different statistical approaches. As a consequence, the separation criterion shifted
from eTh ¼ 16 K (ground-based) towards eTH ¼ 12.5 K for the airborne data.
This systematic shift thus needs to be taken into account when applying airborne
data. Whether further spatial variation of the airborne criterion has to be considered, e.g., due to differences in vegetation or land use, needs further research
(Schuler et al. 2011).
Regional scale use of radiometric data for concept soil map establishment: As
stated earlier, soil and other baseline information is scarce in relation to the northern
Thai highlands. Apart from topographic (and derived) information, radiometric
data represents that with the highest density of sampling points and additionally a
systematic transect-based sampling. As a result, radiometric data was used as one of
the information layers in the concept soil map, based on a statistical approach. In
this case, a random forest approach was chosen (see Sect. 2.3).
The Gini index, representing the relative influence of data layers on map
development (Fig. 2.5) showed that only two data groups had a dominating
influence: i. radiometric data – with K radiation having the greatest influence,
and ii. topographic data – with absolute topographic height above sea level
dominating. On the other hand, parent material exerted only a minimal influence,
and this result needs to be explained. Obviously, data density was an important
determinant of this result. Since only limited soil mapping data were available –
which did not cover all the parent rock types to be found in the northern Thai
highlands, the petrographic data layer had to be further generalised based on an
existing map which had a relatively low resolution (German Geological Mission
1979, 1:250,000). As a consequence, the parent material as a data layer covered
the whole area, but in fact showed the lowest diversity (only categories) and the
lowest data density.
Fig. 2.5 Gini index of all potential predictors used for the establishment of the regional soil map
of Northern Thailand. Where k potassium, th thorium (equivalent), u uranium (equivalent), dem
elevation, aspect aspect, curv curvature, plc planform curvature, prc profile curvature, tri terrain
ruggedness index, vdcn vertical distance above channel network, wi SAGA wetness index, slope
slope, ra relative altitude and pm parent material
2 Beyond the Horizons: Challenges and Prospects for Soil Science and Soil. . .
47
from eTh ¼ 16 K (ground-based) towards eTH ¼ 12.5 K for the airborne data.
This systematic shift thus needs to be taken into account when applying airborne
data. Whether further spatial variation of the airborne criterion has to be considered, e.g., due to differences in vegetation or land use, needs further research
(Schuler et al. 2011).
Regional scale use of radiometric data for concept soil map establishment: As
stated earlier, soil and other baseline information is scarce in relation to the northern
Thai highlands. Apart from topographic (and derived) information, radiometric
data represents that with the highest density of sampling points and additionally a
systematic transect-based sampling. As a result, radiometric data was used as one of
the information layers in the concept soil map, based on a statistical approach. In
this case, a random forest approach was chosen (see Sect. 2.3).
The Gini index, representing the relative influence of data layers on map
development (Fig. 2.5) showed that only two data groups had a dominating
influence: i. radiometric data – with K radiation having the greatest influence,
and ii. topographic data – with absolute topographic height above sea level
dominating. On the other hand, parent material exerted only a minimal influence,
and this result needs to be explained. Obviously, data density was an important
determinant of this result. Since only limited soil mapping data were available –
which did not cover all the parent rock types to be found in the northern Thai
highlands, the petrographic data layer had to be further generalised based on an
existing map which had a relatively low resolution (German Geological Mission
1979, 1:250,000). As a consequence, the parent material as a data layer covered
the whole area, but in fact showed the lowest diversity (only categories) and the
lowest data density.
Fig. 2.5 Gini index of all potential predictors used for the establishment of the regional soil map
of Northern Thailand. Where k potassium, th thorium (equivalent), u uranium (equivalent), dem
elevation, aspect aspect, curv curvature, plc planform curvature, prc profile curvature, tri terrain
ruggedness index, vdcn vertical distance above channel network, wi SAGA wetness index, slope
slope, ra relative altitude and pm parent material
2 Beyond the Horizons: Challenges and Prospects for Soil Science and Soil. . .
47