2. Landscape Function Analysis and Ecological Results
2.1.1 Landscape Function Analysis and Ecological Considerations
263
Spatial resolution is an important aspect of HRS data but is an even more critical issue when 264
quantifying landscapes and ecological processes. This is due to the inherent heterogeneity in 265
processes and structures which often vary, not only spatially and temporally within the 266
same hierarchical level of organisation in a landscape, but may also vary across levels of 267
organisation in a landscape (Lovett et al., 2005). A landscape has been defined as “an area 268
that is spatially heterogeneous in at least one factor of interest” (Lovett et al., 2005). 269
Forman and Godron (1981) defined a landscape as a “kilometres-wide area where a cluster 270
of interacting stands or ecosystems is repeated in similar form.” However, in their scheme, 271
processes and structures of a few metres up to hundreds of metres are at a finer level of 272
scale than a landscape (Forman and Godron, 1981). These structures at a finer scale 273
correspond with the definition of patch given by Lovett et al. (2005) which is “an area that 274
differs from its surroundings in structure and function.” Forman and Godron (1981) were a 275
little more explicit, defining patches as “communities or species assemblages surrounded by 276
a matrix with a dissimilar community structure or composition.” Underlying these 277
definitions of patches and landscapes is scale which Lovett et al. (2005) defined as the 278
11 “spatial or temporal dimension of an object or process, characterized by both grain and 279
extent.” Ludwig et al. (2000) applied power laws to landscape structures and processes and 280
found that at different scales different power functions could be derived and that this 281
change in power function implied scale-dependent thresholds. These authors suggested 282
that at local scales, biological, geochemical and microtopographic redistribution processes 283
are responsible for patch heterogeneity, whereas at landscape scales, vegetation and soil 284
heterogeneity are driven by geological, hydrological and pedogenic processes (Ludwig et al., 285
2000). They further argue that scale is not just a concern in defining landscape structure and 286
processes but also influences our observation processes. The effect of scale in our 287
observations shapes and limits the landscape structures and processes that may be 288
observed, and therefore our awareness of the context within which these structures and 289
processes occur may be incomplete (Ludwig et al., 2000). This is particularly relevant when 290
the HRS level of observation with Hyperion imagery is from 705 km distant from the target 291
where 900 m2 of the Earth’s surface is resolved into a single observation. Whereas, the
292
ecological scale of field measurement as discussed below is collected from a maximum 293
distance from the Earth’s surface of two metres and ranges in area from point measures of a 294
few cm2 to transects covering 180 m.
295
Definitions of patches are at the heart of LFA as the underlying philosophy is that processes 296
active in a landscape structure how that landscape organises itself spatially which in turn 297
feeds back into the processes that are active in that landscape. Therefore by defining the 298
patch types and their organisation relative to each other, the processes active in a landscape 299
can be deduced and evaluated. Landscape Function Analysis (LFA) is a monitoring procedure 300
that assesses how effectively a hill slope is operating as a biophysical system based on the 301
use of visual and tactile assessment of indicators of landscape organisation and soil surface 302
characteristics (Tongway and Hindley, 2004). In brief, the LFA technique requires the 303
establishment of a transect along the major gradient (sometimes referred to as a gradsect) 304
and quantifying the patches and inter-patches along this transect (Tongway and Hindley, 305
2004). These patches are quantified in terms of type, length along transect, and width to a 306
maximum arbitrary distance of 5 m either side of the transect (Tongway and Hindley, 2004). 307
Thereafter, the eleven soil-surface assessment indicators (SSAI) are quantified (Figure 2.1) 308
for each of five replicates for each patch type. The practitioner uses sensory clues, mostly 309
12 visual, to assign values to each SSAI ranging from 1 – 4 to 1 – 10 depending on the particular 310
SSAI (Figure 2.1). These SSAIs are algorithmically resolved into three soil surface indices 311
(SSI), namely stability, infiltration and nutrient cycling, using specific SSAIs for each index 312
(Figure 2.1). From all the above measurements a number of metrics of landscape 313
314 315 316
Figure 2.1 The eleven Soil Surface Assessment Indices (SSAI) measured in the field using
317
the ranking scales shown. The SSAIs are then partitioned and resolved into 318
the three Soil Surface Indices (SSI) of stability, infiltration and nutrient cycling, 319
(from Tongway and Hindley, 2004). t = transported, l = local, n = nil, s = 320
slight, m = moderate, e = extensive, E = sheeting, P = pedicles, T = terracettes, 321
R = rills, S = scalds (Courtesy Tongway and Hindley, 2004). 322
13 organization and process can be derived. These include patch types, patch lengths and 324
widths, patch area index, landscape organization indices, and inter-patch lengths (Tongway 325
and Hindley, 2004). The SSIs are then weighted according to their respective patch 326
contribution to the transect resulting in two SSI measurements: one for each patch type and 327
one for each patch type weighted by the contribution of the patch length to transect length. 328
Landscapes that have low or high values are not necessarily degraded or highly functional, 329
but by comparing to suitable reference sites or to previous measurements at the same site, 330
conclusions can be drawn as to whether a site is losing or gaining resources and therefore 331
experiencing degradation or improved functionality through more tightly retained resources 332
(Tongway and Hindley, 2004). A reference site would be one that illustrates what the 333
landscape would look like under optimum conditions when disturbances are at a minimum, 334
and in a rehabilitation context, would illustrate the target outcome for rehabilitation. 335
However such sites can be difficult to locate for many reasons. Also the nature of the 336
disturbance, gold tailings storage facilities for example where the tailings provide quite 337
unnatural edaphic conditions, may render a site highly unlikely to proceed to a state 338
comparable to undisturbed local landscapes. In such conditions, time series may be more 339
appropriate to define a trajectory for a landscape. If time series are not available, sites may 340
be ranked against each other to provide the range of functionality within the landscape. 341
However it needs to be noted that this merely provides a range under the current 342
conditions and does not necessarily reflect the potential functionality a site may achieve. As 343
the purpose of this study was to test the ability to predict LFA and ecological measurements 344
from remote sensing data, and not to evaluate the degree of functionality and disturbance 345
in the study landscape, no reference site or time series were needed. 346
Most methods developed in South Africa for measuring ecological degradation in 347
rangelands are based on veld condition with the underlying aim of maximizing grazing 348
potential (Tainton, 1999) and therefore may not be suitable for assessment of mine- 349
impacted environments. LFA was developed to monitor changes in landscape heterogeneity 350
and biogeochemical processes in disturbed landscapes in Australia (Tongway and Hindley, 351
2004). It has since been adapted for monitoring rehabilitation processes in Australian mining 352
environments (Tongway and Hindley, 2004) and has been accepted by Australian mining 353
regulators (Lacy et al., 2008). A limitation is that LFA is a subjective method dependant on 354
14 the practitioner’s skill and consistency in correctly evaluating the environmental features 355
concerned. It was therefore deemed worthwhile to include more empirical methods of 356
measuring ecosystem processes within the experimental design. This would provide a 357
broader base to the ecological data, and potentially, some independent verification of the 358
LFA indices. The ecosystem properties that LFA identifies are patch size, soil stability, 359
infiltration and nutrient cycling (Tongway and Hindley, 2004). Patch size and type are 360
descriptors of landscape organization and an aid to organizing and categorizing landscape 361
heterogeneity, whereas soil stability, infiltration and nutrient cycling are closely interrelated 362
characteristics of a soil and local environment. 363