Recent interest in the analysis and monitoring of urban landscapes using remote sensing data has led to the development of a number of classification methodologies. Urban growth pattern has been studied by Masser and Cheng (2003), Rawashdeh and Saleh (2006), etc. Guindon, et al., (2004) generated two independent land cover products, one spectral-based at the pixel level and the other segment-based using Landsat Thematic Mapper (TM). These classifications were then merged through a rule-based approach to generate a final product with enhanced land use classes with producer accuracies of 78% and 73%. Spatio temporal dynamics for urban sprawl has also been reported (Tian, et al., 2005; Yu and Ng, 2007). Most of these studies have been carried out with high spatial resolution data at the micro level such as individual building mapping, road extraction etc. (Barnsley and Barr, 1997; Barr and Barnsley, 2000; Mesev, 2005) and are based on pattern recognition techniques involving supervised or unsupervised classification. Among the various frameworks in which pattern recognition has been traditionally formulated, the statistical approach has been most intensively studied and used in practice.

More recently, neural network techniques and methods based on the statistical learning theory have received increasing attention. The design of a recognition system also involves definition of pattern classes, sensing environment, pattern representation, feature extraction and selection, cluster analysis, classifier design and learning, selection of training and test samples and performance evaluation. Pattern recognition techniques such as neural network, decision tree, fuzzy theory etc. have been widely used with remote sensing data to identify the patterns in land use classes like urban, agriculture land, etc. (Kwan, et al., 1994; Fukushima, et al., 1998; Gori, et al., 1998; Lee, et al., 2006). Both conventional thresholding strategy such as NDVI (Masek,  et al., 2000) and customized indices such as the Normalized Built-up Indicator (Zha, et al., 2003) have been used to segregate urban from non-urban lands. Ridd (1995) while studying ecosystem aspects of Salt Lake City, USA proposed viewing urban pixels as being composed of linear combinations of three generic land cover components, vegetation, impervious surface and soil (V–I–S model). This technique was applied later to Brisbane, Australia (Phinn, et al., 2002) and Cairo, Egypt (Rashed, et al., 2001). Operational impervious surface mapping methods have also been developed at the United States Geological Survey (Yang, et al., 2003). Percent impervious surface is currently being used as a measure of urban intensity as part of the Landsat-based 2001 National Land Cover Dataset (Homer, et al., 2002). Other approaches have attempted to exploit attributes of image spatial features to delineate urban areas, namely, the density and regularity of these features. Measures of image texture, i.e. radiometric variability, have been suggested (Moller-Jensen, 1990; Gluch, 2002; Arumugam, et al., 2003). Gong and Howarth (1990) applied edge detection and smoothing techniques to generate a spatial pseudospectral road-density band to augment conventional spectral bands in classification. Zhang, et al. (2003) applied a similar approach to study urban change in Beijing. In this case, instead of edge, line detection was applied since the latter better conforms to road patterns (Wang, 1993).

The general problem of recognising complex data patterns with arbitrary orientation, location, and scale still present challenges. For example, coarse spatial resolution data acquired from MODIS Aqua/Terra with 250 m to 1 km presents identification and pattern classification challenges for urban pixels since most of the pixels would represent signatures that are combination of various types of urban entities reflectance. These include spectral mixing of diverse land cover components within pixels, spectral confusion with other land cover features such as fallow, agricultural fields and the fact that urban classes of interest are of the land use and not the land cover category. There are various techniques that can be used to improve the spatial resolution of the images including image fusion, resampling, principal component, etc. Once the image is classified, with the details of land features of interest from remotely sensed data, it becomes useful to the regional decision makers to management planners. In this context, study of urban growth from temporal satellite imageries are required for monitoring their status, extent across time, which give insights into many other underlying processes and issues.

Urbanisation drastically alters the drainage characteristics of natural catchments, or drainage areas, by increasing the volume and rate of surface runoff. Drainage systems are unable to cope with the increased volume of water and are often encountered with the blockage due to indiscriminate disposal of solid wastes. Encroachment of wetlands, floodplains, etc. obstructs floodways causing loss of natural flood storage. Konig, et al., (2002) divided damages from urban flooding into two categories:

• Direct damage—typically material damage caused by water or flowing water.
• Indirect damage—social consequences that are negative long term effects of a more psychological character, like decrease of property values in frequently flooded areas and delayed economical development, for e.g. traffic disruptions, administrative and labour costs, production losses, spreading of diseases, etc.

Urban flooding creates considerable infrastructure problems and huge economic losses in terms of production, as well as significant damage to property and goods. Flooding in urban areas causes large damage at buildings and other public and private infrastructure. Besides, street flooding can limit or completely hinder the functioning of traffic systems and has indirect consequences such as loss of business and opportunity. The expected total damage; direct and indirect monetary damage costs as well as possible social consequences is related to the physical properties of the flood, i.e. the water level above ground level, the extend of flooding in terms of water volume escaping from or not being entering the drainage system, and the duration of flooding. With sloped surfaces even the flow velocity on the surface might have an impact on potential flood damage (Schmitt, et al., 2002).

The main hydrologic parameters that are used to represent urbanisation are (a) impermeable area that is the proportion of the surface area for which precipitation enters the storm-water network directly and (b) time of concentration or velocity of the flow through the basin (Campana,  et al., 2001).  SCS model (U.S. Soil Conservation Service, 1972; Shi, et al., 2007) incorporates the parameter CN (runoff potential) to estimate the overland flow volume based on soil type and urban development characteristics like plot size in residential, commercial and industrial areas, among others. Campana and Tucci (1994) used fuzzy mathematics to estimate the impermeable areas from Landsat-TM images of urban developments in Brazil. They used data from the cities of Sao Paulo, Curitiba and Porto Alegre (population, 16, 2.5 and 3 million inhabitants) to establish a relationship between urban density and impermeable area. The time of concentration, or time to hydrograph peak, of an urban basin is usually obtained by empirical equations or by using equations established for rural basins (Diaz and Tucci, 1987; Porto, et al., 1993). These equations are useful, as a first approximation, but do not satisfactorily describe all local conditions. The time of concentration of a basin can be considered as made up of two components; the time taken by overland flow from the basin headwaters to reach the drainage system, and time spent in the runoff channels (Campana, et al., 2001).

Studies using satellite-derived radiant temperature have been termed as the surface temperature heat islands (Streutker, 2002). The phenomenon of UHI using satellite derived land surface temperature (LST) measurements have been conducted using various satellite data products acquired in thermal region of the electromagnetic spectrum. Currently available satellite thermal infrared sensors provide different spatial resolution and temporal coverage data that can be used to estimate LST. The Geostationary Operational Environmental Satellite (GOES) has a 4-km resolution in the thermal infrared, while the NOAA-Advanced Very High Resolution Radiometer (AVHRR) and the Terra and Aqua-MODIS have 1-km spatial resolutions. Significantly high resolution data come from the Terra-Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) which has a 90 m pixel resolution, the Landsat-5 Thematic Mapper (TM) which has a 120 m resolution, and Landsat-7 Enhanced Thematic Mapper (ETM) which has a 60 m resolution. However, these instruments have a repeat cycle of 16 days (Li et. al., 2004).

NOAA AVHRR data were used to derive LST for studying heat island phenomenon (Balling and Brazell, 1988; Gallo, et al., 1993; Gallo and Owen, 1998; Kidder and Wu, 1987, Roth, et al., 1989; Streutker, 2002). The 1.1-km spatial resolution of these data is found suitable only for small-scale urban temperature mapping. The 120 m spatial resolution Landsat TM Thermal Infrared (TM TIR) data have also been utilized to derive surface temperatures. Carnahan and Larson (1990) used TM TIR data to observe meso-scale temperature differences between the urban and the rural area in Indianapolis. Tanaka, et al., (2005) used Landsat TM data to study the characteristics of heat island phenomenon. Nichol (1994) carried out a detailed study using TM thermal data to monitor microclimate for housing estates in Singapore. Weng (2001, 2003) examined LST pattern and its relationship with land cover in Guangzhou and in the urban clusters in the Zhujiang Delta, China. Kato and Yamaguchi, (2005) have used ASTER and ETM+ data for analysis of urban heat island. Nikolakopopulos, et al., (2003) have used Landsat-5 TM and Landsat-7 ETM+ data for creating the temperature profile of Alfios River Basin. Stathopoulou and Cartalis (2007) have used Landsat ETM+ data to identify daytime urban heat island using Corine land cover data for major cities in Greece. Using a Landsat ETM+ imagery of City of Indianapolis, IN, USA, Weng, et al., (2004) examined the surface temperature UHI in the city. They derived LST and analysed their spatial variations using Landsat ETM+ thermal measurements with the urban vegetation abundance and investigated their relationship.

LST is believed to correspond more closely with the UCL (Upper Canopy Layer) heat islands, although a precise transfer function between LST and the near-ground air temperature is not yet available (Nichol, 1994). Research on LST shows that the partitioning of sensible and latent heat fluxes and thus surface radiant temperature response is a function of varying surface soil water content and vegetation cover (Owen, et al., 1998). A higher level of latent heat exchange was found with more vegetated areas, while sensible heat exchange was more favored by sparsely vegetated, such as urban areas (Oke, 1982). This relationship between LST and vegetation abundance (e.g., Carson, et al., 1994; Gallo and Owen, 1998; Gillies and Carlson, 1995; Gillies, et al., 1997; Goward, et al., 2002; Lo, et al., 1997; Weng, 2001), and the relationship to derive biophysical parameters (Carson et al., 1994; Gillies and Carlson, 1995; Gillies, et al., 1997) aid primarily in land cover mapping and change analysis (Lambin and  Ehrlich, 1996; Sobrino and Raissouni, 2000).

Satellite TIR sensors measure top of the atmosphere (TOA) radiances, from which brightness temperatures (also known as blackbody temperatures) can be derived (Dash, et al., 2002). The TOA radiances are the mixing result of emitted radiance from the Earth’s surface, upwelling radiance from the atmosphere, and  downwelling radiance from the sky. The difference between the TOA and land surface brightness temperatures ranges generally from 1 to 5 K in the 10–12 μm spectral region, subject to the influence of the atmospheric conditions (Prata, et al., 1995). Therefore, atmospheric effects, including absorption, upward emission, and downward irradiance reflected from the surface (Franca and Cracknell, 1994), must be corrected before land surface brightness temperatures are obtained. These brightness temperatures should be further corrected with spectral emissivity values prior to the computation of LST to account for the roughness properties of the land surface, the amount and nature of vegetation cover, and the thermal properties and moisture content of the soil (Friedl, 2002). Two approaches have been developed to recover LST from multispectral TIR imagery (Schmugge, et al., 1998). The first approach utilizes a radiative transfer equation to correct the at-sensor radiance to surface radiance, followed by an emissivity model to separate the surface radiance into temperature and emissivity (Schmugge, et al., 1998). The second approach applies the split-window technique for sea surfaces to land surfaces, assuming that the emissivity in the channels used for the split window is similar (Dash, et al., 2002). Land surface brightness temperatures are then calculated as a linear combination of the two channels. A major disadvantage of this approach is that the coefficients are only valid for the data sets used to derive them (Dash, et al., 2002). In other words, a set of thermal responses for a specific landscape phenomenon or process measured using a specific TIR sensor cannot be extrapolated to predict the same TIR measurements either from other sensors, or from images recorded at different times using the same sensor (Quattrochi and Goel, 1995).

Estimation of emissivities for ground objects from passive sensor data has been measured using different techniques including the normalized emissivity method (Gillespie, 1985), thermal spectral indices (Becker and Li, 1990), spectral ratio method (Watson, 1992), Alpha residual method (Kealy and Gabell, 1990), NDVI method (Valor and Caselles, 1996), classification-based estimation (Snyder, et al., 1998), and the temperature emissivity separation method (Gillespie, et al., 1998). These techniques are also useful to separate temperatures from emissivities. Lack of knowledge of emissivity can introduce an error ranging from 0.2 to 1.2 K for mid-latitude summers and from 0.8 to 1.4 K for the winter conditions for an emissivity of 0.98 and at the ground height of 0 km, when a single channel method of LST estimation is used (Dash, et al., 2002). Moreover, it may not be practical to measure emissivity values pixel-by-pixel, since numerous factors are involved. Snyder, et al. (1998) proposed to use kernel methods applied to three bidirectional reflectance distribution function (BRDF) models (a geometric model for sparse vegetation, a volumetric model for dense vegetation, and a specular model for water and ice), so that each pixel can be categorized into 1 of the 14 emissivity classes based on conventional land cover classification and dynamic and seasonal factors.

The basis for using NDVI in LST estimation is that the amount of vegetation present is an important factor and NDVI is used to infer general vegetation conditions. The combination of LST and NDVI by scatter plot results in a triangular shape (Carson, et al., 1994; Gillies and Carlson, 1995; Gillies, et al., 1997). The slope of the LST–NDVI curve has been related to soil moisture conditions (Carson, et al., 1994; Gillies and Carlson, 1995; Gillies, et al., 1997; Goetz, 1997; Goward, et al., 2002), and the evapo-transpiration of the surface (Boegh, et al., 1998). Several methods have been developed to interpret the LST–NDVI space, including: (1) the triangle method using soil–vegetation–atmosphere transfer (SWAT) model (Carson, et al., 1994; Gillies and Carlson, 1995; Gillies, et al., 1997); (2) in situ measurement method (Friedl and Davis, 1994); and (3) remote sensing based method (Betts, et al., 1996). However, difficulties still exist in interpretation of LST for sparse canopies because the measurements have combined the temperature of the soil and that of the vegetation, and the combinations are often nonlinear (Sandholt, et al., 2002).  Interpreting thermal data and images of temperature distribution over an area is often not easy because of many complex factors involved. The most influential factors for controlling the UCL heat island are the distribution of surface cover characteristics, and urban morphology, such as building materials, geometry, and density (Oke, 1982). Each component surface in urban landscapes exhibits a unique radiative, thermal, moisture, and aerodynamic properties, and relates to their surrounding site environment. The myriad of the component surfaces and the spatial complexity, when mosaicked create a limitless array of energy balance and microclimate systems, confiscating urban meteorologists from drawing any generalization (Oke, 1982).