Fernando Bacao

Self-organizing maps as substitutes for k-means clustering

One of the most widely used clustering techniques used in GISc problems is the k-means algorithm. One of the most important issues in the correct use of k-means is the initialization procedure that ultimately determines which part of the solution space will be searched. In this paper we briefly review different initialization procedures, and propose Kohonen’s Self-Organizing Maps as the most convenient method, given the proper training parameters. Furthermore, we show that in the final stages of its training procedure the Self-Organizing Map algorithms is rigorously the same as the k-means algorithm. Thus we propose the use of Self-Organizing Maps as possible substitutes for the more classical k-means clustering algorithms.

Modeling and mapping wildfire ignition risk in Portugal

Portugal has the highest density of wildfire ignitions among southern European countries. The ability to predict the spatial patterns of ignitions constitutes an important tool for managers, helping to improve the effectiveness of fire prevention, detection and firefighting resources allocation. In this study, we analyzed 127 490 ignitions that occurred in Portugal during a 5-year period. We used logistic regression models to predict the likelihood of ignition occurrence, using a set of potentially explanatory variables, and produced an ignition risk map for the Portuguese mainland. Results show that population density, human accessibility, land cover and elevation are important determinants of spatial distribution of fire ignitions. In this paper, we demonstrate that it is possible to predict the spatial patterns of ignitions at the national level with good accuracy and using a small number of easily obtainable variables, which can be useful in decision-making for wildfire management.

The self-organizing map, the Geo-SOM, and relevant variants for geosciences

In this paper we explore the advantages of using Self-Organized Maps (SOMs) when dealing with geo-referenced data. The standard SOM algorithm is presented, together with variants which are relevant in the context of the analysis of geo-referenced data. We present a new SOM architecture, the Geo-SOM, which was especially designed to take into account spatial dependency. The strengths and weaknesses of the different variants proposed are shown through a set of tests based on artificial data. A real world application of these techniques is given through the analysis of geodemographic data from Lisbons metropolitan area.

Digital divide across the European Union

Our research analyses the digital divide within the European Union 27 between the years of 2008 and 2010. To accomplish this we use multivariate statistical methods, more specifically factor and cluster analysis, to address the European digital disparities. Our results lead to an identification of two latent dimensions and five groups of countries. We conclude that a digital gap does, in fact, exist within the European Union. The process of European integration and the economic wealth emerge as explanatory factors for this divide. On the other hand, the educational attendance is not proven to be significant, as one would expect.

Applying genetic algorithms to zone design

Genetic algorithms (GA) have been found to provide global near optimal solutions in a wide range of complex problems. In this paper genetic algorithms have been used to deal with the complex problem of zone design. The zone design problem comprises a large number of geographical tasks, from which electoral districting is probably the most well known. The electoral districting problem is described and formalized mathematically. Different problem encodings, suited to GA optimization, are presented, together with different objective functions. A practical real world example is given and tests performed in order to evaluate the effectiveness of the GA approach.

Geo-Self-Organizing Map (Geo-SOM) for Building and Exploring Homogeneous Regions

Regionalization and uniform/homogeneous region building constitutes one of the most longstanding concerns of geographers. In this paper we explore the Geo-Self-Organizing Map (Geo-SOM) as a tool to develop homogeneous regions and perform geographic pattern detection. The Geo-SOM presents several advantages over other available methods. The possibility of “what-if” analysis, coupled with powerful visualization tools and the accommodation of spatial constraints, constitute some of the most relevant features of the Geo-SOM. In this paper we show the opportunities made available by this tool and explore different features which allow rich exploratory spatial data analysis.

Carto-SOM: cartogram creation using self-organizing maps

The basic idea of a cartogram is to distort a geographical map by substituting the geographic area of a region by some other variable of interest. The objective is to rescale each region according to the value of the variable of interest while keeping the map, as much as possible, recognizable. There are several algorithms for building cartograms. None of these methods has proved to be universally better than any other, since the trade‐offs made to get the correct distortion vary. In this paper we present a new method for building cartograms, based on self‐organizing neural networks (Kohonens self‐organizing maps or SOM). The proposed method is widely available and is easy to carry out, and yet has several appealing properties, such as easy parallelization, making up a good tool for geographic data presentation and analysis. We present a series of tests on different problems, comparing the new algorithm with existing ones. We conclude that it is competitive and, in some circumstances, can perform better then existing algorithms.

Characterizing and modelling the spatial patterns of wildfire ignitions in Portugal: fire initiation and resulting burned area.

According to the statistics Portugal has the highest density of wildfire ignitions among southern European countries. The ability to predict ignition occurrence constitutes an important tool for managers, helping to improve the effectiveness of fire prevention, detection and fire fighting resources allocation. In this study we used a database with information about 127 490 fire ignitions that occurred in Portugal during a five year period. We performed frequency analysis to characterize the occurrence of wildfire ignitions in relation to both human and environmental variables and compared the spatial patterns of ignitions which originated fires larger or smaller than 500 ha. We also used logistic regression models to predict the relative probability of ignition occurrence, as a function of the resulting fire size. Results show that fire ignitions are strongly related to human presence and activity, and that the spatial patterns of ignitions are different for larger or smaller wildfires. Larger wildfires started in areas with lower population density, more distant from the main roads and at higher elevations, when compared to smaller fires, and also started more frequently in shrublands and forested areas. The results obtained can be useful in decision making for fire danger management.

Combining per-pixel and object-based classifications for mapping land cover over large areas

A plethora of national and regional applications need land-cover information covering large areas. Manual classification based on visual interpretation and digital per-pixel classification are the two most commonly applied methods for land-cover mapping over large areas using remote-sensing images, but both present several drawbacks. This paper tests a method with moderate spatial resolution images for deriving a product with a predefined minimum mapping unit (MMU) unconstrained by spatial resolution. The approach consists of a traditional supervised per-pixel classification followed by a post-classification processing that includes image segmentation and semantic map generalization. The approach was tested with AWiFS data collected over a region in Portugal to map 15 land-cover classes with 10 ha MMU. The map presents a thematic accuracy of 72.6 ± 3.7% at the 95% confidence level, which is approximately 10% higher than the per-pixel classification accuracy. The results show that segmentation of moderate-spatial resolution images and semantic map generalization can be used in an operational context to automatically produce land-cover maps with a predefined MMU over large areas.

Exploratory geospatial data analysis using the GeoSOM suite

Abstract Clustering constitutes one of the most popular and important tasks in data analysis. This is true for any type of data, and geographic data is no exception. In fact, in geographic knowledge discovery the aim is, more often than not, to explore and let spatial patterns surface rather than develop predictive models. The size and dimensionality of the existing and future databases stress the need for efficient and robust clustering algorithms. This need has been successfully addressed in the context of general-purpose knowledge discovery. Geographic knowledge discovery, nonetheless can still benefit from better tools, especially if these tools are able to integrate geographic information and aspatial variables in order to assist the geographic analyst’s objectives and needs. Typically, the objectives are related with finding spatial patterns based on the interaction between location and aspatial variables. When performing cluster-based analysis of geographic data, user interaction is essential to understand and explore the emerging patterns, and the lack of appropriate tools for this task hinders a lot of otherwise very good work. In this paper, we present the GeoSOM suite as a tool designed to bridge the gap between clustering and the typical geographic information science objectives and needs. The GeoSOM suite implements the GeoSOM algorithm, which changes the traditional Self-Organizing Map algorithm to explicitly take into account geographic information. We present a case study, based on census data from Lisbon, exploring the GeoSOM suite features and exemplifying its use in the context of exploratory data analysis.