A protected natural area serves multiple functions such as providing a refuge for biodiversity, absorbing air and water pollution, preventing regional soil erosion, protecting
microclimate, providing recreational space, providing carbon sinks to reduce greenhouse gases, and supporting a variety of educational and scientific research goals. The field of landscape ecology has established that the degree to which a protected natural area supports landscape functions, such as these, depends both on the composition and the spatial pattern of its landcover patches and background matrix (Forman & Godron 1986). Walsh et al. (2003, p. 125) writes that “pattern metrics offer an approach for characterizing landscape states and conditions across space and time as an indication of landscape form and function.”
Landscape pattern metrics have been used in conjunction with remotely sensed imagery to characterize landscapes in a variety of contexts, including to compare the level of
similarity/differences in landscape change between two nearby settlements experiencing different community planning (Batistella et al. 2003); compare the similarity between multiple simulated landscapes (Messina & Walsh 2001); understand landscape spatial composition as a function of geographic, biophysical, and socio-economic variables on a household/farm level in the Ecuadorian Amazon (Pan et. al. 2004); statistically identify clusters of slopes within an alpine treeline ecotone on the basis of spatial and compositional patterns (Allen & Walsh 1996); and investigate the effects of landscape patterns on forest crown fire dynamics and the resulting ecological impact (Turner & Romme 1994). In this section, changes in spatial pattern of landcover within GNP are analyzed in relation to the function of the landscape in terms of nature protection, including the protection of specific species, and the evolution of the Latvian cultural landscape in the Park between during the study period (1985 – 2002). These landscape functions were chosen for analysis, because they represent the interests and values identified by Park stakeholders, as presented in
Chapter 2. The timing of the changes in landcover patterns is evaluated with respect to economic, political, and social influences in the Park.
The spatial pattern of the GNP landscape was characterized at each of the four satellite image dates by calculating pattern metrics using algorithms contained within the FRAGSTATS software (McGarigal & Marks 1995). Some metrics were calculated at the
landscape “level”. This refers to a type of measure that summarizes patterns of all landcover classes across the full landscape of interest. Some metrics were calculated at the “class” level. This refers to a type of measure that summarizes spatial patterns of just one landcover class throughout the full landscape of interest. The third and final type of measure that FRAGSTATS calculates was not used for these analyses. This type of measure is termed a “patch” level metric. These metrics calculate spatial pattern characteristics of just one landcover patch.
Due to the difficulty in interpretability of some landscape pattern metrics, a small number of metrics specifically applicable or designed for unique research questions are often used. In a study where Messina et al. (2006) compared landscape pattern metrics between simulated and sample landscapes over time in a protected area in the Ecuadorian Amazon basin, the following metrics were used: number of patches, patch area, mean patch size, and patch density. In the Pan et al. (2004) study investigating landscape composition as a
function of household/farm level predictor variables, the patch density, landscape shape, and contagion metrics were calculated for each individual farm as inputs into regression models. Batistella et al. (2003) analyzed six pattern metrics, focusing on structural changes in the landscape, instead of functional landscape changes, to compare landscape changes in two settlements experiencing different community planning scenarios. Batistella et al. (2003)
applied and interpreted the following metrics: percentage of forest in landscape, the largest patch of forest (for inferences regarding biodiversity), mean patch size, patch size standard deviation, area-weighted mean shape index (to study edge/center relationships of patches), and mean core area index (to look at the average “core area”, or area 90+ meters from the patch edge). In the current study, pattern metrics were calculated for each satellite image date. Similar metrics to those used by Batistella et al. (2003) were specifically chosen to describe protection of core forest areas as well as “patchiness” of the landscape, important for monitoring the evolution of the Latvian cultural landscape and the nature of its
fragmentation.
Table 3.3. Landscape pattern metrics analyzed.
Metric name Level Definition Meaning
Number of patches Landscape Number of patches in the landscape. Total number of patches in GNP. Mean patch area Landscape Total area (in m2) of landscape
divided by the number of patches; the result is divided by 10,000 to get the units in hectares.
Average size in hectares of all landcover patches in GNP.
Area of Largest Patch Landscape Area (in m2) of largest patch divided
by 10,000 to get units in hectares. Size in hectares of largest patch in GNP. Total Core Area Class
(forest) The sum of all core (forest) areas (in m2). Each core area is the area of each (forest) patch greater than a user-specified distance (250 m) from the patch edge.
The total forest area in the Park that is at least 250 meters from its patch edge.
Results
The results are first summarized with maps of the four classified satellite images. Although these maps (see Figures 3.1 through 3.4, below) show coniferous forest and hardwood forest to give the reader a better sense of the Park’s distribution of vegetation, the two forest classes were combined into one forest class for the analyses, since changes from
and to forest are sufficient to characterize landscape changes associated with the evolution of the natural and cultural landscapes. Changes in amounts of gross landcover between image dates are then summarized and discussed in relation to the timing of economic, social, and policy events affecting GNP. Following, from-to landcover change matrices are presented, describing the number of pixels that changed from each landcover class to each landcover class between consecutive satellite images. These changes are then interpreted with respect to the timing of changing economic, social, and policy conditions. Finally, the changes in pattern metrics between image dates are summarized and discussed in terms of likely causes, as well as effects on the protection of nature/biodiversity and the Latvian cultural landscape.