The detection and analysis of clusters has become commonplace within geographic information scien... more The detection and analysis of clusters has become commonplace within geographic information science and has been applied in epidemiology, crime prevention, ecology, demography and other fields. One of the many methods for detecting and analyzing these clusters involves searching the dataset with a flock of boids (bird objects). While boids are effective at searching the dataset once their behaviors are properly configured, it can be difficult to find the proper configuration. Since genetic algorithms have been successfully used to configure neural networks, they may also be useful for configuring parameters guiding boid behaviors. In this paper, we develop a genetic algorithm to evolve the ideal boid behaviors. Preliminary results indicate that, even though the genetic algorithm does not return the same configuration each time, it does converge on configurations that improve over the parameters used when boids were initially proposed for geographic cluster detection. Also, once conf...
The amount of geographic data has rapidly increased throughout the past decade. Impressive data c... more The amount of geographic data has rapidly increased throughout the past decade. Impressive data collection capabilities combine with the ability to merge datasets, resulting in a rapid growth in both the number and size of geographic datasets. Meanwhile, the number and urgency of social, environmental, and public safety applications that use these datasets is so great that it has been argued that "there is an undisputed national imperative for analysis" of these datasets (Openshaw 1995, p. 4). One way to analyze these datasets is by detecting clusters. However, many cluster analysis methods do not scale well to large datasets. In addition, some of these methods return many nearly-identical results for each cluster, which unnecessarily adds to the post-processing burden. Therefore, it is necessary to develop new methods and algorithms to perform cluster analysis on large datasets.
Background: The United States Environmental Protection Agency's Toxic Release Inventory (TRI) dat... more Background: The United States Environmental Protection Agency's Toxic Release Inventory (TRI) data are frequently used to estimate a community's exposure to pollution. However, this estimation process often uses underdeveloped geographic theory. Spatial interaction modeling provides a more realistic approach to this estimation process. This paper uses four sets of data: lung cancer age-adjusted mortality rates from the years 1990 through 2006 inclusive from the National Cancer Institute's Surveillance Epidemiology and End Results (SEER) database, TRI releases of carcinogens from 1987 to 1996, covariates associated with lung cancer, and the EPA's Risk-Screening Environmental Indicators (RSEI) model. Results: The impact of the volume of carcinogenic TRI releases on each county's lung cancer mortality rates was calculated using six spatial interaction functions (containment, buffer, power decay, exponential decay, quadratic decay, and RSEI estimates) and evaluated with four multivariate regression methods (linear, generalized linear, spatial lag, and spatial error). Akaike Information Criterion values and P values of spatial interaction terms were computed. The impacts calculated from the interaction models were also mapped. Buffer and quadratic interaction functions had the lowest AIC values (22298 and 22525 respectively), although the gains from including the spatial interaction terms were diminished with spatial error and spatial lag regression.
International Journal of Geographical Information Science, 2011
ABSTRACT Although there are many algorithms and statistical tests to detect clustering of geograp... more ABSTRACT Although there are many algorithms and statistical tests to detect clustering of geographical phenomena, such as disease cases, the follow-up task of analyzing the cluster to explain its existence and mitigate the cluster is generally left to the researcher. These cluster detection methods are useful only for part of the process of identifying, understanding, and mitigating disease clusters. This research develops and presents a computer program that uses the information about the shape of the cluster to evaluate the hypotheses about potential causes. To achieve this, the shapes of the clusters are represented by image moment invariant statistics developed in the field of computer vision, and these shape statistics are compared against a database of moment invariants of shapes representative of several geographical processes, such as diffusion along roads and wind diffusion. Experiments using simulated data of different types of disease transmission were carried out, and the ability of the program to accurately classify the different types of diffusion demonstrates the viability of this approach to automated cluster analysis.
International Journal of Remote Sensing, Oct 31, 2012
ABSTRACT An implicit assumption of the geographic object-based image analysis GEOBIA literature i... more ABSTRACT An implicit assumption of the geographic object-based image analysis GEOBIA literature is that GEOBIA is more accurate than pixel-based methods for high spatial resolution image classification, but that the benefits of using GEOBIA are likely to be lower when moderate resolution data are employed. This study investigates this assumption within the context of a case study of mapping forest clearings associated with drilling for natural gas. The forest clearings varied from 0.2 to 9.2 ha, with an average size of 0.9 ha. National Aerial Imagery Program data from 2004 to 2010, with 1 m pixel size, were resampled through pixel aggregation to generate imagery with 2, 5, 15, and 30 m pixel sizes. The imagery for each date and at each of the five spatial resolutions was classified into Forest and Non-forest classes, using both maximum likelihood and GEOBIA. Change maps were generated through overlay of the classified images. Accuracy evaluation was carried out using a random sampling approach. The 1 m GEOBIA classification was found to be significantly more accurate than the GEOBIA and per-pixel classifications with either 15 or 30 m resolution. However, at any one particular pixel size e.g. 1 m, the pixel-based classification was not statistically different from the GEOBIA classification. In addition, for the specific class of forest clearings, accuracy varied with the spatial resolution of the imagery. As the pixel size coarsened from 1 to 30 m, accuracy for the per-pixel method increased from 59% to 80%, but decreased from 71% to 58% for the GEOBIA classification. In summary, for studying the impact of forest clearing associated with gas extraction, GEOBIA is more accurate than pixel-based methods, but only at the very finest resolution of 1 m. For coarser spatial resolutions, per-pixel methods are not statistically different from GEOBIA.
ABSTRACT Background: HRQOL research determines the burden of preventable disease, injuries, and d... more ABSTRACT Background: HRQOL research determines the burden of preventable disease, injuries, and disabilities known to impact physical or mental health. Healthy People goals have identified improvement in quality of life as a central goal since 2000. However, geographic variations in HRQOL at the US county level is a relatively unexplored area. This study’s objective was to fill that gap. Methods: Using the CDC Healthy Days measures from the US County Health Rankings data, 1) maps were created showing the geospatial distribution of mean physical and mental unhealthy days (in the past 30 days), 2) a bivariate map was developed to visually represent HRQOL clusters in the US. Residuals from non-spatial multiple regression models (controlling for demographic and behavioral risks) were analyzed to find counties where expected physical and mental unhealthy days deviate from observed; maps were created that showed the geographic distributions of areas that are outperforming or underperforming expectations. Spatial multiple regressions were conducted to examine geographic variability in the impacts of demographic and behavioral risks. Results: Areas with more unhealthy days clustered in central Appalachia and extending to the lower Midwest while areas with fewer unhealthy days clustered in the upper Midwest and the Northeast corridor. Traditional risks accounted for approximately 49% and 36% of the variance in mean unhealthy physical and mental health days, respectively. Results of the residual analysis identified areas that performed better or worse than expected after accounting for traditional risks. Conclusions: Results suggest clusters of mean unhealthy and healthy HRQOL days in the US. Furthermore, some US counties experience relatively more mean unhealthy mental health days than unhealthy physical health days, and vice versa. Implications for interventions and health service needs are discussed.
Spatial analysis techniques are widely used throughout geography. However, as the size of geograp... more Spatial analysis techniques are widely used throughout geography. However, as the size of geographic datasets increases exponentially, limitations to the traditional methods of spatial analysis become apparent. To overcome some of these limitations, many algorithms for exploratory spatial analysis have been developed. This paper presents both a new cluster detection method based on a genetic algorithm, and PROCLUDE (PROgrams for
This Provisional PDF corresponds to the article as it appeared upon acceptance. Fully formatted P... more This Provisional PDF corresponds to the article as it appeared upon acceptance. Fully formatted PDF and full text (HTML) versions will be made available soon. Permitted water pollution discharges and population cancer and non-cancer mortality: toxicity weights and upstream discharge effects in US rural-urban areas
Many digital elevation models (DEMs) now exist, enabling quantitative assessments of topographic ... more Many digital elevation models (DEMs) now exist, enabling quantitative assessments of topographic change through DEM subtraction. However, many of these DEMs are inherently different - their source data, preprocessing, and interpolation procedures introduce varying amounts of error, which are partially correlated with the local topographic situation (e.g. slope, aspect, and elevation of any given location). Understanding of these differences and potential errors is necessary prior to change detection analysis in order to incorporate the inherent uncertainties into the analysis of topographic change. This study evaluated available DEMs in the Mud 7.5′ United States Geological Survey (USGS) topographic quadrangle in West Virginia, USA. We compare DEMs derived from the USGS digital line graphs (DLGs), the Shuttle Radar Topography Mission (SRTM), a statewide photogrammetric DEM, and the Advanced Spaceborne Thermal Emission and Reflection (ASTER) Global DEM v2, to a LiDAR DEM. Using the L...
ABSTRACT This study investigates machine-learning algorithms and measures derived from RapidEye s... more ABSTRACT This study investigates machine-learning algorithms and measures derived from RapidEye satellite imagery and light detection and ranging (lidar) data for geographic object-based image analysis classification of mining and mine reclamation. Support vector machines, random forests, and boosted classification and regression trees classification algorithms were assessed and compared with the k-nearest neighbour (k-NN) classifier. For geographic object-based image analysis classification of mine landscapes, the use of disparate data (i.e. lidar data) improved overall accuracy, whereas the use of complex, object-oriented variables such as object geometry measures, first-order texture, and second-order texture from the grey-level co-occurrence matrix decreased or did not improve the classification accuracy. Support vector machines generally outperformed k-NN and the ensemble tree classifiers when only using the band means. With the incorporation of lidar-descriptive statistics, all four algorithms provided statistically comparable accuracies. K-NN suffered reduced classification accuracy with high-dimensional feature spaces, suggesting that a more complex machine-learning algorithm may be more appropriate when a large number of predictor variables are used.
The detection and analysis of clusters has become commonplace within geographic information scien... more The detection and analysis of clusters has become commonplace within geographic information science and has been applied in epidemiology, crime prevention, ecology, demography and other fields. One of the many methods for detecting and analyzing these clusters involves searching the dataset with a flock of boids (bird objects). While boids are effective at searching the dataset once their behaviors are properly configured, it can be difficult to find the proper configuration. Since genetic algorithms have been successfully used to configure neural networks, they may also be useful for configuring parameters guiding boid behaviors. In this paper, we develop a genetic algorithm to evolve the ideal boid behaviors. Preliminary results indicate that, even though the genetic algorithm does not return the same configuration each time, it does converge on configurations that improve over the parameters used when boids were initially proposed for geographic cluster detection. Also, once conf...
The amount of geographic data has rapidly increased throughout the past decade. Impressive data c... more The amount of geographic data has rapidly increased throughout the past decade. Impressive data collection capabilities combine with the ability to merge datasets, resulting in a rapid growth in both the number and size of geographic datasets. Meanwhile, the number and urgency of social, environmental, and public safety applications that use these datasets is so great that it has been argued that "there is an undisputed national imperative for analysis" of these datasets (Openshaw 1995, p. 4). One way to analyze these datasets is by detecting clusters. However, many cluster analysis methods do not scale well to large datasets. In addition, some of these methods return many nearly-identical results for each cluster, which unnecessarily adds to the post-processing burden. Therefore, it is necessary to develop new methods and algorithms to perform cluster analysis on large datasets.
Background: The United States Environmental Protection Agency's Toxic Release Inventory (TRI) dat... more Background: The United States Environmental Protection Agency's Toxic Release Inventory (TRI) data are frequently used to estimate a community's exposure to pollution. However, this estimation process often uses underdeveloped geographic theory. Spatial interaction modeling provides a more realistic approach to this estimation process. This paper uses four sets of data: lung cancer age-adjusted mortality rates from the years 1990 through 2006 inclusive from the National Cancer Institute's Surveillance Epidemiology and End Results (SEER) database, TRI releases of carcinogens from 1987 to 1996, covariates associated with lung cancer, and the EPA's Risk-Screening Environmental Indicators (RSEI) model. Results: The impact of the volume of carcinogenic TRI releases on each county's lung cancer mortality rates was calculated using six spatial interaction functions (containment, buffer, power decay, exponential decay, quadratic decay, and RSEI estimates) and evaluated with four multivariate regression methods (linear, generalized linear, spatial lag, and spatial error). Akaike Information Criterion values and P values of spatial interaction terms were computed. The impacts calculated from the interaction models were also mapped. Buffer and quadratic interaction functions had the lowest AIC values (22298 and 22525 respectively), although the gains from including the spatial interaction terms were diminished with spatial error and spatial lag regression.
International Journal of Geographical Information Science, 2011
ABSTRACT Although there are many algorithms and statistical tests to detect clustering of geograp... more ABSTRACT Although there are many algorithms and statistical tests to detect clustering of geographical phenomena, such as disease cases, the follow-up task of analyzing the cluster to explain its existence and mitigate the cluster is generally left to the researcher. These cluster detection methods are useful only for part of the process of identifying, understanding, and mitigating disease clusters. This research develops and presents a computer program that uses the information about the shape of the cluster to evaluate the hypotheses about potential causes. To achieve this, the shapes of the clusters are represented by image moment invariant statistics developed in the field of computer vision, and these shape statistics are compared against a database of moment invariants of shapes representative of several geographical processes, such as diffusion along roads and wind diffusion. Experiments using simulated data of different types of disease transmission were carried out, and the ability of the program to accurately classify the different types of diffusion demonstrates the viability of this approach to automated cluster analysis.
International Journal of Remote Sensing, Oct 31, 2012
ABSTRACT An implicit assumption of the geographic object-based image analysis GEOBIA literature i... more ABSTRACT An implicit assumption of the geographic object-based image analysis GEOBIA literature is that GEOBIA is more accurate than pixel-based methods for high spatial resolution image classification, but that the benefits of using GEOBIA are likely to be lower when moderate resolution data are employed. This study investigates this assumption within the context of a case study of mapping forest clearings associated with drilling for natural gas. The forest clearings varied from 0.2 to 9.2 ha, with an average size of 0.9 ha. National Aerial Imagery Program data from 2004 to 2010, with 1 m pixel size, were resampled through pixel aggregation to generate imagery with 2, 5, 15, and 30 m pixel sizes. The imagery for each date and at each of the five spatial resolutions was classified into Forest and Non-forest classes, using both maximum likelihood and GEOBIA. Change maps were generated through overlay of the classified images. Accuracy evaluation was carried out using a random sampling approach. The 1 m GEOBIA classification was found to be significantly more accurate than the GEOBIA and per-pixel classifications with either 15 or 30 m resolution. However, at any one particular pixel size e.g. 1 m, the pixel-based classification was not statistically different from the GEOBIA classification. In addition, for the specific class of forest clearings, accuracy varied with the spatial resolution of the imagery. As the pixel size coarsened from 1 to 30 m, accuracy for the per-pixel method increased from 59% to 80%, but decreased from 71% to 58% for the GEOBIA classification. In summary, for studying the impact of forest clearing associated with gas extraction, GEOBIA is more accurate than pixel-based methods, but only at the very finest resolution of 1 m. For coarser spatial resolutions, per-pixel methods are not statistically different from GEOBIA.
ABSTRACT Background: HRQOL research determines the burden of preventable disease, injuries, and d... more ABSTRACT Background: HRQOL research determines the burden of preventable disease, injuries, and disabilities known to impact physical or mental health. Healthy People goals have identified improvement in quality of life as a central goal since 2000. However, geographic variations in HRQOL at the US county level is a relatively unexplored area. This study’s objective was to fill that gap. Methods: Using the CDC Healthy Days measures from the US County Health Rankings data, 1) maps were created showing the geospatial distribution of mean physical and mental unhealthy days (in the past 30 days), 2) a bivariate map was developed to visually represent HRQOL clusters in the US. Residuals from non-spatial multiple regression models (controlling for demographic and behavioral risks) were analyzed to find counties where expected physical and mental unhealthy days deviate from observed; maps were created that showed the geographic distributions of areas that are outperforming or underperforming expectations. Spatial multiple regressions were conducted to examine geographic variability in the impacts of demographic and behavioral risks. Results: Areas with more unhealthy days clustered in central Appalachia and extending to the lower Midwest while areas with fewer unhealthy days clustered in the upper Midwest and the Northeast corridor. Traditional risks accounted for approximately 49% and 36% of the variance in mean unhealthy physical and mental health days, respectively. Results of the residual analysis identified areas that performed better or worse than expected after accounting for traditional risks. Conclusions: Results suggest clusters of mean unhealthy and healthy HRQOL days in the US. Furthermore, some US counties experience relatively more mean unhealthy mental health days than unhealthy physical health days, and vice versa. Implications for interventions and health service needs are discussed.
Spatial analysis techniques are widely used throughout geography. However, as the size of geograp... more Spatial analysis techniques are widely used throughout geography. However, as the size of geographic datasets increases exponentially, limitations to the traditional methods of spatial analysis become apparent. To overcome some of these limitations, many algorithms for exploratory spatial analysis have been developed. This paper presents both a new cluster detection method based on a genetic algorithm, and PROCLUDE (PROgrams for
This Provisional PDF corresponds to the article as it appeared upon acceptance. Fully formatted P... more This Provisional PDF corresponds to the article as it appeared upon acceptance. Fully formatted PDF and full text (HTML) versions will be made available soon. Permitted water pollution discharges and population cancer and non-cancer mortality: toxicity weights and upstream discharge effects in US rural-urban areas
Many digital elevation models (DEMs) now exist, enabling quantitative assessments of topographic ... more Many digital elevation models (DEMs) now exist, enabling quantitative assessments of topographic change through DEM subtraction. However, many of these DEMs are inherently different - their source data, preprocessing, and interpolation procedures introduce varying amounts of error, which are partially correlated with the local topographic situation (e.g. slope, aspect, and elevation of any given location). Understanding of these differences and potential errors is necessary prior to change detection analysis in order to incorporate the inherent uncertainties into the analysis of topographic change. This study evaluated available DEMs in the Mud 7.5′ United States Geological Survey (USGS) topographic quadrangle in West Virginia, USA. We compare DEMs derived from the USGS digital line graphs (DLGs), the Shuttle Radar Topography Mission (SRTM), a statewide photogrammetric DEM, and the Advanced Spaceborne Thermal Emission and Reflection (ASTER) Global DEM v2, to a LiDAR DEM. Using the L...
ABSTRACT This study investigates machine-learning algorithms and measures derived from RapidEye s... more ABSTRACT This study investigates machine-learning algorithms and measures derived from RapidEye satellite imagery and light detection and ranging (lidar) data for geographic object-based image analysis classification of mining and mine reclamation. Support vector machines, random forests, and boosted classification and regression trees classification algorithms were assessed and compared with the k-nearest neighbour (k-NN) classifier. For geographic object-based image analysis classification of mine landscapes, the use of disparate data (i.e. lidar data) improved overall accuracy, whereas the use of complex, object-oriented variables such as object geometry measures, first-order texture, and second-order texture from the grey-level co-occurrence matrix decreased or did not improve the classification accuracy. Support vector machines generally outperformed k-NN and the ensemble tree classifiers when only using the band means. With the incorporation of lidar-descriptive statistics, all four algorithms provided statistically comparable accuracies. K-NN suffered reduced classification accuracy with high-dimensional feature spaces, suggesting that a more complex machine-learning algorithm may be more appropriate when a large number of predictor variables are used.
This article appeared in a journal published by Elsevier. The attached copy is furnished to the a... more This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier's archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright a b s t r a c t Coffee retailers often court discriminate consumers through the marketing of single origin coffees with distinct flavor attributes. The Cup of Excellence has become a key mechanism for locating and certifying single origin coffees. In this paper use hedonic regression analysis to examine the price premiums for farmers generated by the origin attributes of 607 Cup of Excellence certified coffees from Central America. We find that while the greatest impact on the price premiums for a pound of coffee is the jury score awarded by coffee cuppers, several geographic factors, including the altitude of the farm, coffee farm size, country of origin and even sub-region of origin have significant effects on price premiums. While the price premiums conferred represent opportunities for farmers who have access to such markets, we argue that the appetite for single origin coffees may also contribute to uneven development.
Uploads
Papers by Jamison Conley