Spectral and human sensors : hyperspectral remote sensing and participatory GIS for mapping livestock grazing intensity and vegetation in transhumant Mediterranean conservation areas

Increasing shortage of pasture resources due to land use conversion
constitutes a major challenge to traditional transhumance systems.
Reduction of transhumance and related activities leaves the non
converted areas abandoned. This may lead to change in grazing
intensity, which might result into change in species composition and
vegetation pattern. A reduction in grazing intensity might thus
influence the biodiversity and forage quality of previously more
intensively grazed areas. Proper management of Mediterranean
grasslands would require insight on how grazing intensity varies
across a landscape and how it influences the distribution and
abundance of plant species.
The aim of this study was to investigate methods for mapping of
livestock grazing intensity and vegetation, using hyperspectral remote
sensing, geographic information systems (GIS) and participatory GIS
(PGIS). Investigations were undertaken at two main levels. A
greenhouse experiment was used to investigate the effects of
defoliation and defoliation time for two species grown in mono and
mixed culture on the height and dry matter yield as measures of
regrowth and competitive ability of two livestock forage grasses
selected from a transhumant Mediterranean area. Narrow band
hyperspectral reflectance, indices and the red-edge position were
investigated to see if they may be used to study these effects. At field
landscape level, we tested the use of local people’s knowledge in
mapping grazing intensity through the application of PGIS.
The results from the greenhouse experiment showed that the species
with higher dry matter yield (Lolium multiflorum) had a significantly
higher relative regrowth rate and possibly higher competitive ability
than its competitor Dactylis glomerata (P < 0.05). Increase in dry
matter yield was shown as the trait that determines competitive
ability in the early established stage of the two grass species (period
of 13 to 18 weeks after sowing). The experiment also provided insight
on the persistence of forage species that are of grazing preference.
Selective clipping did not alter the competitive ability of D. glomerata
to surpass that of L. multiflorum when the former was clipped at
lower clipping intensity to simulate selective grazing.
The hyperspectral remote sensing variables that may be used to
estimate the effect of species types, cultures and defoliation
treatments were: the physiological reflectance index (PRI), the Carter
index, R694, the ration of the Transformed Chlorophyll Absorption in
Reflectance Index to the Optimized Soil-Adjusted Vegetation Index
(TCARI/OSAVI) and the red-edge position. The PRI was found to be
the most sensitive index. A significant increase (p < 0.001) in PRI
was associated with the higher competitive ability of L. multiflorum
than D. glomerata when the two were mixed. The response of the PRI
from negative to positive over the measurement time in relation to
height and dry matter yield suggest that the PRI may be used to
study competitive ability because the related growth characteristics
are indicators of competitive ability. This encourages further
investigation of this method as a potential simpler and quicker
alternative to the existing canopy height and pasture growth models.
This may lead to efficient assessment and improved understanding of
the condition and spatial patterns of forage vegetation species at field
level.
At field landscape level, using participatory GIS (PGIS), spatial
knowledge on grazing intensity from pastoralists and local range
ecology experts was elicited and relevant criteria generated and used
to classify grazing intensity. Local pastoralists appeared to be more
knowledgeable than local range ecology experts, possibly because of
the pastoralists’ superior familiarity with the rangeland and better
perceptions about the distribution of palatable species but the experts
represented the grazing intensity better on a map. Local pastoralists
have potential to contribute better to this process if the PGIS includes
adequate training in the map making process. The local experts
showed the capability to produce data and synthesize spatial
variables, but it was also shown that the expert-based PGIS maps
may not always be reliable. Using a proposition that “This area or
pixel belongs to the high, medium, or low grazing intensity class
because the local expert(s) says (say) so”, we tested for uncertainty
in the PGIS-maps produced by different local experts using spatial
tools such as evidential belief functions (EBFs).
Evaluating the classification uncertainty in the different grazing
intensity maps revealed that the maps with the lowest uncertainty
were based on the composition of palatable vegetation species as the
mapping criterion. This criterion may be used for mapping grazing
intensity because it relates to measures of forage condition such as
ground cover and quality, but it may be limited in use if other
parameters such as vegetation composition and quantity are not
integrated. If the definition of grazing intensity also includes these
parameters and also livestock vegetation use factor and impacts on
vegetation, then the proposition for EBF evaluation would be that:
“This pixel or area is a specific grazing intensity class because of the
level of livestock grazing use and its impacts on species composition,
ground cover, quantity and quality. These parameters may be
efficiently estimated using hyperspectral remote sensing. In order to
include local knowledge in such an evaluation, research should
establish how local pastoralists and experts may process the various
parameters and how they may apply such a proposition.
Since more than one criterion proved cumbersome for the local
experts as evidenced by a weak correlation between the grazing
intensity map and a grazing suitability index (r =0.35 (p < 0.01)),
spatial multiple criteria tools may be useful for synthesizing the
different mapping criteria.
Overall, this study showed that high spectral resolution sensors can
detect the effect of grazing and competitive interactions among
forage plants through narrow band channels across the spectrum,
while the local people perceive a few broad grazing intensity classes
and spatially represent them using a few criteria. The two are
complementary. The spectral sensor provides detailed information on
the status and spatial patterns of vegetation, while local participants
provide the spatial information on a more general coarse scale that
may be used as baseline for hyperspectral remote sensing research.