CHAPTER 5 - LANDSCAPE CHANGE DESCRIBED BY SELECTED METRICS
'Quando eu percebi, isso aqui tava tudo mudado'
5.1. Why study landscape change in Rondônia?
The interest of naturalists and ecologists in landscape spatial patterns is extensive (Urban et al. 1987, Turner et al. Predicting 1989). This approach, responsible for an ecological perspective about the geographic space, is today represented by landscape ecology. The term was initially proposed by Troll (1939) and used by Schmithusen (1942) and Neef (1956), among others. The tradition in regional geography and vegetation ecology studies was in the origin of this recent science (Bertrand 1968, Godron et al. 1968, Long 1974, Jurdant et al. 1977). Its historical development was widely reviewed by Naveh (1982) and Naveh and Lieberman (1984).
The concept of landscape was always present in the history of civilization, induced by artistic motivation, as a complementary descriptor on the delimitation of territories. Recently, landscapes became objects of study, analysis, and synthesis, including new perspectives about the distribution of ecological systems. The landscape is no longer considered just 'a portion of the earth surface captured by human eyes' (Amandier 1973). It is now understood as a spatially heterogeneous mosaic (Forman and Godron 1981) to be studied from the reciprocal effects among spatial patterns and ecological processes (Pickett and Cadenasso 1995). Others have emphasized the human dimension underlying landscape outcomes (Naveh and Lieberman 1994). The study of these relations confers a practical dimension to landscape ecology, through the establishment of scientific bases for planning, management, conservation, and development of territories (Leser and Rodd 1991).
Since the expression of such studies is spatially represented, the issue of scale and resolution is central (Allen and Starr 1982, Meentemeyer and Box 1987, Pickett and Candenasso 1995). Recent empirical tests have focused on the role of scale and resolution for understanding relations among patterns and processes of landscape change. Changing spatial resolutions, for instance, may affect our ability to extrapolate information across different scales (Turner and Gardner 1991). Traditionally, many researchers have assumed that ecological processes affecting populations and communities operate at local scales (Dunning et al. 1992). Meanwhile, habitat variations respond to different scales (Wiens 1989), making the problem of spatial dynamics one of the frontiers of ecology (Levin 1992, Kareiva 1994). The current interest in biodiversity, within the landscape context, unites the research on population dynamics and ecological processes (Ricklefs 1987, Norton and Ulanowicz 1992, M. Turner et al. 1995). Perhaps, an important methodological problem for landscape ecological studies may be the difficulty of repeating observations through time and space. For this reason, quantitative approaches, through models of analysis and simulation, still dominate (Sklar and Costanza 1991).
In landscape ecology, the need for studies at multiple scales suggests the use of spatial data analysis (Turner et al. Predicting 1989). This has been done through modern approaches to address spatial patterns and ecological processes (Turner et al. Effects 1989, Turner 1990, Flamm and Turner 1994, Wickham and Norton 1994). The parallel development of geographic information science (Goodchild 1992) and landscape ecology (Forman and Godron 1986) provides new opportunities for multi-disciplinary studies on ecological modeling of spatial data (Raper and Livingstone 1995). However, the nature of spatial data is diverse and such applications must take the actual nature of the ecological phenomena into account instead of just testing algorithms. Like the 'illusion of objectivity' inherent to analyses of statistical data (Berger and Berry 1988), the analysis of spatial data also includes a variety of pitfalls. However, the need for quantitative methods is an incentive to search for standards. In a world of constant change from global to local scales, it is urgent to overcome the limitations of spatial representations and find better ways to handle their intrinsic problems.
One of the primary steps of spatial analytical initiatives is to identify their underlying assumptions. Frequently, the assumptions are so strong that even the choice of methodological techniques to be used is affected. For example, Anselin (1989) emphasized that the uniqueness of spatial data is expressed by three characteristics. First, it is primarily based on two continuous dimensions (x,y). Second, it presents spatial dependence: 'the propensity for nearby locations to influence each other and to possess similar attributes'. Third, geographical data is distributed over the curved surface of the Earth (from projections to the sphere). The field of geostatistics has followed the assumptions of continuity and spatial dependence (Rossi et al. 1992). It is reasonable to expect such characteristics when dealing with spatial data, until there is a boundary. As human-altered landscapes are full of sharp boundaries (Forman 1997), difficulties have been faced to integrate geostatistics and landscape ecology.
Another relevant issue when dealing with spatial data is that spatial representation can assume multiple forms. Areal data, point data, network data and directional data are the most common ones (Burt and Barber 1996). The purpose of these spatial representations is to mimic a range of phenomena. Thus, examples include land-use/land-cover maps as areal data, vegetation samples as point data, drainage systems as network data, and wind or water flow as directional data.
Several representation techniques have been tested through statistical approaches to allow integration of distinct spatial distributions. Although there are methods to convert data from different spatial representations (e.g., point data into areal data and vice-versa), the procedure is not always recommended. Recently, development efforts are willing to integrate this distinct group of techniques in a more friendly way to handle spatial data (Goodchild et al. 1992, Burrough and Frank 1995). Geostatistics techniques (Issaks and Srivastava 1989), spatial analysis (Burrough 1990, Baker and Cai 1992, Fotheringham and Rogerson 1995, McGarigal and Marks 1995), and GIS capabilities (Burrough and McDonnell 1998, DeMers 2000) have provided new opportunities to explore spatial- and scale-related matters (Withers and Meenteneyer 1999).
The potential of such an integrative approach to handle spatial phenomena for the study of landscapes is promising. However, this functionality is still not implemented in a friendly way that allows a reasonable manipulation of different spatial data representations through complementary techniques. Also, although the integration has been frequently suggested, it is rare to see studies in landscape ecology dealing with point data, for example. This is probably related to the rationale behind the study of landscape structure, based on the concepts of matrix, patch, and corridor (Forman and Godron 1986, Forman 1997). Landscape mosaics imply discreteness of elements and the existence of clear boundaries between neighboring patches (Hansson et al. 1995). Thus, spatial statistics has been used to describe the degree of spatial autocorrelation or spatial dependency between values of a variable that has been sampled at various geographic coordinates, while landscape metrics characterize the geometric and spatial properties of a mosaic of patches (Fortin 1999).
The applicability of these concepts to spatially explicit ecological studies is clear. In a world where human-altered landscapes are increasingly created, processes of disturbance need to be spatially quantified and understood. Disturbance can be defined as any relatively discrete event in time that disrupts the ecosystem, community, or population structure and changes resources, substrate availability, or the physical environment (Pickett and White, 1985). The propagation of disturbance in heterogeneous landscapes depends on the structure of the landscape, as well as on the intensity and frequency of disturbances. M. Turner et al. (1995) highlighted the importance of new conceptual approaches when studying disturbance within landscapes. For instance, a broader view of the equilibrium concept should expect a return to normal dynamics rather than to an artificial 'undisturbed' state. Moreover, as disturbed sites recover deterministically through succession, stability must be assessed through multi-temporal and -spatial approaches, taking into account the scale-dependent nature of concepts of landscape equilibrium (M. Turner et al. 1993).
When studied through the landscape ecology approach, the structure of 'disturbance landscapes' is controlled by characteristics of the disturbance regimes, including the distribution of disturbance sizes and intervals, and the rotation time. In this case, the structure of mosaics of disturbance patches (e.g., patch size and shape) is an important parameter to assess landscape structure (Forman and Godron 1981). Both the number and size of patch births (i.e., patch turnover) govern the response of landscapes to changing disturbance regimes (Baker 1995).
Several methods based on the concept of landscape structure have been developed to address processes of disturbance within landscapes. Landscape metrics have been widely used for this purpose (Baker and Cai 1992). The integration of spatial data in GIS has improved this approach (Haines-Young et al. 1996, Forman 1997, Frohn 1998). Using GIS and landscape metrics to relate disturbance and spatial heterogeneity allows the study of environmental composition and configuration at scales broader than the community or ecosystem (Sample 1994). There are metrics related to landscape composition, referring to features associated with the presence and amount of each patch type within the landscape but without being spatially explicit. Others are related to landscape configuration, referring to the physical distribution or spatial character of patches within the landscape (Burrough 1981, Mandelbrot 1983, McGarigal and Marks 1995).
Perhaps, one of the most frequent examples of landscape disturbance in the tropics is derived from LULC change, particularly forest fragmentation. The process occurs when forested areas are progressively subdivided into smaller and more isolated forest fragments, mainly as a result of human land-use activities. Landscape heterogeneity can either increase or decrease, depending on the parameter and spatial scale examined (Krummel et al. 1987). In general, the disturbed landscape has more small forest patches and fewer large, matrix patches than the intact landscape (Lovejoy et al. 1986, Mladenoff et al. 1993, Malcolm 1994, Lovejoy 1997).