Date of submission: 12
Predicting Change – Markov Chain
Download 0.72 Mb.
|
- Navigate this page:
- 4. Study Area
- 5. Methodology
- 5.1 Data Sources
3.7 Predicting Change – Markov ChainAfter the transition potentials have been created the Land Change Modeler provides tools for a dynamic land cover change prediction. After specifying an end date, the quantity of change in each transition can either be modelled through a Markov Chain analysis. A Markov chain is a process with a finite number of states in which the probability of being in a particular state at step n + 1 depends only on the state occupied at step n and not on the sequence of events that preceded it. The MARKOV module is used in the LCM to calculate the transition probability matrix from two input maps and uses this to create a projection of the transition potentials into the future. 4. Study AreaThe study comprises the urbanised area of Austin, Texas, as delimited by the US Census Bureau in 2010. Austin is located in central Texas, which is in the American South West. 
Figure 3: Location of Austin in the USA and Texas Austin is the state capital of Texas and the urban area delineated by the US Census bureau has an area of 1365 km2. It is predominantly located within Travis county, though also covers parts of Hays, Williamson and a very small part of Caldwell county. Austin is the thirteenth most populous city in the USA, with a population of 820,611 as at 2011 (US Census Bureau, 2011). The Austin urban area is shown below, along with the primary and secondary roads in the region. 
Figure 4: Austin, Texas. The primary/secondary roads downloaded from the US Census bureau are present Austin presents a very interesting area for studying urban growth. The Sierra Club, an influential American environmental organisation, ranked Austin highly amongst sprawl threatened cities in the US in a 1998 report (Sierra Club 1998) and noted that from 1990-1998 on average 1000 people per month had moved to the Austin area. The city has continued to develop rapidly thanks to a bouyant local economy and between 2000 and 2006 was the third fastest growing large city in the US (CNN Money, 2007). The city is considered a major centre for high technology and a number of global companies have operations within the city, including Apple, Oracle and Google, amongst many others. Urban sprawl has been acknowledged to be an issue and the amount of land occupied by Austin’s urban area has continued to rise. The map below, produced by the city of Austin GIS team, displays Austin’s urbanised area from 1970 – 2004 and shows a steady expansion. 
Figure 5: Austin’s urban area expansion from 1970 – 2004. Source: City of Austin GIS Team In 1998 Smart Growth planning was initiated by Mayor Kirk Watson in order to combat sprawl and make Austin a more environmentally conscious city, along the lines of a city like Portland, Oregon (Summerville, 2011). Tax incentives and expedited building approvals have been made to encourage developments in the Central Business District of Austin and whilst this has created a large amount of residential development within the urban core of the city, the thousands of new units being created there only represent a tiny percentage of the total residential units created on the urban periphery (City of Austin, 2012). There simply are very few land availability constraints in the territory surrounding the city, so urban sprawl remains a concern in Austin. 5. MethodologyDOWNLOAD Landsat Data US Census files Reference Material IMAGE PRE-PROCESSING Geo-correction Projection
1988
1995 2003
2010 ΔNDVI Epoch 1 ΔNDVI Epoch 2 ΔNDVI Epoch 3 Minus Agricultural Variability INCREASE/DECREASE IN NDVI 3x3 Modal Filter & Accuracy Assessment URBAN GROWTH This chapter provides an overview of the methodology employed in order to answer the aims and objectives set out. The chapter is broadly spilt into two parts – 1) methods for change detection analysis of the satellite imagery and 2) methods for applying the Multi-layer Perceptron Neural Network model to the change detection results to firstly create a map for 2010 which could be validated against the actual land cover change and secondly produce a prediction for the future.
The materials for this study relied completely on what was available on the internet. Satellite imagery was downloaded from the United States Geological Survey’s “Earth Explorer” website http://earthexplorer.usgs.gov/ which provides over 16,000 Landsat images that are freely available for download. Landsat 5 was launched in 1984 and this (along with Landsat 4) marked an introduction of the Thematic Mapper (TM) sensor. This provided a significant improvement over the previous Landsat missions as the spatial resolution was increased to 30 metres and three additional spectral bands were added, two in shortwave infrared and one in thermal infrared. (Williams et al, 2006). Landsat 5 has an altitude of 705km which provides a ground swath width of 185 kilometres. The temporal resolution is 16 days. Remarkably, Landsat 5 has significantly exceeded its designed life expectancy and as of March 2012 celebrated 28 years in space, 25 years beyond the original 3 year mission that was planned. Due in part to its significant temporal archive, the Landsat-5 TM dataset was utilised for this study and four Landsat scenes were chosen. This provided the core data that was used in this study. Images from mid-July to mid-August were selected as these can provide better land cover detection in the fast vegetation growth season (Ji et al, 2006) and the importance of choosing near anniversary dates was taken into account to avoid seasonal differences in vegetation skewing the NDVI results. Images were also selected that displayed zero cloud cover over the study area. All data were received in the UTM projection Zone 14N, using the World Geodetic System 1984 datum.
5.1.2 – US Census Bureau shapefiles A number of other pieces of data were also used in this study. The US Census TIGER files for 2010 were used to download county information for Texas, the US urban areas (clipped to the Austin area), census tracts for Austin and primary and secondary road data (US Census Bureau, 2010). This data was all downloaded in shapefile format and was in the Global Coordinate System North American Datum of 1983 (GCS NAD83). Table 3: US census bureau shapefiles
5.1.3 –Reference data Various data was also downloaded from the City of Austin GIS datasets website ftp://ftp.ci.austin.tx.us/GIS-Data/Regional/coa_gis.html in order to help with accuracy assessment for the study. This excellent resource provided a number of pieces of data crucial to the study, including aerial photography which aided in classification and accuracy assessment as well as shapefiles for the City of Austin parks and land use information. The parks layer was used in the accuracy assessment stage and the land use information played a critical role in utilising the NDVI differencing method, as this gave information on the parts of the study area that were classified as agricultural. Using this, it was possible to filter out the agricultural signal in the study area, which can otherwise cause significant noise when using this method for classifying urban growth. All of the data downloaded from the Austin GIS website was in the Texas Central State Plane projection, using the North American Datum of 1983. Table 4: City of Austin GIS Data, used as reference data for the study
Download 0.72 Mb. Share with your friends:
|