a Department of Geography, San Diego State University, 5500 Campanile Dr., San Diego, CA 92182-4493, United States
John R. Weeks
a Department of Geography, San Diego State University, 5500 Campanile Dr., San Diego, CA 92182-4493, United States
Douglas A. Stow
a Department of Geography, San Diego State University, 5500 Campanile Dr., San Diego, CA 92182-4493, United States
b Department of Geography, 1832 Ellison Hall, University of California Santa Barbara, Santa Barbara, CA 93106-4060, United States
Keith C. Clarke
b Department of Geography, 1832 Ellison Hall, University of California Santa Barbara, Santa Barbara, CA 93106-4060, United States
Ter Sub-Saharan Africa rapid urban growth combined with rising poverty is creating diverse urban environments, the nature of which are not adequately captured by a elementary urban-rural dichotomy. This paper proposes an alternative classification scheme for urban mapping based on a gradient treatment for the southern portion of the Westelijk African country of Ghana. Landsat Enhanced Thematic Mapper Plus (ETM+) and European Remote Sensing Satellite-2 (ERS-2) synthetic aperture radar (SAR) imagery are used to generate a pattern based definition of the urban setting. Spectral combination analysis (SMA) is used to classify a Landsat toneel into Built, Vegetation and Other land covers. Landscape metrics are estimated for Built and Vegetation land covers for a 450 peettante uniform grid covering the examine area. A measure of texture is extracted from the SAR imagery and classified spil Built/Non-built. SMA based measures of Built and Vegetation fragmentation are combined with SAR texture based Built/Non-built maps through a decision tree classifier to generate a nine class urban setting schrijfmap capturing the transition from unsettled land at one end of the gradient to the klein urban core at the other end. Training and testing of the decision tree classifier wasgoed done using very high spatial resolution reference imagery from Google Earth. An overall classification agreement of 77% wasgoed determined for the nine-class urban setting opbergmap, with user’,s accuracy (commission errors) being lower than producer’,s accuracy (omission errors). Nine urban contexts were classified and then compared with gegevens from the 2000 Census of Ghana. Results suggest that the urban classes appropriately differentiate areas along the urban gradient.
Ter the coming decades most of the world’,s land voorkant and land use switch (LCLUC) is predicted to take place ter the tropics, where population is growing the fastest (DeFries, Asner and Foley 2006). United Nations’, projections estimate that virtually all of the world’,s population inbetween now and the middle of this century will emerge ter the cities of the developing world, (United Nations Population Division 2012) driven by natural increase te both urban and rural areas, and by continued migration from rural to urban areas spil people search for economic opportunities (Lee 2007). Urbanization is shaping landscapes te and around cities through densification and sprawl, while at the same time enlargened interaction among cities is creating fresh hybrid landscapes where rural and urban livelihoods overlap (Lambin et reeds. 2001, Seto et alreeds. 2012). The rapid rhythm of latest urbanization is reshaping the morphology and function of cities around the world (Longley 2002), and while research has found that urban growth and the request for land conversion has bot driving habitat fragmentation (Wickham, O’,Neill and Jones 2000), little is known about how the urban landscape itself is switching spil cities grow (Liu and Herold 2007, Seto and Shepherd 2009). Urban environments are becoming increasingly diverse and a elementary urban-rural dichotomy fails to capture that diversity (Champ and Hugo 2004).
Urban mapping increasingly relies on the use of satellite imagery through the development of objective, automated and replicable methodologies for the identification of human-induced land covers (Pumain 2004). The physical characteristics of urban places generate spatial and spectral signatures that are readily captured ter remotely sensed gegevens (Elvidge et reeds. 2004). Spil a result, detection and monitoring of the urban environment at global, regional and local scales depends more and more on the use of such gegevens (Potere et alhoewel. 2009, Puny 2005, Lu and Weng 2006). Te developing countries, where urbanization is taking place at the fastest rates (United Nations Population Division 2014), the geographic comprehensiveness of satellite imagery has made it a useful contraption for quantifying and monitoring the distribution and growth of human settlements (Harris and Longley 2002, Weeks 2004). The Landsat Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+) and Operational Land Imager (OLI) satellite systems provide an extensive and accessible archive of moderate spatial resolution (
30 m) imagery that has bot successfully used to monitor urban areas and settlements ter a broad range of environments (Puny 2005, Seto and Fragkias 2005, Lu and Weng 2008). Ter equatorial and tropical regions where cloud voorkant is a common problem for optical remote sensing, radar imagery is an alternative gegevens source (Rogan et nu. 2003) that has successfully bot used for human settlement detection (Stasolla and Gamba 2008) and for urban mapping (Haack and Bechdol 2000). While measures of texture extracted from the Synthetic Aperture Radar (SAR) imagery have bot found to improve land voorkant and land use mapping (Herold, Haack and Solomon 2004), detect building density (Dell’,Acqua and Gamba 2003) and differentiate informal from formal settlements (Dell’,Acqua, Stasolla and Gamba 2006), applications that combine radar and optical imagery have shown to successfully detect human settlements (Haack et ofschoon. 2002, Tatem, Noor and Hay 2004).
Definitions of settlements spil urban are generally based on an arbitrary threshold set spil the split inbetween rural and urban places without accounting for differences te land use power, function or heterogeneity (Seto et alhoewel. 2012). However, ter urban environments different types and densities of buildings and built surface materials, spil well spil vegetation, can vary within brief distances (Cadenasso, Pickett and Schwarz 2007). Rising suburbanization trends are forming edge cities that are increasingly facilitating urban spread into rural areas (Zipperer et reeds. 2000) and blurring the distinctions inbetween rural and urban places (Hugo, Champ and Lattes 2003). The diffuse transition inbetween urban centers and the countryside is described by Antrop (2004) spil a elaborate combination of land uses with diverse and fragmented morphology. This heterogeneous transition zone that extends inbetween urban and rural places requires further identification and classification.
Albeit most research on urban spaces proceeds to use a elementary urban-rural dichotomy, there have bot attempts to characterize urban environments through gradient approaches based on measures of landscape fragmentation. Research on urban ecosystems has focused on examining the interaction inbetween habitat fragmentation and ecological function (Breuste, Niemelä, and Snep 2008, Kü,hn and Klotz 2006). Measures of landscape fragmentation have also bot used te studies of spatial patterns of urban form (Yang et nu. 2011, Van den Voorde, Pandjesjas and Canters 2011) and growth (Luck and Wu 2002, Weng 2007). Research te the fields of landscape ecology and population have proposed the use of continuous measures of degree of urbanization that combine proportions of land voorkant with population characteristics (McDonnell and Hahs 2008, Weeks, Larson and Rashed 2003), and measures of landscape pattern with socio-economic indicators (Toit and Cilliers 2011, Weeks, Larson and Fugate 2005). Thesis studies that integrate gegevens collected ter censuses or surveys with imagery derived gegevens have two disadvantages: (a) an urban gradient cannot be calculated ter the absence of those socioeconomic gegevens, and (b) since its definition depends upon such gegevens, an urban gradient cannot –,without becoming tautological-be used directly to predict a population’,s socioeconomic characteristics. The objective of this examine is to develop and test a pattern-based classification scheme for the urban setting, using an gradient treatment based solely on remotely sensed imagery that exploits quantitative measures of spatial patterns of built and vegetation land voorkant for the purpose of advancing population and health studies. This pattern-based definition of the urban setting permits differentiating a range of urban environments deepening the understanding of spaces defined spil place of residence ter demographic and public health studies. Gegevens and the applications setting are drawn from a investigate area te southern Ghana.
Two. Investigate area and methodology
Two.1. Investigate area and period
Urbanization te Ghana is spreading at a swifter rhythm than among most of its Westelijk African neighbors. The 2010 Census of Population and Housing exposed that more than half of the country’,s population resided ter urban areas, a figure that the UN projects to reach three quarters by 2050. Ghana Statistical Service (GSS) estimates that population te the Greater Accra Metropolitan Area enhanced from under 1.Five million te 1984 to almost Trio million ter 2000, and then to the Four million mark te 2010. However, urbanization is taking place not only te the capital (Accra) and other major cities (especially Kumasi), but also te smaller settlements both close to and far away from cities (Moller-Jensen and Knudsen 2008).
Studies of land voorkant and land use switch ter Ghana have found that migration is linked to decreasing woodlands te northern Ghana (Pabi 2007, Braimoh 2004), that ter the Western region the most predominant switches are linked to mining, farming, lumbering, fuel wood collection and urbanization (Kusimi 2008), and that ter the Accra region urbanization is the major driver of landscape transformation (Yorke and Margai 2007). Te the capital city of Accra, urban expansion wasgoed mapped inbetween 1985 and 2002 with Landsat imagery, showcasing a swift and unplanned spread of the city into its hinterland (Mø,ller-Jensen and Yankson 1994, Mø,ller-Jensen, Kofie and Yankson 2005). Yeboah (2003) describes the emergence of higher-quality residential sprawl te the peri-urban and rural localities adjacent to Accra’,s metropolitan area.
The explore area from which gegevens are drawn for this analysis is located te southern Ghana, consisting of Legitimate districts, including all of the Greater Accra Region (which comprised Five districts te 2000) and 13 adjacent districts ter the Central, Eastern and Volta regions shown te figure 1 .The coastal regions of Ghana have seen a stable increase ter population growth spil the capital city Accra attracts a stable flow of migrants te search for opportunities. Accra’,s metropolitan area alone witnessed its population dual inbetween the mid 1980’,s and the beginning of 2000, when the last census took place. The examine period for this research is the early part of the decade beginning te 2000. The probe area includes Accra and Tema, and their metropolitan fringes, periphery and hinterland. The districts selected for this investigate open up overheen portions of Accra’,s neighboring regions defined here spil areas that will likely be influenced by urban sprawl and other effects from switches ter Accra ter the near future. It is composed of a diverse landscape ranging from purely rural to central city (i.e., core) urban.
Explore area depicted te crimson within Ghana (gray) and the Gulf of Guinea te blue. Polygons within the explore area represent census districts, while ter the surplus of Ghana represent regions (states).
The year 2000 wasgoed selected spil the probe period to coincide with the Ghanaian population and housing census which permitted drawing comparisons inbetween the landscape pattern based definition of the urban setting and a range of demographic variables. All the analyzed and classified imagery wasgoed selected to match spil closely spil possible to 2000 time framework spil wasgoed all the very high spatial resolution imagery used spil reference gegevens.
Urban setting is characterized here using a uniform grid covering the explore area through the use of satellite imagery and geographic information system (GIS) technologies. Landsat ETM + imagery wasgoed analyzed through spectral combination analysis (SMA) and classified into Built and Vegetation land covers. Synthetic aperture radar imagery from the ERS-2 satellite wasgoed used to estimate a measure of radar backscatter texture and classified into a Built/Non Built land voorkant ordner. Landscape metrics are estimated for the SMA based Built and Vegetation land covers and combined with the radar texture based Built/Non Built schrijfmap through a decision tree classifier te order to generate a classification of degree of urbanization ( Figure Two )
Flow chart for urban setting classification based on the combination of Landsat 2000 imagery and ERS-2 2000 imagery
Two.Two.1 Landsat imagery and processing
A cloud-free 30 m spatial resolution Landsat ETM+ terrain corrected photo captured for path 193 and row 56 on 26 December 2002 wasgoed selected–,he only cloud-free ETM+ photo captured within the period 1999–,2003. Pre-processing of the picture consisted of masking waterbodies, sand flats and fire scars to minimize the confusion of land voorkant classes. Spectral combination analysis (SMA) wasgoed applied to the masked ETM+ pic to estimate sub-pixel fractions of endmembers and the derived fraction photos were used to generate a ordner of Built and land voorkant based on a hard (majority) classification.
The pre-processing of the Landsat ETM+ toneel included applying waterbodies, sandbars and fire-scar masks. The waterbodies mask wasgoed extracted from a Land Use land Voorkant Schrijfmap for 2000 digitized on Landsat Imagery by the Center for Remote Sensing and Geographic Information (CERSGIS) of the University of Ghana, Legon. The CERSGIS waterbodies layer included reservoirs, dams and rivers. Te addition to the CERSGIS waterbodies layer an unsupervised classification wasgoed used to incorporate smaller lagoons and reservoirs that had bot missed by the land use land voorkant schrijfmap. The resulting improved waterbodies layer wasgoed by hand edited to include salt ponds and wetlands by digitizing directly on the Landsat ETM+ toneel and corroborating visually with Google Earth imagery. The Google Earth Imagery used to verify added water features include pansharpened Landsat (15m), Spot Five (Two.5m) and DigitalGlobe QuickBird Two (65 cm) that correspond to the most current available dates. Atmospheric correction wasgoed considered unnecessary given that a single date Landsat picture wasgoed classified based on signatures derived from the same picture (Song et alhoewel. 2001).
Given the spectral similarity of bright sand caf and impervious surfaces, the decision wasgoed made to mask out sand flats te order to reduce confusion inbetween the two land voorkant classes. Bright sand flats were digitized on Google Earth using the most current available very high spatial resolution imagery which includes Pansharpened QuickBird Two, Spot Five. Wij assumed that sand flats are unlikely to have converted from built or vegetated land voorkant, which led us to determine to use the most current very high spatial resolution imagery available. A fire scar mask wasgoed also created to eliminate areas of savanna vegetation burned instantaneously prior to the photo acquisition date, to avoid confusion with the dark (shade) endmember. The fire scar mask wasgoed created using a supervised classification of a principal components transformed photo (Hudak and Brockett 2004).
The resulting masked picture of digital number (DN) values for six (all multispectral except thermal infrared) wavebands wasgoed analyzed using spectral combination analysis (SMA). SMA extracts sub-pixel information by assuming that the spectral reflectance of a pixel is the product of the linear combination of the spectra of zuivere components or endmembers (Lu and Weng 2008). Even however SMA wasgoed originally developed to classify natural environments (Adams, Smith and Gillespie 1993, Roberts et ofschoon. 1998), the mechanism wasgoed adapted to urban landscapes by Ridd (1995) to represent the land voorkant of Salt Lake City spil a combination of vegetation, impervious surface and soil (VIS). The pixel un-mixing algorithm constrains the resulting fractions to sum to 1 for each pixel while each individual fraction is non-negative (Phinn et hoewel. 2002), spil is described ter the following equation:
where spectral combination Riλ, is modeled at location spil the sum of the fractions fmi of M pic end-members rmλ, plus a residual ε,iλ, at waveband λ,. Ter addition to estimating fractions for each end-member the proefje generates a root-mean-square error (RMS) picture that assesses the proefje gezond spil described ter the following equation:
where N is the number of bands and ε,iλ, is a residual term calculated for all pixels at waveband λ,.
The accuracy of the proportions generated by SMA depends on the selection of spectral end-members used to represent unspoiled classes ter the un-mixing process. End-member spectra collected directly from the imagery were supported by a pixel purity index (PPI) (Phinn et reeds. 2002, Rashed et nu. 2003) which ranks pixel values based on how often they are repeated ter the extremes of the spectral distribution of the picture (Boardman, Kruse and Green 1995). Candidate pixels were visually explored on the pan-sharpened (15 m) toneel and on very high spatial resolution satellite pics te Google Earth (Pansharpened QuickBird Two (0.7 m). Given the lack of very high spatial resolution imagery matching the date of the Landsat ETM+ toneel (only 2% of the probe area), the decision wasgoed made to expand the time framework for the reference imagery to voorkant 1998 to 2004 (27% of the investigate area) ( Figure Trio ).
Extent of the investigate area delineated with a crimson boundary on Google Earth and extent of the very high spatial resolution imagery available on Google Earth ter the 1998 to 2004 time period. 27% of the investigate area is covered by very high spatial resolution imagery te the 1998–,2004 time period.
SMA models were run on different sets of candidate end-members and the resulting fractions and RMS pictures were evaluated for goodness of gezond. Models producing fractions inbetween 0 and 1 and maximum RMS error under a threshold were considered good models. Models that didn’,t getraind those parameters had their end-member refined ter an iterative process until the optimum set of end-members wasgoed identified. The final end-member selection consisted of five zuivere signatures, one for green vegetation (pixels selected from forested areas), non-photosynthetic vegetation (pixels selected from savannah areas), soil (pixels selected from patches of nude soil or mud), impervious surface (pixels selected from built patches) and shade (pixels selected from areas te the shadows of ridges) ( Figure Four ).
DN values for endmembers selected using the pixel purity index for tape Trio (crimson) vs tape Four (near infrared). Selected Five end-members include Impervious surfaces, Non-photosynthetic vegetation, Shade, Soil and Vegetation.
The resulting Landsat-derived SMA fractions were input to a series of discrete threshold classifiers to identify and ordner Vegetation and Built land voorkant classes. ETM+ pixels with more than fifty procent impervious surface were classified spil Built land voorkant. The land voorkant proportions resulting from the SMA demonstrated that within urban areas shade played an significant role ter capturing building shadows and dark pavement. Ter order to capture shadows and dark pavement voorkant, contextual information wasgoed used to enhance a threshold classifier. Large settlements were delineated through visual inspection of the pan-sharpened (15 peettante) Landsat ETM+ pic, and pixels found within those areas with proportions of overheen fifty procent shade and twenty five procent impervious surfaces were also classified spil Built. Pixels modeled spil having more than fifty procent vegetation were classified spil Vegetation voorkant. Results from the SMA confirm previous research identifying that shade is also largely associated with vegetated areas where trees personages and contain substantial amounts of shade (Lu, Moran and Batistella 2003). A normalized difference vegetation index (NDVI) wasgoed calculated from ETM+ wavebands Three and Four, and compared to the proportions of vegetation and shade produced by the SMA, confirming the overlap of vegetation and shade ter more strongly vegetated areas. Te order to capture the portion of shade found within the vegetation voorkant, pixels with more than 50% shade and 25% vegetation voorkant were also classified spil Vegetation. The resulting classification product is a 30 m raster land voorkant schrijfmap of the explore area containing Built, Vegetation and Other land voorkant classes.
Two.Two.Three SAR imagery and processing
While optical sensors are limited by lack of transmission of brief to medium wavelength electromagnetic energy through clouds and precipitation, synthetic aperture radar (SAR) sensors are capable of transmitting and receiving microwave energy that is sensitive to physical characteristics of land surfaces such spil roughness, morphology and geometry ter most atmospheric conditions (Soergel 2010). Applications of SAR imagery for urban and built area mapping have proven to be very effective, given the high come back characteristic of man-made features (Haack and Bechdol 2000).
ERS-2 radar imagery collected te the C plakband (Five.6 cm) with 12.Five m spatial resolution wasgoed acquired for the probe area from the European Space Agency for three orbits: 18370 collected on October 25 1998, 19601 collected on January Nineteen 1999, and 41373 collected on March 20 2003. Pre-processing and processing of the radar imagery wasgoed conducted for each orbital pass separately. Pre-processing involved applying a terrain correction algorithm and a speckle reduction filterzakje while the processing included the estimation of a measure of texture that is then classified spil Built or Non-Built land voorkant.
Ground range photos were pre-processed using the NEST toolbox developed by the European Space Agency (Engdahl et alhoewel. 2012). A range Doppler terrain correction algorithm wasgoed implemented for terrain correction and radiometric normalization, using a 30 m spatial resolution Global Digital Elevation Prototype (GDEM V2) derived from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) satellite sensor and precise orbit files from the Delft Institute for Earth-Oriented Space Research. Ter addition, a SAR simulation for orthorectification wasgoed used to generate a layover mask. Adaptive filters are commonly used for speckle reduction ter radar imagery because they have the capacity to reduce multiplicative noise (Lee et reeds. 1994). A refined Lee filterzakje wasgoed used to reduce speckle noise generated by the interference of individual scatterers by examining variance te a 7 ×, 7 window and establishing a threshold that detects edges, (Lee et alhoewel. 2009). The terrain-corrected SAR pic and layover mask were closely tested against very high spatial resolution imagery on Google Earth and the DEM te order to verify that areas susceptible to terrain distortion were masked from subsequent processing. Research te settlement mapping has shown that radar imagery is particularly useful ter areas with little terrain where background classes can be defined spil vapid undeveloped surfaces with low radar comebacks against which artificial structures with high comebacks lightly stand up (Haack and Slonecker 1994). Through visual inspection, areas located at higher elevations were identified spil irregular naked rock formations generating mixed comebacks and foreshortening distortions which appeared to be missed by the terrain correction and layover mask. After close examination of the radar backscatter against optical imagery, the decision wasgoed made to expand the layover mask te areas located above 200 m elevation using a 200 m buffer to eliminate any remaining foreshortening distortions. The expanded mask helps to ensure that the radar backscatter captured by the sensor is only minimally influenced by the radar rafter interacting with the terrain, and is largely a product of its interaction with man-made structures. Eventually, the same waterbodies mask used for the Landsat toneel wasgoed used to mask all water features, salt ponds and wetlands.
Researchers exploit the capability of radar imagery to detect structures and forms through the use of measures of texture. The use of texture extracted from radar imagery permits for the delineation of features and has bot found to improve picture classification of land voorkant and land use (Herold et ofschoon. 2004, Dell’,Acqua and Gamba 2003, Dell’,Acqua et hoewel. 2006). Several measures of texture were tested on the filtered radar imagery and a 9 ×, 9 window wasgoed selected to estimate the standard deviation of the radar backscatter values within the moving window. The selection of the 9 ×, 9 pixels moving window wasgoed based on the assumption that an area of 112.Five m by 112.Five m harshly correspond to the size of a city block that would identify a significant cluster of buildings. The standard deviation texture pic wasgoed then smoothed using a Trio ×, Three pixel moving window te order to liquidate outliers. The resulting variance of radar backscatter wasgoed used spil an indicator of spatial composition of the built environment, where heterogeneous comes back are associated with ingewikkeld artificial landscapes such spil the man-made features characteristic of settlements. A GIS layer depicting settlement locations from Ghana Statistical Service wasgoed used ter combination with very high spatial resolution imagery from Google Earth spil reference to establish a threshold te the radar texture that maximizes the detection of populated areas. The three orbits processed independently were classified spil Built/Non-built land covers based on the defined threshold of radar texture and then mosaicked into a single raster opstopping covering the entire investigate area.
Two.Two.Four Landscape metrics of Built and Vegetation patches
Researchers studying urban form have found that landscape metrics of multi-class land voorkant land use maps derived from classified remotely sensed imagery efficiently portray the complexity of cities (Herold, Scepan and Clarke 2002, Luck and Wu 2002, Pesaresi and Bianchin 2003, Herold, Goldstein and Clarke 2003) and of smaller rural settlements (Wang and Caldas 2014). Studies focusing on capturing the morphological transition inbetween urban and rural places have shown that patch density, mean patch size and patch size variability describe best how fragmented, dispersed and heterogeneous the built environment is (Luck and Wu 2002, Herold et alhoewel. 2003, Seto and Fragkias 2005).
To probe urban structure with landscape metrics requires partitioning the city into homogenous units of analysis (Herold, Couclelis and Clarke 2005). This explore uses a uniform grid cell treatment to estimate landscape fragmentation via the investigate area. Six different cell sizes were tested, ranging from 450 m by 450 m to 14400 m by 14400 m. Our analysis of the resulting landscape metrics indicated that the smaller cells maximized the detection of heterogeneous landscape patterns. Wij concluded that cell sizes larger than 450 m by 450 m denigrated our capability to derive meaningful distinctions among the resulting classes, especially given the spatial resolution of the imagery available to us.
Class and landscape metrics were estimated for the SMA-based Built and Vegetation land voorkant classes for the 450 m grid that corresponds to a 15 by 15 pixel cell which is defined spil the landscape unit of analysis. The degree of landscape fragmentation, dispersal and complexity wasgoed studied by examining spatial patterns of Built and Vegetation patches within the 450 m cell along the urban transition.
The metrics were calculated using FRAGSTATS software (McGarigal and Marks 1995). Class metrics for the Built and Vegetation land voorkant classes included procent land voorkant, patch density, coefficient of variation of patch area and area weighted mean fractal dimension of patches. The heterogeneity of the patterns of Built and Vegetation land voorkant within each 450 m cell wasgoed quantified using percentage of land voorkant and density of patches for each land voorkant class according to the following equation:
where ni is the number of patches of class i and A is the total area te m Two which is then converted into density vanaf 100 hectares.
The variability te patch sizes for the Built and Vegetation land covers wasgoed estimated for each 450 m cell with a coefficient of variation of patch area with the following equation:
where aji is the area of patch ij of class i and ni is the number of patches.
The complexity of the shapes of urban patches wasgoed assessed through estimates of area-weighted mean patch fractal dimension, which have shown to help differentiate inbetween klein dense urban areas and the patchy urban fringe (Batty and Longley 1988, Mesev et alreeds. 1995). Area weighted mean fractal dimension wasgoed estimated for Built and Vegetation patches within each 450 peettante cell according to the following equation:
where aji is the area of patch ij (class i),pij is the perimeter of m number of classes (patch types) and TA is the total area.
Ter addition, an index of contagion wasgoed used to evaluate adjacency and compactness ter the landscape, describing the spatial orkestratie of different land covers within the landscape unit (Yeh and Huang 2009, Herold et alhoewel. 2003, Dietzel et alreeds. 2005).
where pi is the proportion of the landscape that is occupied by patch type (class)i, gik is the number of adjacencies (joins) inbetween pixels of patch types i and k based on the dual count method and m is the number of patch types present te the landscape.
Two.Two.Five Defining the urban setting
The fusion of optical and radar based land voorkant products provides an chance to improve the accuracy of the land voorkant classification obtained from individual sensors. The radar imagery wasgoed particularly useful to detect puny settlements that were missed by the SMA-based treatment on the optical imagery. A scheme is proposed to describe landscape patterns of the built and vegetation land covers that combines measures of landscape fragmentation extracted from the classification of optical imagery with a measure variability of the built class extracted from radar imagery. A nine-class scheme wasgoed created to describe te a rank-order categorical manner the continuous transition of the urban setting. The scheme represents an urban gradient defined spil: Klein urban core, Fragmented large urban patches, Dense and dispersed puny urban patches, Fragmented sub-urban, Scattered settlements, Sparsely populated, Fragmented transition, Fragmented unsettled and Unsettled land ( Figure Five ).
Urban setting classification scheme. On the most urban end klein and dense built land voorkant predominates, fragmentation increases ter both built and vegetation land voorkant spil classes transition towards rural environments. On the most rural end of the scheme klein and dense vegetation land voorkant predominates.
The conceptualization of the nine class scheme is the product of the combination of measures of land voorkant fragmentation (SMA-based classification) with a measure of variability of the radar-based Built class through a series of rules. The rules were defined a priori by analyzing the frequency distributions of six measures of land voorkant fragmentation and the standard deviation of the radar-based Built variable, splitting each variable te two (high and low values) using a natural cracks classification scheme that minimizes within-class variance and maximizes between-class variance. The natural pauze split for each variable wasgoed used to define a series of consecutive rules where the resulting classes ( Figure 6 ) were named and validated by examining a set of representative cells for each class against very high spatial resolution imagery on Google Earth.
Rules used to define the 9 class urban setting classification scheme. Six measures of landscape fragmentation were analyzed and combined with an aggregated measure of radar texture to define the urban setting classification scheme.
Decision tree classifiers are non-parametric models that overeenkomst efficiently with numerical and categorical gegevens, making them a suitable treatment to classify urban setting based on imagery extracted variables such spil land voorkant and measures of texture and morphology from different gegevens sources. A decision tree classifier wasgoed used to classify the 450 m cells or landscape units into one of nine urban setting classes, using the measures of landscape fragmentation estimated on the SMA-based classification of Built and Vegetation land covers and the aggregated radar texture based Built voorkant. This classification technology takes advantage of the spectral characteristics of the optical imagery, the pattern characteristics of the landscape metrics, and the structural characteristics of the radar imagery to generate a range of urban setting classes that describe the varying physical characteristics of the landscape. Nine cell level measures of landscape fragmentation and the cell level classified radar texture were used spil inputs for the decision tree classifier to generate the pattern based classification of the urban setting.
The VIS based Built land voorkant class wasgoed used to stratify the examine area into high procent Built, medium procent Built and low procent Built, clipped to the section of the examine area covered by very high spatial resolution imagery on Google Earth inbetween 1998 and 2004 and then overlaid to the 450 m uniform grid to draw a stratified random sample of 690 cells for training and validating the decision tree. Given that built land represents a very petite portion of the examine area, the more urban strata were oversampled ter order to select a ondergrens of 60 cells vanaf class. The reference cells were visually studied on very high spatial resolution Google Earth imagery and assigned to one of the nine classes based on a series of rules describing the predominant land voorkant type within the cell and level of fragmentation of the land covers found within the cell ( Table 1 ).
Urban setting rules for classification of reference gegevens te Google Earth
The 690 cells were visually studied and assigned to one of the nine classes, and then were partitioned into training and testing samples for a C5.0 boosted tree (Ten trials) to prototype urban setting based on landscape and texture metrics. The boosted tree is an iterative process to improve on the previous tree and reduce the number of errors. A random sample of 349 of the reference cells were used to train the tree and the remaining 344 were used to validate the tree. The violates produced by the boosted tree ( Figure 7 ) are based on the input variables for the reference gegevens.
Schematic illustrating C5.0 boosted tree (Ten trials) input, classification split values and output. The yellow boxes indicate the classified urban setting classes.
Two.Two.6 Accuracy assessment
Accuracy of the Built and Vegetation land covers classified from Landsat-derived SMA fractions wasgoed assessed by comparing the land voorkant classification to very high spatial resolution imagery from Google Earth for the 2000–,2004 timeframe for a random sample of 1000 points. The sample size wasgoed enlargened until a ondergrens of 50 points wasgoed reached for each of the classes (Congalton 1991). Accuracy of the Built/Non-Built classification based on radar texture wasgoed assessed by comparing it to very high spatial resolution imagery from Google Earth for the same time framework for an independent random sample of 900 points. The sample size wasgoed enlargened until a ondergrens of 50 points wasgoed reached for the Built class. Confusion matrices and overall agreement statistics were estimated for each of the classifications.
Accuracy of the urban setting classification wasgoed assessed by comparing the by hand classified 450 by 450 m cells from the validation portion of the reference gegevens against classes predicted by the decision tree for the same sample, a confusion matrix and overall agreement statistics were calculated. The use of a confusion matrix identifies how much misclassification is taking place for each one of the classes but does not permit measuring the magnitude of the errors. Errors that could be considered minor arise when a continuous scale is converted into discrete categories and areas that are relatively similar are assigned to two different but contiguous classes (Foody 2002). Given the gradient nature of the urban setting classification scheme, confusion of adjacent classes wasgoed expected to be substantial but not-problematic. A fuzzy measure of accuracy wasgoed used to differentiate minor and major misclassification errors te the urban setting classification. An independent stratified random sample of 375 cells wasgoed selected, oversampling the most urban classes until a ondergrens of 30 cells wasgoed reached for each class. Following a linguistic scale developed by Woodcock and Gopal (2000), the predicted class for the sample of cells are evaluated te detail against very high spatial resolution Google Earth imagery and scored on a scale of 1 to Five, with 1 meaning absolutely wrong and Five meaning absolutely right ( Table Two ). Cells that are scored spil 1 or Two are considered major errors while cells scored spil Three or Four are considered minor errors and cells scored Five are considered accurately classified.
Fuzzy accuracy linguistic score
Trio.1 SMA based vegetation and built land voorkant ordner
The land voorkant schrijfmap produced using SMA is depicted ter Figure 8 a ) and shows that there is very little separation inbetween Accra and Tema, the two large and sprawling metropolitan areas that predominate the urban system te the region. A network of smaller settlements, such spil the town of Agona Swedru ( Figure 8 b and c ), can be observed spreading east-west following the coastline and scattered mid-size towns extend inland following major roads.
(a) Built (>,50% impervious &, >,25% impervious+>,50% shade) and Vegetation (>,50% vegetation &, >,25% vegetation+>,50% shade) land voorkant extracted from SMA, (b) Landsat ETM+ false color infrared (bands 4-3-2) enlargement of the town of Agona Swedru (c) Built and Vegetation land voorkant extracted from SMA enlargement of the town of Agona Swedru.
An examination of the confusion matrix ( Table Three ) for the SMA-derived land voorkant schrijfmap indicates a high overall agreement of 91%, with producer’,s and user’,s accuracies for both the Built and Vegetation land voorkant classes >, 80%. The producer’,s accuracy indicates the probability that the reference pixels are accurately classified and represent a measure of omission errors while the user’,s accuracy indicates the probability that the classified pixels represent the right category on the ground and are a measure of commission error. The VIS-based Built class has a 15% omission error which indicates that the classification is successfully detecting most of the built environment while a 19% commission error points to a persistent level of confusion inbetween the Built and Other land voorkant class.
SMA based land voorkant classification confusion matrix
Results from the final SMA prototype suggest that distinguishing soil and built land voorkant classes is challenging given the spectral similarity of both classes, and also because of the high prevalence of mixing that occurs ter cities of the developing world where many of the streets remain unpaved (Ridd 1995, Powell and Roberts 2008), or where soil is deposited on paved street surfaces, especially spil runoff after rainy periods.
Trio.Two SAR texture-based Built class
The resulting confusion matrix ( Table Four ) demonstrates that while the opbergmap of the Built class derived from SAR texture has a user’,s accuracy of 67 %, it has a much higher producer’,s accuracy of 94%. The low levels of omission error indicate that the radar texture-based measure of the built environment is successful at detecting the majority of manmade features ter the investigate area, while the high commission error indicates that there is a fair amount of confusion inbetween Built and Non-built classes.
Radar-based Built/Non-built classification confusion matrix
Close visual inspection of the SMA-and radar texture-based maps of the Built land voorkant te conjunction with Google Earth imagery indicated that the radar texture-derived schrijfmap captures Built features for a broader range of settlement sizes ter the examine area ( Figure 9 ). The petite towns of Kwame Adewe and Nsutapon on figure 9 b–,d and e–,g illustrate how the radar-based ordner is able to detect puny towns that are missed by the SMA-based classification of the Built class. While the Built schrijfmap extracted from SMA seems to have a fair amount of omission, the radar texture-extracted Built class, given its finer spatial resolution and imaging mode, is capable of identifying much smaller towns.
(a) Radar-derived Built opbergmap, (b) enlargement of radar-derived Built opbergmap near the town of Kwame Adewe, (c) enlargement of SMA-based land voorkant schrijfmap of the town of Kwame Adwene, (d) Google Earth pic from the town of Kwame Adewene 2003, (e) Radar based built class zoom on the town of Nsutapon, (f) SMA based land voorkant classification zoom on the town of Nsutapon, and (g) Google Earth picture from the town of Nsutapon 2000.
Trio.Trio Classification of the urban setting
Landscape and class metrics estimated for the 450 m cells (58,000 cells covering the explore area) described ter the Methods section and a cell level standard deviation of the radar texture-based Built class were used spil input for the decision tree classifier. The resulting classification combined the ten intermediate variables extracted from optical and radar imagery into a nine class urban setting classification that describes the spatial composition of land voorkant via the urban transition.
The urban setting classification ordner ( Figure Ten ) identifies almost 870 (1.5%) out of the 58,000 cells spil Klein urban areas. Thesis cells are located mainly ter Greater Accra and Tema, te the centers of cities such spil Koforidua and Winneba, and within major coastal and inland settlements ( Figure Ten ). Cells classified spil Fragmented large urban patches compose 1.3% of the explore area and are mostly located within the central areas of major cities and settlements. A similar number of cells is identified spil Dense and dispersed puny urban patches and are found closer to the outskirts of larger cities such spil the area located inbetween Accra and Tema. The Fragmented sub-urban class is restricted to the outskirts of large cities found almost entirely te coastal areas, where urbanization is spreading at a quick tempo. Scattered settlements covering 0.8% of the explore area, on the other arm, are spread around the periphery of intermediate towns, most of them inland. Cells classified spil sparsely populated areas voorkant 1295 cells (Two.2%) and are scattered across the investigate area extending beyond the peripheries of consolidated towns. Ultimately, cells identified spil transitional classes spread into unsettled land following a verhouding pattern that expands beyond the periphery of lodged areas.
450 peettante cell urban setting ordner. Accra, the capital city, and the port of Tema te the central coast are the two largest urban centers within the probe area.
The confusion matrix ( Table Five ) indicates an overall agreement of 77%, it is overduidelijk that user’,s accuracy is lower than producer’,s accuracy. The highest omission errors were found both on the least urban and most urban classes ter the Fragmented transition and Fragmented large urban patches classes. High commission errors are found on the most and least urban classes, spil indicated by figure Ten . The most urban of the setting classes (e.g., Klein urban and Fragmented large urban patches classes), exhibit a fair amount of confusion while the more rural Fragmented unsettled land and Fragmented transition classes voorstelling some confusion. The matrix and figure 11 indicate that while there is some confusion inbetween similar classes (i.e., adjacent classes te a rank-order sense), the automated classification of urban setting classes shows close agreement with the reference gegevens derived from the visual interpretation of very high spatial resolution imagery.
Distribution of predicted versus reference classes. Commission error is present on both completes of the classification scheme. The dark crimson portion of the Fragmented large urban patches tapkast indicates confusion with the Klein urban core class. The dark green portion of the Fragmented unsettled class brochure indicates confusion with the Unsettled land class.
Urban setting classification confusion matrix
1: klein urban
Two: Fragmented large urban patches
Three: Fragmented dense and dispersed urban patches
Four: Fragmented sub-urban
Five: Scattered settlements
6: Sparsely populated
7: Fragmented transition
8: Fragmented unsettled
Results from the fuzzy accuracy assessment indicate a high degree of correspondence when using a score of agreement with the reference gegevens. The right class katern on table 6 includes scores 1 to Three corresponding to “,absolutely right,”, “,good response”, and “,reasonable reaction,”, while the precies class only captures score 1-absolutely right. The average agreement inbetween classification and reference gegevens improved from 61% to 87% when using the right class overheen the precies class. Given the continuous nature of the classification scheme, the use of a broader definition of agreement shows up to be an suitable way of assessing accuracy. The improvements te agreement levels through the use of the fuzzy accuracy treatment indicate that there is notable overlap inbetween adjacent classes, an artifact of the gradient treatment.
Four. Discussion and Conclusions
Rapid urbanization is reshaping the morphology and function of cities globally. By portraying rural and urban areas spil clearly distinct spaces, dichotomous rural/urban classifications overlook the importance of flows of people and products that connect thesis spaces and belie how urban processes are remaking global landscapes far beyond urban areas. Ter this paper, wij have created and evaluated an urban setting classification scheme that attempts to characterize different urban contexts that exist along a gradient inbetween the arbitrary extremes of urban and rural. Wij demonstrate a novel characterization of the urban setting based exclusively on the pattern characteristics of land voorkant distributions. A series of landscape metrics were computed for Built and Vegetated land voorkant maps with the aim of differentiating areas based on the degree of landscape fragmentation.
Two intermediate and independent land voorkant classifications were generated and then integrated, based on our analysis of optical and radar imagery it is overduidelijk that each of the approaches has advantages and limitations. The workflow for the processing of the optical imagery estimates sub-pixel proportions of land voorkant and then classifies them into discrete land voorkant classes ter order to generate distinct land voorkant patches that are further analyzed through landscape metrics. Even tho’ the classification of the Landsat-derived SMA fractions leads to a significant loss of sub-pixel land voorkant detail, the categorical product permits for the analysis of spatial patterns of land voorkant patches that would not be possible to achieve with more continuous spatial gegevens. Results from the accuracy assessment for the SMA based classification of land voorkant indicate that there is a fair amount of omission, meaning that wij are missing some built and vegetation patches, most likely because of their smaller proportions within the 30 m pixel. On the other palm, the radar-based classification had higher commission errors, meaning that there is at least some confusion inbetween the Built and Non-Built classes using just that method. By combining the SMA-and radar-based land voorkant classes the attempt wasgoed made to overcome the limitations of each independent classification and generate a more accurate depiction of the built environment ter the examine area.
Patterns of land voorkant fragmentation were estimated using a 15 pixel by 15 pixel landscape unit to assess heterogeneity and complexity ter patch sizes, and dispersion and interspersion of the land voorkant. This analysis provides an in-depth portrayal of the spatial patterns of land voorkant found within the probe area. The generated urban setting schrijfmap with a spatial resolution of 450 m does not identify individual objects or land voorkant classes but it categorizes the landscape based on their spatial patterns following a gradient treatment. The assumption te our method is that spil city or settlement centers become more densely urbanized the built environment becomes more klein, whereas towards the outskirts of the city the land voorkant conversion brought by urban expansion means higher fragmentation and dispersion. The pattern-based urban setting definition is based on the relative fragmentation of both the built environment and the vegetation land voorkant. A klein urban core bounds the most urbanized end of the spectrum, with a predominant Built land voorkant class and very low levels of landscape fragmentation. Spil distance from the klein urban core increases, the built environment becomes increasingly fragmented providing way to dispersion and interspersion of Vegetation and Built land covers.
At distances from the city center reaching beyond the city thresholds, the landscape switches to scattered settlements and sparsely populated areas, fragmentation of the built environment peaks and is little by little substituted by areas transitioning from their natural state into cleared spaces suggestive of potential settlement. At the least urbanized end of the spectrum, unsettled grounds are identified by lower levels of fragmentation te the vegetation land voorkant, while te transitioning spaces wij start to observe clearings linked to growing vegetation voorkant fragmentation. The pattern-based definition of the urban setting used te this examine captures a broader range of urban environments than do traditional rural/urban classifications. By differentiating the klein urban city center from very fragmented suburban areas and scattered settlements, the urban setting definition identifies significant pattern differences among inhabited spaces. This probe proposes defining urban setting based on characteristics of landscape fragmentation, an treatment that is lightly replicable te other gegevens poor countries. At the same time, it is significant to recognize that the urban setting classification derived ter this investigate is a relative measure of degree of urbanization that is based on the fragmentation characteristics of this particular landscape. Its replicability ter different geographic settings is a subject for future research.
The pattern based scheme wasgoed developed solely on the poot of imagery, the test of its utility lies ter its capability to differentiate socioeconomic characteristics derived from independent sources. While a finish test of its utility remains beyond the scope of this paper, the urban setting classification wasgoed compared to census gegevens summarized at the enumeration area (EA) level (average size 13 Km Two ). A random sample of 2000 EAs wasgoed selected from a total of 5000 covering the examine area and wasgoed categorized based on the predominant urban setting class covering each sampled EA (i.e. the class that covers the majority of the EA). EA level measures of population density and percentage of population occupied ter agriculture were estimated and examined against the EA’,s predominant urban setting class. Results indicate that the highest percentages of population employed ter agriculture are concentrated at the most rural end of the spectrum for scattered settlements, sparsely populated, fragmented transition, fragmented unsettled and unsettled land ( Figure 12 ). It is interesting to note that the klein urban core has a higher percentage of population employed ter agriculture than the fragmented sub-urban class, a result that would seem unlikely at very first view. However it is also worth pointing out that even however the klein urban core class captures the dense city center of Accra and Tema it also captures dense and klein city centers of intermediate cities and major towns. Te those smaller cities the proportion of population working ter agriculture is much higher than te sub-urban areas which are only concentrated around Accra. On the other mitt, population density is highest te the most urban end of the spectrum for the klein urban core, fragmented large urban patches, and dense and disperse puny urban patches classes and decreases significantly with the transition into the unsettled end of the scheme ( Figure 13 ).
Procent working te agriculture across the urban setting classification. The highest percentage of population working ter the agricultural sector is found on the most rural end of the scheme, Fragmented transition, fragmented unsettled and unsettled land.