Machine learning for geospatial data : algorithms, software tools and case studies

Résumé Cette thèse est consacrée à l'analyse, la modélisation et la visualisation de données environnementales à référence spatiale à l'aide d'algorithmes d'apprentissage automatique (Machine Learning). L'apprentissage automatique peut être considéré au sens large comme une sous-catégorie de l'intelligence artificielle qui concerne particulièrement le développement de techniques et d'algorithmes permettant à une machine d'apprendre à partir de données. Dans cette thèse, les algorithmes d'apprentissage automatique sont adaptés pour être appliqués à des données environnementales et à la prédiction spatiale. Pourquoi l'apprentissage automatique ? Parce que la majorité des algorithmes d'apprentissage automatiques sont universels, adaptatifs, non-linéaires, robustes et efficaces pour la modélisation. Ils peuvent résoudre des problèmes de classification, de régression et de modélisation de densité de probabilités dans des espaces à haute dimension, composés de variables informatives spatialisées (« géo-features ») en plus des coordonnées géographiques. De plus, ils sont idéaux pour être implémentés en tant qu'outils d'aide à la décision pour des questions environnementales allant de la reconnaissance de pattern à la modélisation et la prédiction en passant par la cartographie automatique. Leur efficacité est comparable au modèles géostatistiques dans l'espace des coordonnées géographiques, mais ils sont indispensables pour des données à hautes dimensions incluant des géo-features. Les algorithmes d'apprentissage automatique les plus importants et les plus populaires sont présentés théoriquement et implémentés sous forme de logiciels pour les sciences environnementales. Les principaux algorithmes décrits sont le Perceptron multicouches (MultiLayer Perceptron, MLP) - l'algorithme le plus connu dans l'intelligence artificielle, le réseau de neurones de régression généralisée (General Regression Neural Networks, GRNN), le réseau de neurones probabiliste (Probabilistic Neural Networks, PNN), les cartes auto-organisées (SelfOrganized Maps, SOM), les modèles à mixture Gaussiennes (Gaussian Mixture Models, GMM), les réseaux à fonctions de base radiales (Radial Basis Functions Networks, RBF) et les réseaux à mixture de densité (Mixture Density Networks, MDN). Cette gamme d'algorithmes permet de couvrir des tâches variées telle que la classification, la régression ou l'estimation de densité de probabilité. L'analyse exploratoire des données (Exploratory Data Analysis, EDA) est le premier pas de toute analyse de données. Dans cette thèse les concepts d'analyse exploratoire de données spatiales (Exploratory Spatial Data Analysis, ESDA) sont traités selon l'approche traditionnelle de la géostatistique avec la variographie expérimentale et selon les principes de l'apprentissage automatique. La variographie expérimentale, qui étudie les relations entre pairs de points, est un outil de base pour l'analyse géostatistique de corrélations spatiales anisotropiques qui permet de détecter la présence de patterns spatiaux descriptible par une statistique. L'approche de l'apprentissage automatique pour l'ESDA est présentée à travers l'application de la méthode des k plus proches voisins qui est très simple et possède d'excellentes qualités d'interprétation et de visualisation. Une part importante de la thèse traite de sujets d'actualité comme la cartographie automatique de données spatiales. Le réseau de neurones de régression généralisée est proposé pour résoudre cette tâche efficacement. Les performances du GRNN sont démontrées par des données de Comparaison d'Interpolation Spatiale (SIC) de 2004 pour lesquelles le GRNN bat significativement toutes les autres méthodes, particulièrement lors de situations d'urgence. La thèse est composée de quatre chapitres : théorie, applications, outils logiciels et des exemples guidés. Une partie importante du travail consiste en une collection de logiciels : Machine Learning Office. Cette collection de logiciels a été développée durant les 15 dernières années et a été utilisée pour l'enseignement de nombreux cours, dont des workshops internationaux en Chine, France, Italie, Irlande et Suisse ainsi que dans des projets de recherche fondamentaux et appliqués. Les cas d'études considérés couvrent un vaste spectre de problèmes géoenvironnementaux réels à basse et haute dimensionnalité, tels que la pollution de l'air, du sol et de l'eau par des produits radioactifs et des métaux lourds, la classification de types de sols et d'unités hydrogéologiques, la cartographie des incertitudes pour l'aide à la décision et l'estimation de risques naturels (glissements de terrain, avalanches). Des outils complémentaires pour l'analyse exploratoire des données et la visualisation ont également été développés en prenant soin de créer une interface conviviale et facile à l'utilisation. Machine Learning for geospatial data: algorithms, software tools and case studies Abstract The thesis is devoted to the analysis, modeling and visualisation of spatial environmental data using machine learning algorithms. In a broad sense machine learning can be considered as a subfield of artificial intelligence. It mainly concerns with the development of techniques and algorithms that allow computers to learn from data. In this thesis machine learning algorithms are adapted to learn from spatial environmental data and to make spatial predictions. Why machine learning? In few words most of machine learning algorithms are universal, adaptive, nonlinear, robust and efficient modeling tools. They can find solutions for the classification, regression, and probability density modeling problems in high-dimensional geo-feature spaces, composed of geographical space and additional relevant spatially referenced features. They are well-suited to be implemented as predictive engines in decision support systems, for the purposes of environmental data mining including pattern recognition, modeling and predictions as well as automatic data mapping. They have competitive efficiency to the geostatistical models in low dimensional geographical spaces but are indispensable in high-dimensional geo-feature spaces. The most important and popular machine learning algorithms and models interesting for geo- and environmental sciences are presented in details: from theoretical description of the concepts to the software implementation. The main algorithms and models considered are the following: multi-layer perceptron (a workhorse of machine learning), general regression neural networks, probabilistic neural networks, self-organising (Kohonen) maps, Gaussian mixture models, radial basis functions networks, mixture density networks. This set of models covers machine learning tasks such as classification, regression, and density estimation. Exploratory data analysis (EDA) is initial and very important part of data analysis. In this thesis the concepts of exploratory spatial data analysis (ESDA) is considered using both traditional geostatistical approach such as_experimental variography and machine learning. Experimental variography is a basic tool for geostatistical analysis of anisotropic spatial correlations which helps to understand the presence of spatial patterns, at least described by two-point statistics. A machine learning approach for ESDA is presented by applying the k-nearest neighbors (k-NN) method which is simple and has very good interpretation and visualization properties. Important part of the thesis deals with a hot topic of nowadays, namely, an automatic mapping of geospatial data. General regression neural networks (GRNN) is proposed as efficient model to solve this task. Performance of the GRNN model is demonstrated on Spatial Interpolation Comparison (SIC) 2004 data where GRNN model significantly outperformed all other approaches, especially in case of emergency conditions. The thesis consists of four chapters and has the following structure: theory, applications, software tools, and how-to-do-it examples. An important part of the work is a collection of software tools - Machine Learning Office. Machine Learning Office tools were developed during last 15 years and was used both for many teaching courses, including international workshops in China, France, Italy, Ireland, Switzerland and for realizing fundamental and applied research projects. Case studies considered cover wide spectrum of the real-life low and high-dimensional geo- and environmental problems, such as air, soil and water pollution by radionuclides and heavy metals, soil types and hydro-geological units classification, decision-oriented mapping with uncertainties, natural hazards (landslides, avalanches) assessments and susceptibility mapping. Complementary tools useful for the exploratory data analysis and visualisation were developed as well. The software is user friendly and easy to use.

Spatial visualization of DNA in solution.

An image analysis method is presented which allows for the reconstruction of the three-dimensional path of filamentous objects from two of their projections. Starting with stereo pairs, this method is used to trace the trajectory of DNA molecules embedded in vitreous ice and leads to a faithful representation of their three-dimensional shape in solution. This computer-aided reconstruction is superior to the subjective three-dimensional impression generated by observation of stereo pairs of micrographs because it enables one to look at the reconstructed molecules from any chosen direction and distance and allows quantitative analysis such as determination of distances, curvature, persistence length, and writhe of DNA molecules in solution.

Spatial Scale of Clustering of Motor Vehicle Crash Types and Appropriate Countermeasures, 2009

This project analyzes the characteristics and spatial distributions of motor vehicle crash types in order to evaluate the degree and scale of their spatial clustering. Crashes occur as the result of a variety of vehicle, roadway, and human factors and thus vary in their clustering behavior. Clustering can occur at a variety of scales, from the intersection level, to the corridor level, to the area level. Conversely, other crash types are less linked to geographic factors and are more spatially “random.” The degree and scale of clustering have implications for the use of strategies to promote transportation safety. In this project, Iowa's crash database, geographic information systems, and recent advances in spatial statistics methodologies and software tools were used to analyze the degree and spatial scale of clustering for several crash types within the counties of the Iowa Northland Regional Council of Governments. A statistical measure called the K function was used to analyze the clustering behavior of crashes. Several methodological issues, related to the application of this spatial statistical technique in the context of motor vehicle crashes on a road network, were identified and addressed. These methods facilitated the identification of crash clusters at appropriate scales of analysis for each crash type. This clustering information is useful for improving transportation safety through focused countermeasures directly linked to crash causes and the spatial extent of identified problem locations, as well as through the identification of less location-based crash types better suited to non-spatial countermeasures. The results of the K function analysis point to the usefulness of the procedure in identifying the degree and scale at which crashes cluster, or do not cluster, relative to each other. Moreover, for many individual crash types, different patterns and processes and potentially different countermeasures appeared at different scales of analysis. This finding highlights the importance of scale considerations in problem identification and countermeasure formulation.

Revisiting spatial distribution and biochemical composition of calcium-containing crystals in human osteoarthritic articular cartilage.

INTRODUCTION: Calcium-containing (CaC) crystals, including basic calcium phosphate (BCP) and calcium pyrophosphate dihydrate (CPP), are associated with destructive forms of osteoarthritis (OA). We assessed their distribution and biochemical and morphologic features in human knee OA cartilage. METHODS: We prospectively included 20 patients who underwent total knee replacement (TKR) for primary OA. CaC crystal characterization and identification involved Fourier-transform infra-red spectrometry and scanning electron microscopy of 8 to 10 cartilage zones of each knee, including medial and lateral femoral condyles and tibial plateaux and the intercondyle zone. Differential expression of genes involved in the mineralization process between cartilage with and without calcification was assessed in samples from 8 different patients by RT-PCR. Immunohistochemistry and histology studies were performed in 6 different patients. RESULTS: Mean (SEM) age and body mass index of patients at the time of TKR was 74.6 (1.7) years and 28.1 (1.6) kg/m², respectively. Preoperative X-rays showed joint calcifications (chondrocalcinosis) in 4 cases only. The medial femoro-tibial compartment was the most severely affected in all cases, and mean (SEM) Kellgren-Lawrence score was 3.8 (0.1). All 20 OA cartilages showed CaC crystals. The mineral content represented 7.7% (8.1%) of the cartilage weight. All patients showed BCP crystals, which were associated with CPP crystals for 8 joints. CaC crystals were present in all knee joint compartments and in a mean of 4.6 (1.7) of the 8 studied areas. Crystal content was similar between superficial and deep layers and between medial and femoral compartments. BCP samples showed spherical structures, typical of biological apatite, and CPP samples showed rod-shaped or cubic structures. The expression of several genes involved in mineralization, including human homolog of progressive ankylosis, plasma-cell-membrane glycoprotein 1 and tissue-nonspecific alkaline phosphatase, was upregulated in OA chondrocytes isolated from CaC crystal-containing cartilages. CONCLUSIONS: CaC crystal deposition is a widespread phenomenon in human OA articular cartilage involving the entire knee cartilage including macroscopically normal and less weight-bearing zones. Cartilage calcification is associated with altered expression of genes involved in the mineralisation process.

GPS to LRM: Integration of Spatial Point Features with Linear Referencing Methods, 2001

Global positioning systems (GPS) offer a cost-effective and efficient method to input and update transportation data. The spatial location of objects provided by GPS is easily integrated into geographic information systems (GIS). The storage, manipulation, and analysis of spatial data are also relatively simple in a GIS. However, many data storage and reporting methods at transportation agencies rely on linear referencing methods (LRMs); consequently, GPS data must be able to link with linear referencing. Unfortunately, the two systems are fundamentally incompatible in the way data are collected, integrated, and manipulated. In order for the spatial data collected using GPS to be integrated into a linear referencing system or shared among LRMs, a number of issues need to be addressed. This report documents and evaluates several of those issues and offers recommendations. In order to evaluate the issues associated with integrating GPS data with a LRM, a pilot study was created. To perform the pilot study, point features, a linear datum, and a spatial representation of a LRM were created for six test roadway segments that were located within the boundaries of the pilot study conducted by the Iowa Department of Transportation linear referencing system project team. Various issues in integrating point features with a LRM or between LRMs are discussed and recommendations provided. The accuracy of the GPS is discussed, including issues such as point features mapping to the wrong segment. Another topic is the loss of spatial information that occurs when a three-dimensional or two-dimensional spatial point feature is converted to a one-dimensional representation on a LRM. Recommendations such as storing point features as spatial objects if necessary or preserving information such as coordinates and elevation are suggested. The lack of spatial accuracy characteristic of most cartography, on which LRM are often based, is another topic discussed. The associated issues include linear and horizontal offset error. The final topic discussed is some of the issues in transferring point feature data between LRMs.

Spatial mapping of exciton lifetimes in single ZnO nanowires

We investigate the spatial dependence of the exciton lifetimes in single ZnO nanowires. We have found that the free exciton and bound exciton lifetimes exhibit a maximum at the center of nanowires, while they decrease by 30% towards the tips. This dependence is explained by considering the cavity-like properties of the nanowires in combination with the Purcell effect. We show that the lifetime of the bound-excitons scales with the localization energy to the power of 3/2, which validates the model of Rashba and Gurgenishvili at the nanoscale.

Dissociation between components of spatial memory in rats after recovery from the effects of retrohippocampal lesions

A series of 4 experiments examined the performance of rats with retrohippocampal lesions on a spatial water-maze task. The animals were trained to find and escape onto a hidden platform after swimming in a large pool of opaque water. The platform was invisible and could not be located using olfactory cues. Successful escape performance required the rats to develop strategies of approaching the correct location with reference solely to distal extramaze cues. The lesions encompassed the entire rostro-caudal extent of the lateral and medial entorhinal cortex, and included parts of the pre- and para-subiculum, angular bundle and subiculum. Groups ECR 1 and 2 sustained only partial damage of the subiculum, while Group ECR+S sustained extensive damage. These groups were compared with sham-lesion and unoperated control groups. In Expt 1A, a profound deficit in spatial localisation was found in groups ECR 1 and ECR+S, the rats receiving all training postoperatively. In Expt 1B, these two groups showed hyperactivity in an open-field. In Expt 2, extensive preoperative training caused a transitory saving in performance of the spatial task by group ECR 2, but comparisons with the groups of Expt 1A revealed no sustained improvement, except on one measure of performance in a post-training transfer test. All rats were then given (Expt 3) training on a cueing procedure using a visible platform. The spatial deficit disappeared but, on returning to the normal hidden platform procedure, it reappeared. Nevertheless, a final transfer test, during which the platform was removed from the apparatus, revealed a dissociation between two independent measures of performance: the rats with ECR lesions failed to search for the hidden platform but repeatedly crossed its correct location accurately during traverses of the entire pool. This partial recovery of performance was not (Expt 4) associated with any ability to discriminate between two locations in the pool. The apparently selective recovery of aspects of spatial memory is discussed in relation to O'Keefe and Nadel's (1978) spatial mapping theory of hippocampal function. We propose a modification of the theory in terms of a dissociation between procedural and declarative subcomponents of spatial memory. The declarative component is a flexible access system in which information is stored in a form independent of action. It is permanently lost after the lesion. The procedural component is "unmasked" by the retrohippocampal lesion giving rise to the partial recovery of spatial localisation performance.

Sensitivity of Spring Phenology to Warming Across Temporal and Spatial Climate Gradients in Two Independent Databases

Disparate ecological datasets are often organized into databases post hoc and then analyzed and interpreted in ways that may diverge from the purposes of the original data collections. Few studies, however, have attempted to quantify how biases inherent in these data (for example, species richness, replication, climate) affect their suitability for addressing broad scientific questions, especially in under-represented systems (for example, deserts, tropical forests) and wild communities. Here, we quantitatively compare the sensitivity of species first flowering and leafing dates to spring warmth in two phenological databases from the Northern Hemisphere. One-PEP725-has high replication within and across sites, but has low species diversity and spans a limited climate gradient. The other-NECTAR-includes many more species and a wider range of climates, but has fewer sites and low replication of species across sites. PEP725, despite low species diversity and relatively low seasonality, accurately captures the magnitude and seasonality of warming responses at climatically similar NECTAR sites, with most species showing earlier phenological events in response to warming. In NECTAR, the prevalence of temperature responders significantly declines with increasing mean annual temperature, a pattern that cannot be detected across the limited climate gradient spanned by the PEP725 flowering and leafing data. Our results showcase broad areas of agreement between the two databases, despite significant differences in species richness and geographic coverage, while also noting areas where including data across broader climate gradients may provide added value. Such comparisons help to identify gaps in our observations and knowledge base that can be addressed by ongoing monitoring and research efforts. Resolving these issues will be critical for improving predictions in understudied and under-sampled systems outside of the temperature seasonal mid-latitudes.

Vision, spatial cognition and intellectual disability.

Vision is the most synthetic sensory channel and it provides specific information about the relative position of distant landmarks during visual exploration. In this paper we propose that visual exploration, as assessed by the recording of eye movements, offers an original method to analyze spatial cognition and to reveal alternative adaptation strategies in people with intellectual disabilities (ID). Our general assumption is that eye movement exploration may simultaneously reveal whether, why, and how, compensatory strategies point to specific difficulties related to neurological symptoms. An understanding of these strategies will also help in the development of optimal rehabilitation procedures.

Spatial polarization of the ecological footprint distribution

The international allocation of natural resources is determined, not by any ethical or ecological criteria, but by the dominance of market mechanisms. From a core-periphery perspective, this allocation may even be driven by historically determined structural patterns, with a core group of countries whose consumption appropriates most available natural resources, and another group, having low natural resource consumption, which plays a peripheral role. This article consists of an empirical distributional analysis of natural resource consumption (as measured by Ecological Footprints) whose purpose is to assess the extent to which the distribution of consumption responds to polarization (as opposed to mere inequality). To assess this, we estimate and decompose different polarization indices for a balanced sample of 119 countries over the period 1961 to 2007. Our results points toward a polarized distribution which is consistent with a core-periphery framework. Keywords: Polarization, Core-Periphery, Ecological Footprint

What matters for predicting spatial distributions of trees: techniques, data, or species' characteristics?

Data characteristics and species traits are expected to influence the accuracy with which species' distributions can be modeled and predicted. We compare 10 modeling techniques in terms of predictive power and sensitivity to location error, change in map resolution, and sample size, and assess whether some species traits can explain variation in model performance. We focused on 30 native tree species in Switzerland and used presence-only data to model current distribution, which we evaluated against independent presence-absence data. While there are important differences between the predictive performance of modeling methods, the variance in model performance is greater among species than among techniques. Within the range of data perturbations in this study, some extrinsic parameters of data affect model performance more than others: location error and sample size reduced performance of many techniques, whereas grain had little effect on most techniques. No technique can rescue species that are difficult to predict. The predictive power of species-distribution models can partly be predicted from a series of species characteristics and traits based on growth rate, elevational distribution range, and maximum elevation. Slow-growing species or species with narrow and specialized niches tend to be better modeled. The Swiss presence-only tree data produce models that are reliable enough to be useful in planning and management applications.

The Spatial Effects of Trade Openness: A Survey

This paper surveys the literature on the implications of trade liberalisation for intra-national economic geographies. Three results stand out. First, neither urban systems models nor new economic geography models imply a robust prediction for the impact of trade openness on spatial concentration. Whether trade promotes concentration or dispersion depends on subtle modelling choices among which it is impossible to adjudicate a priori. Second, empirical evidence mirrors the theoretical indeterminacy: a majority of cross-country studies find no significant effect of openness on urban concentration or regional inequality. Third, the available models predict that, other things equal, regions with inherently less costly access to foreign markets, such as border or port regions, stand to reap the largest gains from trade liberalisation. This prediction is confirmed by the available evidence. Whether trade liberalisation raises or lowers regional inequality therefore depends on each country's specific geography.

Metabolic activation pattern of distinct hippocampal subregions during spatial learning and memory retrieval.

Activation dynamics of hippocampal subregions during spatial learning and their interplay with neocortical regions is an important dimension in the understanding of hippocampal function. Using the (14C)-2-deoxyglucose autoradiographic method, we have characterized the metabolic changes occurring in hippocampal subregions in mice while learning an eight-arm radial maze task. Autoradiogram densitometry revealed a heterogeneous and evolving pattern of enhanced metabolic activity throughout the hippocampus during the training period and on recall. In the early stages of training, activity was enhanced in the CA1 area from the intermediate portion to the posterior end as well as in the CA3 area within the intermediate portion of the hippocampus. At later stages, CA1 and CA3 activations spread over the entire longitudinal axis, while dentate gyrus (DG) activation occurred from the anterior to the intermediate zone. Activation of the retrosplenial cortex but not the amygdala was also observed during the learning process. On recall, only DG activation was observed in the same anterior part of the hippocampus. These results suggest the existence of a functional segmentation of the hippocampus, each subregion being dynamically but also differentially recruited along the acquisition, consolidation, and retrieval process in parallel with some neocortical sites.

Ibotenate Lesions of Hippocampus and/or Subiculum: Dissociating Components of Allocentric Spatial Learning

This study examined the effects of ibotenic acid-induced lesions of the hippocampus, subiculum and hippocampus +/- subiculum upon the capacity of rats to learn and perform a series of allocentric spatial learning tasks in an open-field water maze. The lesions were made by infusing small volumes of the neurotoxin at a total of 26 (hippocampus) or 20 (subiculum) sites intended to achieve complete target cell loss but minimal extratarget damage. The regional extent and axon-sparing nature of these lesions was evaluated using both cresyl violet and Fink - Heimer stained sections. The behavioural findings indicated that both the hippocampus and subiculum lesions caused impairment to the initial postoperative acquisition of place navigation but did not prevent eventual learning to levels of performance almost as effective as those of controls. However, overtraining of the hippocampus + subiculum lesioned rats did not result in significant place learning. Qualitative observations of the paths taken to find a hidden escape platform indicated that different strategies were deployed by hippocampal and subiculum lesioned groups. Subsequent training on a delayed matching to place task revealed a deficit in all lesioned groups across a range of sample choice intervals, but the subiculum lesioned group was less impaired than the group with the hippocampal lesion. Finally, unoperated control rats given both the initial training and overtraining were later given either a hippocampal lesion or sham surgery. The hippocampal lesioned rats were impaired during a subsequent retention/relearning phase. Together, these findings suggest that total hippocampal cell loss may cause a dual deficit: a slower rate of place learning and a separate navigational impairment. The prospect of unravelling dissociable components of allocentric spatial learning is discussed.