Home Articles Integration of RS and GIS to assess human impact on ecosystem change...

Integration of RS and GIS to assess human impact on ecosystem change in Lianos area (Venezuela)

1Yanning Guan (China), 2Steven M. de Jong, 3Johan de Meijere (Netherlands)
1Institute of Remote Sensing Applications, Chinese Academy of Sciences,
P. Box 9718, Beijing 100101, China
Tel: 86-10-64889540
E-mail: [email protected]
2Center of Geoimformation, Wageningen University and Research Center
3International Institute for Aerospace Survey and Earth Sciences

Abstract:
Landscape ecology emphasizes large areas and ecological effects of the spatial patterning of ecosystem. Recent developments in landscape ecology have emphasized the important relationship between spatial patterns and many ecological processes. Quantitative methods in landscape ecology link spatial patterns and ecological processes at broad spatial and temporal scales. In turn the increased attention on temporal change of ecosystem and human impact has highlighted the need for quantitative methods that can analyze patterns, determine the importance of spatially explicit processes, and develop reliable model. This research applies quantitative methods to assess the human impact on ecosystem temporal change in the flooding savanna area. Remote sensing offers the temporal change of ecosystem on landscape characteristics and disturbed by human. GIS as an important tool use in ecosystem temporal change analysis, landscape fragmentation analysis, and a probability model was developed to assess the relationship between ecosystem change and anthropogenic variables. The results are calculated in different scales.

1. Introduction
Landscape processes are dynamic and various climatic, lithological, landscape and anthropological factors contribute to significant spatial and temporal variability in environmental phenomena. Recognition, interpretation and mapping of the variability of process controlling variables and land conditions are imperative for assessing impact on ecological environment of human being and implementation of effective natural resources management.

Remote sensing image analyses systems and Geographic Information System (GIS) show great functionality for the integration of a wide variety of spatial information supporting tasks such as natural resource management, regional planning, and environmental monitoring. Current remote sensing programs are based on a variety of sensors that provide temporal and spatial earth observation on a global scale, thereby offering the opportunity for analyses of various phenomena synoptically from local to global scales. These attributes make remote sensing appealing for application to landscape ecology. GIS offers efficient tools for handling, manipulating, analyzing and presenting spatial data. Working in an integrated remote sensing and GIS environment allows taking advantage of both GIS and remote sensing image analysis techniques.

Remote sensing and GIS are powerful tools for the integration of a wide variety of spatial information applying tasks such as natural resource management and environmental change monitoring. Remote sensing technology and GIS offer the ability to facilitate ecosystem change investigations leading to a more complete understanding of human impact on the ecosystem. The research gives the approaches to derived the ecosystem temporal change and landscape fragmentation analysis by integrated remote sensing and GIS, and to model the human impact on the ecosystem change.

2. The Study Area
The flooding savannas of the Llanos area, which cover around 16,000 km2, are in the west central part of Venezuela, and occupy portions of Guarico, Cojedes, Portuguesa, Barinas and Apure States. The study area is in the Apure and locates in the central of Llanos.

In 1967, there was a serious flooding in Apure, where about 50,000 km2 were inundated. The government carried out a project named “Modulos of Apure” to control flooding and improve the hydrologic management of the region. There are 188, 000 ha of modulated area constructed. The advantages of the project were evident when productivity indexes of the modulated savannas under rational management were compared against the indexes of the natural savanna under traditional management. A series of dikes have been built for experiment. The dikes are more or less perpendicular to the riverbanks, and parallel to each other. The area closed by these dikes and riverbanks is called a ‘module’. A module can be defined as: “a sector of savanna that has been provided with low earth dikes to retain rainfall water in order to maintain green pastures throughout the year. These dikes have also the additional use as communication ways (roads)” (Gisele, 1998).

The construction of dikes has had an enormous effect on the economy of the Apure area. It not only impacts the local economy situation, but also affects the ecosystem in numerous ways. Change in hydrological factors, such as increased or decreased spate frequency and magnitude, change in frequency of drying events or shortening of drying, can affect the ecosystems extremely. The faunas characterised by extraordinarily rapid development. Similarly, biotic communities of the module area are disturbed. The stability of the habitats in these disturbed ecosystems may, in some case, provide a refuge for many species on other hand the construction of dikes led to an increase in influences on ecosystem by human being.

3. Research Methodology
The framework for remote sensing application in landscape ecology must be considered in reference to both the spatial and temporal condition. The individual research question associated with space, time and dynamics. (Quattrochi, D.A., and Pelletier, R.E., 1990) An appropriate data collection design can be employed within the scope of an investigation to provide answers to the specific landscape questions under consideration. For to assess the human impact on the ecosystem change these questions are not all-encompassing and should be addressed. The landscape processes associated with the landscape spatial and temporal dynamics should be observed and measured.

Figure 1 Methods of implemented integrated GIS and RS assess to human impact in the Llanos area
To take into account the landscape ecological characteristics in the study area and the research question and the limitation of available data, the framework of implemented methods in this research is:

  • Constructing the spatial ancillary database including road map and residential area map etc.
  • Interpreting the ecosystem map from Landsat TM.
  • Employing the change detection technique to derive the ecosystem temporal change 1960 to 1988.
  • Extracting the main human impact factor ¾ dikes by using knowledge based and fuzzy classification.
  • Carrying out the landscape fragmentation analysis in different scales ¾ to calculate the landscape pattern index that correlated with the degree of human manipulation of landscape in GIS environment to compare the spatial difference of landscape change in modular area and non-modular area.
  • Applying GIS to model the human impact on ecosystem change to take into account the anthropogenic factors such as distance to dikes etc.

Figure 1 gives an overview about implemented methods of integrated GIS and remote sensing to assess human impact in the study area.

4. Results
Temporal changes in a landscape could include change in (1) patch number, (2) patch size, (3) number and type of corridors, (4) number and type of dispersal barriers, (5) probability and spread of disturbance. (Turner, 1988) The emphases in this study are 1,2,4 and 5.

The change detection is a technique that detects the changes between two data sets, the Ecosystem map 1960 and the Ecosystem map 1988 pixel by pixel. The implementing of the post-classification is a program in GIS environment (ARC/INFO). Based on the results of landscape change detection, the Ecosystem change map, the area of ecosystem temporal change can be identified.

The patch number and the total boundary length in 1988 increase enormously comparing with in 1960, the average patch size is shrinking. The construction of dikes is the dissection, dissect or subdivide of the landscapes into sections. Fragmentation predominates in the present phases of landscape change.

Analysis of the landscape type change, the River forest and the Hyper-seasonal savanna, which dominated the study area and relative stable comparing with the other types. However, it is remarkable that the area of Hyper-seasonal savanna inundated by water is more than the Semi-seasonal savanna, and the Semi-seasonal savanna is the lower part in the study area. The trend of the landscape change indicates the landscape transformation that is not a natural process.

Using GIS the overlay analysis function, the ecosystem map 1960 was intersected with the dike map. The patches that were dissected by dikes were reselected. The areas of each landscape type were counted as disturbed by the construction of dikes. For the landscape types within the study area, the frequencies of disturbance are more than 80%.

The landscape indices represent the quantify aspects of spatial pattern that can be correlated with the spatial process. One of the indices, the fractal dimension is shown to be correlated with the degree of human manipulation of the landscape. The fractal dimension is an index of the complexity of shapes on the landscape. If the landscape is composed of simple geometric shapes like squares and rectangles, the fractal dimension will be small, approaching 1.0. If the landscape contains many patches with complex and convoluted shapes, the fractal dimension will be large. The fractal dimensions of 20 coverage were calculated in various scales for the modulated area and non-modulated area (Table 1).

Table 1 The fractal dimension on multi-scales in the modulated area and non-modulated area

Scale Modulated area
Fractal dimension
Non-modulated area
Fractal dimension
1960 1988 1960 1988
Patch area >0. 5 ha 1.200 1.406 1.402
Patch area > 5 ha 1.196 1.390 1.086 1.386
Patch area > 10 ha 1.214 1.372 1.086 1.346
Patch area > 25 ha 1.216 1.244 1.126 1.368
Patch area > 50 ha 1.156 1.016 1.100 1.306

In the non-modulated area, the fractal dimensions increase from 1960 to 1988 in multi-scale. These differences of the landscape indices between the two periods illustrate the natural fragmentation processes. The patches were fragmented to small pieces and the shapes of the patch became more complex.

In the modulated area, the fractal dimensions increase from 1960 to 1988 on the patch scales of: patch area >0. 5 ha, patch area > 5 ha, patch area >10 ha and patch area >25 ha. The differences indicate the natural fragmentation processes on the four patch scales as well. In the contrast, the fractal dimension decreases from 1960 to 1988 on the patch scale large than 50 ha. The change shows an opposite direction of the natural fragmentation process. The shape patch became simple like square and rectangle. It is anthropogenic process that can be associated with the construction of dikes.

The landscape properties is highly scale dependent (Keitt,T.H. 1997). Ecological studies more often have a spatial component and include landscape scale parameters (Baker W. And Cai Y., 1992). The results show that on the scale of patch area > 50 ha. Landscape indices captures major features of pattern ¾ the effects of the construction of dikes. The landscape indices are sensitive to the calculation scale (O’Neill, R.V.; Hunsaker, C.T.; et al. 1996).

An ecosystem change assessing model is developed by using GIS and logistic regression analysis. The logistic regression model has been studied the Mt.Graham red squirrel habitat (Pereira, J.M.C., and Itami, R.M., 1991), crane habitat (Herr,A.M., and Queen, L.P., 1993) and bobwhite habitat (Roseberry,J.L.,et.al.1994). The logistic regression analysis can contain numeric as well as categorical data. The change of ecosystem is the dependent variable. The independent variables included the environmental variables (ecosystem units) and anthropogenic variables. (eg. Distance to dike) Using GIS function the shortest distances will be measured between the ecosystem change and anthropogenic features. The land use anthropogenic features (dikes etc.) are extracted from TM.

The logistic model and the probability of ecosystem change developed in this study is represented in the following:


Where Y is the exponent of the logistic equation;
P is the probability ecosystem change at a particular location.

The output probability values range from 0 to 1 indicating 0 to 100 percent probability of ecosystem change. The result can be illustrated by GIS in a digital map.

The descriptive statistics of the anthropogenic independent variables are the Distance to dike, Distance to residential area, Distance to main road, Distance to dirt road, Distance to path and Distance to airport. The ecosystem change map as the dependent variable is only having two codes, 0 is non-changed and 1 is changed. The dependent and the independent variables are exactly identical in location and pixel size.

Applying the logistic model, which the pixel size is 150m by 150m and the sample area is 30pixel by 30pixel, assessed the ecosystem change probability. A raster format was used for it is more effective at presenting spatially continuous phenomena than is a vector format. The result is a probability of the ecosystem changing effected by the anthropogenic factors (Figure 2).

5. Summary
In general agreement with the ground truths, the models suggested that the construction of the dikes have the unique directional influence on the ecosystem change. Close to the dikes, the probabilities of the ecosystem change increase. This may suggest the construction of the dikes is the human disturbances that are widely spread in the Llanos modular area, and roads, airport and the residential sites provided the important impacts for the ecosystem change as well. This study should contribute to understand the human impact on the flooding savanna in Llanos area.

Figure 2 The probability of the ecosystem changing effected by the anthropogenic factors
The structure and function of a landscape can be perceived differently at different scales, and it is important for the observer to decide upon appropriate scales for a study. To develop the GIS model, not only the sample scale effects the result of model; the cell size also influences the model establishing. This is the particular issue of modelling in the GIS environment.

References

  • Baker W. And Cai Y., 1992. The r.le programs for multiscale analysis of landscape structure using the GRASS geographical information system. Landscape Ecology, vol.7(4), p291-302.
  • Gisele, C. L., 1998. GIS support for SWOT analysis applied to land evaluation, Msc. thesis of Wageningen Agriculture University, the Netherlands.
  • Herr,A.M., and Queen, L. P., 1993. Crane nabitat evaluation using GIS and remote sensing, Photogrammetric Engineering & Remote Sensing, 59(10): 1531-1538.
  • Keitt, T. H. et. al., 1997. Detecting critical scales in fragmented landscapes,
  • Pereira, J.M.C., and Itami, R.M., 1991. GIS-based habitat modeling using logistic multiple regression: a study of the Mt. Graham red squirrel, Photogrammetric Engineering & Remote Sensing, 57(11) 1475-1486.
  • Quattrochi, D.A., and Pelletier, R. E., 1990. Remote sensing for analysis of landscapes: an introduction, Quantitative Methods in Landscape Ecology, ed. Turner, M. G., and Gardner, R.H. 53-76.
  • Roseberry, J. L., et.al.1994. Assessing the potential impact of conservation reserve program lands on bobwhite habitat using remote sensing, GIS, and habitat modeling, Photogrammetric Engineering & Remote Sensing, 60(9): 1139-1143.