Remote Sensing And GIS Analysis Research Paper

1. Remote Sensing Systems

A number of satellite sensors have been acquiring multispectral data over the entire globe since the 1970s, at a spatial resolution of dozens of meters. The most widely used among these sensors are Landsat Multispectral Scanner (MSS) (resolution of 80 × 80 m), Landsat Thematic Mapper (TM) (resolution of 30 × 30 m), and SPOT High Resolution Visible (HRV) Imaging Instrument (resolution of 20 × 20 m for the multispectral product and 10 × 10 m for the panchromatic product) (Colwell et al. 1983, Lillesand and Kiefer 1994).

Radiance data measured from the Advanced Very High Resolution Radiometer (AVHRR) on the NOAA series of orbiting platforms have been used widely for monitoring changes in the attributes of the Earth’s surface. These data cover the period from June 1981 to date. The daily NOAA AVHRR Local Area Coverage (LAC) data have a spatial resolution of 1.1 km at nadir, which is well suited to monitor seasonal vegetation dynamics at the scale of the landscape. Other useful sensors with a 1 km resolution are ERS’s Along Track Scanning Radiometer (ATSR), SPOT VEGETATION and the Moderate Resolution Imaging Spectroradiometer (MODIS).

Several active microwave sensors are also available: the Synthetic Aperture Radar (SAR) on the ERS-1 and JERS-1 satellites, and the Advanced Synthetic Aperture Radar (ASAR) sensor as part of the forth- coming ENVISAT mission. The cloud-penetrating capability of microwaves make them an ideal tool to work in areas with a permanent cloud cover. Both active and passive microwave sensing are used. Active systems supply their own source of energy or illumination while passive systems sense energy that is naturally emitted or reflected from the Earth’s surface, at extremely low levels however. Note that this enumeration of sensors used for land applications is not exhaustive (see Colwell et al. 1983, Lillesand and Kiefer 1994).

A new generation of fine spatial resolution sensors is due to be launched around the turn of the twenty-first century. These sensors are capable of generating imagery with spatial resolutions as fine as 1 m in panchromatic mode and 4 m in multispectral mode. These systems are also characterized by high geometric precision, short revisit intervals, and rapid data supply. Such imagery will provide greater spatial details on the land surface and open new applications relevant for social sciences (e.g., for urban studies).

2. Land Surface Attributes Measured By Remote Sensing

The electromagnetic radiation reflected and emitted in different wavelength ranges by the surface is controlled by surface attributes, such as the nature of the material, surface moisture, state of the vegetation, canopy structure, etc. Different features can be identified on an image based on the specific ‘spectral signature’ of different surfaces. A good understanding of the biophysical processes controlling the radiation of a surface is necessary to interpret the remotely sensed signal in terms of surface attributes.

2.1 In The Spectral Domain

In the visible and near-infrared range of the electromagnetic spectrum, the land surface generally is characterized using vegetation indices, which are arithmetic combinations of spectral bands. The most widely used of these indices is the normalized difference vegetation index (NDVI). Vegetation indices derived from remote sensing data are related to several variables quantifying vegetation activity and state, such as the fraction of photosynthetically active radiation absorbed by the vegetation canopy, canopy attributes (e.g., green biomass or green leaf area index), state of the vegetation (i.e., vegetation vigor or stress) and instantaneous rates associated with the activity of the vegetation. Seasonal variations of vegetation indices are related to vegetation phenology and biome seasonality.

In the thermal range of the spectrum, land surface ‘skin’ brightness temperature (Ts) can be derived from the thermal channels of satellite sensors. Ts is related, through the surface energy balance equation, to surface moisture availability and evapotranspiration, as a function of latent heat flux. Empirical studies have demonstrated the usefulness of thermal data to locate dense forest boundaries and monitor the growth cycle of crops, including for irrigated systems.

In the microwave range of the spectrum, the intensity of microwave returning from a given surface depends on the geometric characteristics of the surface (i.e., shape and/orientation of objects, as well as their surface roughness) and its electrical characteristics (i.e., reflectivity and conductivity of materials, as measured by their complex dielectric constant). The all-weather capacity of microwave remote sensing combined with the fine spatial resolution of the data opens promising avenues for land cover assessment and monitoring. This technology has however not yet reached the same operational level as optical remote sensing.

2.2 In The Spatial Domain

A major attribute of a landscape is its spatial pattern—i.e., the arrangement in space of its different elements. The concept of landscape spatial pattern covers, for example, the patch size distribution of residual forests, the location of agricultural plots, the shapes of fields, or the number and configuration of landscape elements (i.e., their spatial heterogeneity). Landscape spatial pattern is seldom static due both to natural changes in vegetation and human intervention. Remote sensing offers the possibility to analyze changes in spatial pattern at the scale of landscapes. Textural information of fine spatial resolution remote sensing images can be measured using a variety of indices such as variance, variograms, mathematical morphology, or fractals.

3. Processing And Analysis Of Remotely Sensed Data

3.1 Preprocessing

Technical issues of geometric correction and rectification, radiometric correction, geometric and radiometric enhancement, and multispectral transformations of image data are described in Richards (1993). The purpose of these operations is to georeference and to transform the image data in a form that can be readily interpreted in terms of Earth surface attributes. The analysis and interpretation of remote sensing data typically is performed through image classification, to generate land-cover maps, or through change detection, to monitor land-cover changes.

3.2 Image Classification

Classification is a method by which labels are attached to spatial sampling units (i.e., the pixels) in view of their character. This character is generally their spectral response in different spectral ranges. It may also include their spatial attributes (i.e., texture) or temporal changes. This labeling is implemented through pattern recognition procedures, the patterns being pixel vectors. The most commonly used classification methodology is based on the maximum likelihood, a probabilistic classification method (Richards 1993). More advanced classification techniques have been developed. For example, contextual classifiers label a pixel in the context of its neighbors in space. They exploit spatial information of an image. Neural networks are classifier networks, having a decision tree structure, where a collection of simple classifiers is brought together to solve a complex problem.

3.3 Change Detection

Time series of remote sensing data can be created to monitor changes in landscapes at local to regional scales. Classic land-cover change detection techniques are based on the comparison of sequential land cover maps derived from remote sensing data or other sources for the same area. For every sampling unit of the maps, the land cover categories at the two dates are compared. The comparison of successive maps fails to detect subtle changes within broad land-cover classes. Thus, rather than detecting changes on the basis of land-cover categories, change detection is better performed on the basis of the continuous variables defining these categories, whether these are reflectance values measured by a satellite sensor or biophysical attributes derived by model inversion. Coppin and Bauer (1996) review techniques used for this comparison. Recent studies have detected changes in land-cover between successive dates by combining different methods (e.g., Macleod and Congalton 1998).

4. Components Of A GIS

A GIS is a powerful tool for handling spatial data. It is designed for the collection, storage, and analysis of objects and phenomena where geographic location is an important characteristic (Aronoff 1989, Maguire et al. 1991). The data input component of a GIS converts data from their existing form to one that can be used by the GIS. Two types of data are entered in a GIS: spatial data such as points, lines and areas that represent the geographic locations of features, and associated nonspatial attribute data that provide descriptive information. The data management component includes those functions needed to store and retrieve data from the database. In a GIS, data are maintained in a digital format. The representation of the spatial component of geographic information follows either the raster model—the homogeneous units are regular cells—or the vector model—the spatial units are the points, lines, and polygons. The data analysis and manipulation functions determine the information that can be generated by the GIS. A GIS automates certain activities and data interpretation procedures. Some of the most commonly used GIS functions are classification, overlay operations, neighborhood operations, spatial interpolations, geo-statistical analyses, and connectivity functions. The output or reporting functions of a GIS produce maps, tables, or statistical results.

GIS facilitates the interpretation of remote sensing data, by linking biophysical information measured by remote sensing to maps of natural and cultural landscape attributes (e.g., soil type, topography, accessibility as measured by various distances), to field measurements (e.g., crop yield or water quality) and to socioeconomic information derived through household surveys (e.g., indicators of well-being, rate of population growth, agricultural input use). For a discussion of the ideas and social practices that have emerged with the development of new forms of data handling and spatial representation in GIS, see Pickles (1995).

5. The Contribution Of Remote Sensing And GIS To Social Science Studies

The scientific literature is becoming increasingly rich in case studies using remote sensing to study social phenomena. Actually, remote sensing techniques are growing in importance for studies aimed at understanding land use dynamics, its driving forces and its impact on society (see Behrens 1994, Liverman et al. 1998). The output of remote sensing and GIS analyses are major inputs for spatial models. Remote sensing data is supporting land-use change analysis in a variety of ways.

5.1 Land-Cover Mapping

Remote sensing is used widely to map land cover, land use, and agricultural patterns. The current classification algorithms succeed in identifying from 10–15 land-cover classes, including several succession stages of vegetation. Achieving this level of detail requires however a set of georeferenced field observations on a sample of locations. Current sensors do not identify very well settlements or different cropping systems. Spatially heterogeneous or fragmented classes are identified as mosaics of cover types. The further analysis of land use land cover maps can lead to the separation of ‘landscape units,’ characterized by different combinations of land use and biophysical attributes.

5.2 Monitoring Land-Cover Change

Remote sensing facilitates the introduction of a dynamic perspective in land use studies, by monitoring changes in the spatial patterns of landscapes caused by land use. Sequential remotely sensed data allows to detect changes in land use, such as an expansion of the area under permanent or shifting agriculture, urban growth, or increase in eroded areas. For example, quantitative estimates of rates of deforestation are derived from time series of remote sensing data. The spatial extent and location of secondary growths on abandoned plots is also well detected by remote sensing. Remote sensing has been used in semi-arid regions to assess dryland degradation and to identify transformations of land use and cultivation. Measuring rates of dryland degradation is a complex challenge, however, given the strong interaction between erratic fluctuations in rainfall due to climatic variability and anthropogenic changes in vegetation cover.

5.3 Identification Of Land-Cover Change ‘Hot Spots’

Land-cover change ‘hot spots’ are defined as areas where high rates of land-cover changes have been observed. Remote sensing-based methods for identifying land-cover change hot spots at broad spatial scales have been developed. Such methods are based on a series of indicators, such as land-cover fragmentation, occurrence of fires, opening of new roads, etc., which are detectable, by remote sensing. Deforestation, for example, is strongly related to the proximity to forested deforested edge, the proximity to an access and the degree of fragmentation of forests. Forest fragmentation can be quantified easily from remote sensing data.

5.4 Identification Of Land-Use Attributes

Remote sensing allows identification and measurement of key socioeconomic and ecological characteristics of land use systems. In the Guinea Highlands, Gilruth and Hutchinson (1990) have been able to discriminate, on the basis of remote sensing techniques, between permanent agriculture (in the form of home gardens) and shifting cultivation. The spatial distribution of shifting cultivation areas can also be mapped from fine to medium resolution satellite data. Guyer and Lambin (1993) succeeded in discriminating and quantifying the total area and proportion of tractor-cleared and hand-cleared fields in a region of Nigeria by using several shape criteria derived from multispectral SPOT data. These authors also computed the crop-fallow cycle and the importance of the land reserve using a remote sensing-based land-cover map.

5.5 Spatial Stratification And Spatial Extrapolation

Remote sensing allows testing of how widely local observations apply to larger areas. Actually, a common problem of in-depth investigations of land-use dynamics is that it requires a large data collection effort in the field, at the household and plot levels. Remote sensing, through the measurement of indicators of land use, helps to define regions characterized by a similar land use, that can lead to a spatial stratification of a study area in terms of land use patterns. As a result, findings obtained over specific sites within a stratum, that is homogenous in terms of land use, can be extrapolated with more confidence to that stratum. If such a spatial stratification is performed before field surveys, it is a powerful tool to design a sampling strategy and locate villages or households that will be used for in depth data collection and investigation. This leads to an optimization of the field data collection effort.

In a study of land use intensification (Guyer and Lambin 1993), remote sensing analysis provided a means of matching the patterns from a haphazard sample of households to the population, based on a few variables that were estimated both by household survey (on the sample) and by remote sensing (on the entire population). The results tended to give greater confidence to inferences from the sample about ecological dynamics.

5.6 Input Into Spatial Models

The spatial patterns, which are detectable at a variety of spatial scales on remotely-sensed data, provide the most obvious link between spatially-explicit land-use change models and remote sensing analysis. With the application of remote sensing technology to the monitoring of land-use changes, it becomes possible to integrate more explicitly spatial relationships into models of land-use change. It is also possible to apply and test these models at a finer spatial resolution and with samples comprising smaller geographic units— e.g., at the scale of individual landscape units.

Spatial, statistical models are born from the combination of remote sensing, GIS, and multivariate mathematical models. Their emphasis is on the spatial distribution of landscape elements and on changes in landscape patterns. The goal of these models is the projection and display, in a cartographic form, of future landscape patterns, which would result from the continuation of current land-management practices or the lack thereof. The approach analyses the location of land-cover changes in relation to maps of natural and cultural landscape variables in a GIS. A model is built to describe the relationship between the dependent variable—e.g., land-cover changes—and the independent landscape variables. Spatial, statistical analyses of tropical deforestation or dryland degradation have already been developed for several tropical regions (see review by Lambin 1997). Spatial models of desertification hazard have also been developed by integrating indicators of water and wind erosion, vegetation degradation, range utilization, and human settlement.

5.7 Validation Of Spatially-Explicit Models

Dynamic spatial simulation, actor-based models have been developed to predict spatial patterns of land-use changes at the landscape scale—e.g., for tropical deforestation in roadside areas occupied by shifting cultivators (e.g., Wilkie and Finn 1988) or in regions of official resettlement schemes associated with road development (Southworth et al. 1991). These grid-cell models combine spatially-explicit ecological information with socio-economic factors related to land use decisions by farmers. The modeling is based on rules of behavior determined by locational attributes, land tenure system, and vegetation successions. Remote sensing offers a powerful tool to calibrate and/or validate these models by comparing actual and predicted land-use cover patterns, and by comparing the level of fragmentation of the landscape.

6. ‘Socializing The Pixel’

By ‘socializing the pixel’ one means to discern information embedded within spatial imagery that is directly relevant to the core themes of the social sciences, and use it to inform the concepts and theories pertinent to those themes (Geoghean et al. 1998). Creating a direct link between spatially-explicit land cover information, as derived by remote sensing, and information on land-use change processes requires the development of new methods and models which are merging landscape data with data on human behavior.

6.1 Linking Household-Le El Data With Remote Sensing Data

A synthesis of the methods for merging socio-economic household survey data and remote sensing-based information is given in Liverman et al. (1998). The major challenges of this approach are: (a) the definition of the appropriate spatial observation units, i.e., the appropriate level of aggregation of information derived from the domains of social phenomena and natural environment, and (b) the development of the appropriate linkages between household-level and remote sensing datasets. While conceptually straightforward, these links can be difficult to implement operationally. In remote sensing, the unit of observation is the pixel that is not directly associated to any social science unit of observation, e.g., individuals, households, or villages. Establishing the correspondence between biophysical and socioeconomic variables requires taking into account the unit of decision-making, units of landscape transformation, and the spatial scale of ecological processes (e.g., a watershed).

Linking remote sensing observations to socio-economic data at the scale of the administrative units which were used for the collection of socioeconomic data leads to a loss of information as it obscures the variability within the units. More recently, research efforts have attempted to integrate remote-sensing observations and field surveys at finer levels of aggregation, i.e., at the scale of individuals, households, or villages (e.g., Entwisle et al. 1998). There are difficulties in relating remotely-sensed patterns of land-cover changes with field observations of land-use changes since, in many cases, people live in nucleated villages away from their fields, since households may cultivate multiple non-contiguous plots and since people move—especially pastoralists. For this reason, the integration of the two datasets is generally performed at the village level.

Individual household data allow for a better understanding of the land-use practices within each village, as most land-use decisions are made by individuals and households. Moran and Brondizio (1998) have investigated the linkages between remotely sensed data and traditional field methods in the social and biological sciences. Their results have led them to new research questions, including examination of the role of the developmental cycle of the household in shaping their trajectory of land-use and deforestation.

6.2 Landscape Spatial Patterns

The spatial pattern of a changing landscape have some information content on the processes of land-cover change. Certain categories of land-use changes tend to fragment the landscape (e.g., expansion of smallholder farming, small-scale logging, overgrazing around deep wells). Other land-use changes increase landscape homogeneity (e.g., mechanized cultivation or ranching over large areas). There is a good correlation between remotely-sensed spatial patterns and some important characteristics of farming systems. For example, Lambin (1988) found statistically that, in Burkina Faso, spatial patterns interpreted from Landsat MSS data correlate more closely with ethnic groups than with any other physical or cultural landscape variables. Other authors demonstrated that the analysis of the spatial pattern of forest-nonforest interfaces allows identification of regions affected by different deforestation processes, which are controlled by different driving forces.

The spatial patterns of fires, which are well detected from space, are also good indicators of human activities. In tropical forests, regions of fire concentration indicate frontier areas and variations in the spatial organization of fires are associated strongly with political boundaries and farming practices.

6.3 GIS For Health And The Environment

An emerging area of research concerns the application of remote sensing and GIS techniques to analyze the relationships between people, their environment and their health (de Savigny and Wijeyaratne 1995). GIS provides information about the spatial distribution of and interaction between disease risk factors, patterns of morbidity and mortality, and the allocation of health resources. Vector-borne diseases (e.g., malaria, trypanosomiasis) have received most emphasis so far, as their occurrence is strongly controlled by environmental attributes which are detectable by remote sensing (Barinaga 1993). However, there is a large potential for GIS application on any environment related health problem. A classic example is the spatial correlation between cholera deaths and contaminated water supplies.

Bibliography:

1. Aronoff S 1989 Geographic Information Systems: A Management Perspective. WDL Publications, Ottawa, Canada
2. Barinaga M 1993 Satellite data rocket disease control efforts into orbit. Nature 261: 31–32
3. Behrens C A 1994 Recent advances in the regional analysis of indigenous land use and tropical deforestation: Introduction. Human Ecology 22: 243–47
4. Colwell R N, Simonet D S, Estes J E 1983 Manual of Remote Sensing. American Society of Photogrammetry, Falls Church, VI
5. Coppin P R, Bauer M E 1996 Digital change-detection in forest ecosystems with remote sensing imagery. Remote Sensing Review 13: 207–34
6. de Savigny D, Wijeyaratne P 1995 GIS for Health and the Environment. International Development Research Center, Ottawa, ON
7. Entwisle B, Walsh S J, Rindfuss R R, Chamratrithirong A 1998 Land-use land-cover and population dynamics, Nang Rong, Thailand. In: Liverman D, Moran E F, Rindfuss R R, Stern P C (eds.) People and Pixels: Linking Remote Sensing and Social Science. National Acadamy Press, Washington, DC
8. Geoghegan J, Pritchard L, Ogneva-Himmelberger Y, Chowdhury R R, Sanderson S, Turner II B L 1998 ‘Socializing the pixel’ and ‘pixelizing the social’ in land-use and landcover change. In: Liverman D, Moran E F, Rindfuss R R, Stern P C (eds.) People and Pixels: Linking Remote Sensing and Social Science. National Acadamy Press, Washington, DC
9. Gilruth P T, Hutchinson C F 1990 Assessing deforestation in the Guinea highlands of West Africa using remote sensing. Photogram Engineering and Remote Sensing 56: 1375–82
10. Guyer J, Lambin E F 1993 Land use in an urban hinterland: Ethnography and remote sensing in the study of African intensification. American Anthropologist 95: 839–59
11. Lambin E F 1988 L’apport de la teledetection dans l’etude des systemes agraires d’Afrique; l’exemple du Burkina Faso. Africa 58: 337–52
12. Lambin E F 1997 Modelling and monitoring land-cover change processes in tropical regions. Progress Physical Geography 21: 375–93
13. Lillesand T M, Kiefer R W 1994 Remote Sensing and Image Interpretation. Wiley, New York
14. Liverman D, Moran E F, Rindfuss R R, Stern P C (eds.) 1998 People and Pixels: Linking Remote Sensing and Social Science. National Academy Press, Washington, DC
15. Macleod R D, Congalton R G 1998 A quantitative comparison of change-detection algorithms for monitoring eelgrass from remotely sensed data. Photogram Engineering and Remote Sensing 64: 207–16
16. Maguire D J, Goodchild M F, Rhind D W 1991 Geographical Information Systems: Principles and Applications. Longman, Harlow, UK
17. Moran E F, Brondizio E 1998 Land-use change after deforestation in Amazonia. In: Liverman D, Moran E F, Rindfuss R R, Stern P C (eds.) People and Pixels: Linking Remote Sensing and Social Science. National Acadamy Press, Washington, DC
18. Pickles J 1995 Ground Truth: The Social Implications of Geographic Information Systems. The Guildford Press, New York
19. Richards J A 1993 Remote Sensing Digital Image Analysis. Springer-Verlag, Berlin
20. Southworth F, Dale V H, O’Neill R V 1991 Contrasting patterns of land use in Rondonia, Brazil: simulating the effects on carbon release. International Social Science Journal 130: 681–98
21. Wilkie D S, Finn J T 1988 A spatial model of land use and forest regeneration in the Ituri forest of Northeastern Zaire. Ecological Modelling 41: 307–23