Jaroslaw Jasiewicz

LandEx—A GeoWeb Tool for Query and Retrieval of Spatial Patterns in Land Cover Datasets

The vast amount of data collected by satellites via remote sensing is a valuable resource, however, it lacks machine search capabilities. In particular, large land cover datasets, such as the 30-m/cell NLCD 2006 covering the entire conterminous United States, are rarely analyzed as a whole due to the lack of tools beyond the basic statistics and SQL queries. Consequently, the NLCD is underutilized relative to its potential. We address this issue by introducing LandEx-a GeoWeb application for real time, content-based exploration and mining of land cover patterns in large datasets. By combining the functionality of online computerized maps with the power of the pattern recognition algorithm, LandEx provides an easy to use visual search engine for the entire extent of the NLCD at its full resolution. The user selects a pattern of interest (a query) and the tool produces a similarity map indicating the spatial distribution of locations having patterns of land cover similar to that in the query. Pattern-based query and retrieval addresses the issue of structural similarity between landscapes. The core of the method is the similarity function between two patterns which is based on 2D land cover class/clump size histograms and the Jensen-Shannon divergence. The search relies on exhaustive evaluation using an overlapping sliding window approach. LandEx is implemented using Free Open Source Software (FOSS) software and adheres to the Open Geospatial Consortium (OGC) standards. The wait time for an answer to a query is only several seconds due to the high level of system optimization. The methodology and implementation of LandEx are described in detail and illustrative examples of its application to different domains, including agriculture, forestry, and urbanization are given.

Example-Based Retrieval of Alike Land-Cover Scenes From NLCD2006 Database

Query-by-image-content (QBIC) tools are in demand in geospatial community because they enable exploration and mining of the rapidly increasing database of remotely sensed images. Accompanying the growth of the imagery database is the increase in the number of image-derived products, such as high-resolution large-spatial-extent maps of land cover/land use (LCLU). QBIC-like tools for exploration and mining of such products would significantly enhance their value. In this letter, we present a method for retrieval of alike scenes from a category-valued geospatial database of which an LCLU map is a particular example. Alikeness between the two scenes is tantamount to similarity between their spatial patterns of class labels. Our method works on the principle of query by example, its input is a reference scene, and its output is a similarity map indicating a degree of alikeness between a location on the map and the reference. The two core components of the method are as follows: scene signature—an encapsulation of the scene pattern by means of probability distribution of class labels and the sizes of the patches that they form, and scene similarity–a mutual-information-based function that assigns a level of similarity between any two scenes based on their signatures. The method is described in detail and applied to the National Land Cover Dataset 2006. Two examples of queries on this data set are presented and discussed. The applicability of the method to other data sets is discussed.

Spatial association between dissection density and environmental factors over the entire conterminous United States

Previous studies of land dissection density (D) often find contradictory results regarding factors controlling its spatial variation. We hypothesize that the dominant controlling factors (and the interactions between them) vary from region to region due to differences in each regions local characteristics and geologic history. We test this hypothesis by applying a geographical detector method to eight physiographic divisions of the conterminous United States and identify the dominant factor(s) in each. The geographical detector method computes the power of determinant (q) that quantitatively measures the affinity between the factor considered and D. Results show that the factor (or factor combination) with the largest q value is different for physiographic regions with different characteristics and geologic histories. For example, lithology dominates in mountainous regions, curvature dominates in plains, and glaciation dominates in previously glaciated areas. The geographical detector method offers an objective framework for revealing factors controlling Earth surface processes.

GeoPAT: A toolbox for pattern-based information retrieval from large geospatial databases

Abstract Geospatial Pattern Analysis Toolbox (GeoPAT) is a collection of GRASS GIS modules for carrying out pattern-based geospatial analysis of images and other spatial datasets. The need for pattern-based analysis arises when images/rasters contain rich spatial information either because of their very high resolution or their very large spatial extent. Elementary units of pattern-based analysis are scenes – patches of surface consisting of a complex arrangement of individual pixels (patterns). GeoPAT modules implement popular GIS algorithms, such as query, overlay, and segmentation, to operate on the grid of scenes. To achieve these capabilities GeoPAT includes a library of scene signatures – compact numerical descriptors of patterns, and a library of distance functions – providing numerical means of assessing dissimilarity between scenes. Ancillary GeoPAT modules use these functions to construct a grid of scenes or to assign signatures to individual scenes having regular or irregular geometries. Thus GeoPAT combines knowledge retrieval from patterns with mapping tasks within a single integrated GIS environment. GeoPAT is designed to identify and analyze complex, highly generalized classes in spatial datasets. Examples include distinguishing between different styles of urban settlements using VHR images, delineating different landscape types in land cover maps, and mapping physiographic units from DEM. The concept of pattern-based spatial analysis is explained and the roles of all modules and functions are described. A case study example pertaining to delineation of landscape types in a subregion of NLCD is given. Performance evaluation is included to highlight GeoPATs applicability to very large datasets. The GeoPAT toolbox is available for download from http://sil.uc.edu/ .

Unsupervised regionalization of the United States into landscape pattern types

ABSTRACT We present a pattern-based regionalization of the conterminous US – a partitioning of the country into a number of mutually exclusive and exhaustive regions that maximizes the intra-region stationarity of land cover patterns and inter-region disparity between those patterns. The result is a discretization of the land surface into a number of landscape pattern types (LPTs) – spatial units each containing a unique quasi-stationary pattern of land cover classes. To achieve this goal, we use a recently developed method which utilizes machine vision techniques. First, the entire National Land Cover Dataset (NLCD) is partitioned into a grid of square-size blocks of cells, called motifels. The size of a motifel defines the spatial scale of a local landscape. The land cover classes of cells within a motifel form a local landscape pattern which is mathematically represented by a histogram of co-occurrence features. Using the Jensen–Shannon divergence as a dissimilarity function between patterns we group the motifels into several LPTs. The grouping procedure consists of two phases. First, the grid of motifels is partitioned spatially using a region-growing segmentation algorithm. Then, the resulting segments of this grid, each represented by its medoid, are clustered using a hierarchical algorithm with Ward’s linkage. The broad-extent maps of progressively more generalized LPTs resulting from this procedure are shown and discussed. Our delineated LPTs agree well with the perceptual patterns seen in the NLCD map.

Multi-scale segmentation algorithm for pattern-based partitioning of large categorical rasters

Abstract Analyzing large Earth Observation (EO) data on the broad spatial scales frequently involves regionalization of patterns. To automate this process we present a segmentation algorithm designed specifically to delineate segments containing quasi-stationary patterns. The algorithm is designed to work with patterns of a categorical variable. This makes it possible to analyze very large spatial datasets (for example, a global land cover) in their entirety. An input categorical raster is first tessellated into small square tiles to form a new, coarser, grid of tiles. A mosaic of categories within each tile forms a local pattern, and the segmentation algorithm partitions the grid of tiles while maintaining the cohesion of pattern in each segment. The algorithm is based on the principle of seeded region growing (SRG) but it also includes segment merging and other enhancements to segmentation quality. Our key contribution is an extension of the concept of segmentation to grids in which each cell has a non-negligible size and contains a complex data structure (histograms of pattern features). Specific modification of a standard SRG algorithm include: working in a distance space with complex data objects, introducing six-connected “brick wall” topology of the grid to decrease artifacts associated with tessellation of geographical space, constructing the SRG priority queue of seeds on the basis of local homogeneity of patterns, and using a content-dependent value of segment-growing threshold. The detailed description of the algorithm is given followed by an assessment of its performance on test datasets representing three pertinent themes of land cover, topography, and a high-resolution image. Pattern-based segmentation algorithm will find application in ecology, forestry, geomorphology, land management, and agriculture. The algorithm is implemented as a module of GeoPAT – an already existing, open source toolbox for performing pattern-based analysis of categorical rasters.

Geo-spatial modelling with unbalanced data:modelling the spatial pattern of human activityduring the Stone Age

Abstract With the increasing availability of data, geoscience provides many methods to model the spatial extent of various phenomena.Acquiring representative, high quality data is the most important criterion to assess the value of any spatial analysis, however, there are many situations in which these criteria cannot be fulfilled. Archived data, collected in the past, for which analysis cannot be repeated or supplemented is a very common information source. Archaeological data collected at a regional extent during years of field work and superficial observations are an additional example. Such data rarely provide representative samples and are usually imbalanced; only very few examples contain useful data, while many examples remain without any archaeological traces. In spite of these limitations archaeological information presented in the form of maps can be a useful and helpful tool to analyse the spatial patterns of some phenomena and, from a more practical point of view, a tool to predict the location of undiscovered occurrences. The primary goal of this paper is to present a methodology for modelling spatial patterns based on imbalanced categorical data which do not fulfil the criteria of spatial representation and incorporates uncertainty in its decision process. This concept will be discussed using a collection of Stone Age sites and set of environmental variables from the postglacial lowlands in Western Poland. We will propose a machine-learning system which adopts CART through bootstrap simulation to incorporate uncertainty into the spatial model and utilise that uncertainty in the decision-making process. Finally, we will describe the relationships between the model and environmental variables and present our results in cartographic form using the principles of decision-tree cartography.

Retrieval of pattern-based information from giga-cells categorical rasters — Concept and new software

Rapid development of computer technology together with the growing availability of giga-scale data sources brings new possibilities to geo-spatial analysis [1, 2]. We define giga-scale datasets as those having size exceeding 109 cells, regardless of their physical scale. They may represent local regions at ultra-high resolution (of the order of centimeters) offered by LiDAR technology or global mosaics of satellite imagery or digital elevation models (DEMs) at medium resolution (of the order of 10-100 meters). Frequently, these giga-scale datasets are categorical rasters - products derived from processing of original data. Examples include land cover/land use (LCLU), landforms, vegetation, and urban maps. In such rasters important information is stored not only at the level of individual cells, but also, and maybe predominantly, at the level of patterns of the categories [3, 4]. Urban structures, plant habitats, geomorphological surfaces, and landscapes are examples of such patterns; they have collective functions and meaning and thus contain valuable information that cannot be inferred at the level of cell-based analysis.

Query and retrieval of land cover patterns

Ability to explore and mine land use/land cover (LULC) maps having high resolution and large spatial extent could significantly enhance the value of such datasets. In this paper we present a method for retrieval of alike scenes from large LULC datasets. Alikeness between the two scenes is defined as similarity between their spatial patterns of class labels. Our method works on the principle of query-by-example; the inputs are the LULC map and a reference scene and the output is a similarity map indicating a degree of alikeness between a given location on the map and the reference scene. The similarity measure is described and its applicability to the National Land Cover Dataset 2006 (NLCD2006) is discussed.

Mapping past human activity using low representative site location datasets and elevation data

Archaeological maps based on the location of sites are strongly biased by the degree of archaeological recognition and inform little about the real pattern of past human activities, especially on areas poorly covered by surveys. Continuous maps and spatial models, independent of the degree of archaeological recognition of the area, can used as a tool for explanation of the patterns of past human activity [1]. There are several methods (see: [2, 3, 4, 1] for details) which couple information about the location of archaeological remnants and variables derived from natural datasets and social and economic variables. These methods use Geographic Information Science technology and statistical algorithms and result in maps of past human activity. Correct models require the user to know the importance of variables what is difficult to proceed on insensibly contrasted areas like temperate lowlands (Jasiewicz, Hildebrandt-Radke 2009). and requires a large amount of representative data so its application is limited only to well recognized areas where data representativeness is not questionable - which does not occur very often. Furthermore, archaeological remnants tend to be clustered also that some areas were examined more thoroughly than others. This leads to the problem of data imbalance. The term refers to any dataset that exhibits an radically unequal distribution between its classes [5, 6].