with MLP 1715.3.2. Modeling daily precipitation with MLP 1735.3.2.1. Data description 1735.3.2.2. Practical issues of MLP modeling 1745.3.2.3. The use of elevation and analysis of the results 1775.3.3. Hybrid models: NNRK and NNRS 1795.3.3.1. Neural network residual kriging 1795.3.3.2. Neural network residual simulations 1825.3.4. Conclusions 1845.4. Automatic mapping and classification of spatial data using machine learning 1855.4.1. k-nearest neighbor algorithm 1855.4.1.1. Number of neighbors with cross-validation 1875.4.2. Automatic mapping of spatial data 1875.4.2.1. KNN modeling 1885.4.2.2. GRNN modeling 1905.4.3. Automatic classification of spatial data 1925.4.3.1. KNN classification 1935.4.3.2. PNN classification 1945.4.3.3. Indicator kriging classification 1975.4.4. Automatic mapping – conclusions 1995.5. Self-organizing maps for spatial data – case studies 2005.5.1. SOM analysis of sediment contamination 2005.5.2. Mapping of socio-economic data with SOM 2045.6. Indicator kriging and sequential Gaussian simulations for probability mapping. Indoor radon case study 2095.6.1. Indoor radon measurements 2095.6.2. Probability mapping 2115.6.3. Exploratory data analysis 2125.6.4. Radon data variography 2165.6.4.1. Variogram for indicators 2165.6.4.2. Variogram for Nscores 2175.6.5. Neighborhood parameters 2185.6.6. Prediction and probability maps 2195.6.6.1. Probability maps with IK 2195.6.6.2. Probability maps with SGS 2205.6.7. Analysis and validation of results 2215.6.7.1. Influence of the simulation net and the number of neighbors 2215.6.7.2. Decision maps and validation of results 2225.6.8. Conclusions 2255.7. Natural hazards forecasting with support vector machines – case study: snow avalanches 2255.7.1. Decision support systems for natural hazards 2275.7.2. Reminder on support vector machines 2285.7.2.1. Probabilistic interpretation of SVM 2295.7.3. Implementing an SVM for avalanche forecasting 2305.7.4. Temporal forecasts 2305.7.4.1. Feature selection 2315.7.4.2. Training the SVM classifier 2325.7.4.3. Adapting SVM forecasts for decision support 2335.7.5. Extending the SVM to spatial avalanche predictions 2375.7.5.1. Data preparation 2375.7.5.2. Spatial avalanche forecasting 2395.7.6. Conclusions 2415.8. Conclusion 2415.9. References 242Chapter 6. Bayesian Maximum Entropy – BME 247G. CHRISTAKOS6.1. Conceptual framework 2476.2. Technical review of BME 2516.2.1. The spatiotemporal continuum 2516.2.2. Separable metric structures 2536.2.3. Composite metric structures 2556.2.4. Fractal metric structures 2566.3. Spatiotemporal random field theory 2576.3.1. Pragmatic S/TRF tools 2586.3.2. Space-time lag dependence: ordinary S/TRF 2606.3.3. Fractal S/TRF 2626.3.4. Space-time heterogenous dependence: generalized S/TRF 2646.4. About BME 2676.4.1. The fundamental equations 2676.4.2. A methodological outline 2736.4.3. Implementation of BME: the SEKS-GUI 2756.5. A brief review of applications 2816.5.1. Earth and atmospheric sciences 2826.5.2. Health, human exposure and epidemiology 2916.6. References 299List of Authors 307Index 309

"It gives a good overview, is clearly written, is concise, and includes many references to papers published in the different areas." (Zentralblatt MATH, 2011)