Télédétection et ressources en eau/Remote sensing and water resources
189
RS to map land use/cover to help determine the SCS Curve Number (Sharma, 1992;
Zevenbergen et al., 1988). These report very good results and without taking the PAC
phenomenon into account.
Nonetheless, Tauer and Humborg (1992) caution that the CN system is not relevant for the
Sahel because it is callibrated for extreme events in the USA and generates virtually no runoff
below 40 mm rainfall, which is larger than most rain events in the Sahel. In any case, the size of
catchments amenable to this approach are larger than the catchment area commonly used in WH.
The fundamental irony in attempting to use RS to map WH potential is that the scale ‘visible’ to
satellite sensors is suited to assessing Macro WH, but the most widely used and socially realistic
system is Micro WH because it is based on individual ownership, which involves catchment areas
invisible to RS.
From the ground up: techniques relevant for assessing MICRO WH suitability
The other approach taken specifically in the Sahel to WH starts from the ground and aggregates
upwards to the ‘point of visibility’ on RS imagery and is of greater relevance to WH because it
deals with surfaces as small as 100 m
2
, still within the Micro WH category. This approach in
semi-arid Africa has been the work of French scientists and driven by the opportunity to take
advantage of an existing resource, the Catalogue, by linking it to RS. As the Catalogue is based
on ‘elemental surfaces’ of 1 m
2 the problem has been how to aggregate up to about the 1 ha
minimum discernible area or ‘effective pixel’ on SPOT or TM [once various georeferencing
errors add up]. There are a number of possible ways of doing this.
Casenave and Valentin (1989) established a relationship between a classic classification of
‘visible features’ of the landscape by RS in terms of geomorphic features such as ‘rocky slopes’
visible on the imagery and their constituent surfaces. The surfaces typically linked to each RS
class are grouped in a manner analogous to soil associations and consist of, for example, 75%
surface type A and 25% surface type B. The relationship allowed prediction of runoff on an
unmapped catchment by calculating from the Catalogue the net runoff as a function of the
proportion of the catchment occupied by each association and the presumed proportion of surface
types within each association as previously determined from training areas. PAC was not
considered but gave a regression coefficient of 0.80 with respect to measured runoff. The other
approach would be to try to pick up the reflectance characteristics of the elemental associations
directly where they occur in large enough patches to serve as training areas. This was attempted
by Puech and Laily (1990) but only four of twelve associations could be distinguished. In
response, an effort was made to correlate the associations to vegetation and soil types, as these
are visible to RS (Puech, 1994) and this has proved to be feasible. Their current research involves
the development of quick transect based surveys based on the Catalogue, designed to furnish a
large enough sample to accurately interpret imagery on an operational basis (Lamachère and
Puech, 1995).
Yet another possibility is to start on the ground not just with Rsim but also with radiometric
reflectance; ie to relate the two in situ, and then relate this correspondance to satellite imagery. A
hand held spectoradiometer is being used in the pilot area to determine whether a relationship can
be established between propensity to generate runoff and spectral signature either directly or via
proxy measure(s).
189
RS to map land use/cover to help determine the SCS Curve Number (Sharma, 1992;
Zevenbergen et al., 1988). These report very good results and without taking the PAC
phenomenon into account.
Nonetheless, Tauer and Humborg (1992) caution that the CN system is not relevant for the
Sahel because it is callibrated for extreme events in the USA and generates virtually no runoff
below 40 mm rainfall, which is larger than most rain events in the Sahel. In any case, the size of
catchments amenable to this approach are larger than the catchment area commonly used in WH.
The fundamental irony in attempting to use RS to map WH potential is that the scale ‘visible’ to
satellite sensors is suited to assessing Macro WH, but the most widely used and socially realistic
system is Micro WH because it is based on individual ownership, which involves catchment areas
invisible to RS.
From the ground up: techniques relevant for assessing MICRO WH suitability
The other approach taken specifically in the Sahel to WH starts from the ground and aggregates
upwards to the ‘point of visibility’ on RS imagery and is of greater relevance to WH because it
deals with surfaces as small as 100 m
2
, still within the Micro WH category. This approach in
semi-arid Africa has been the work of French scientists and driven by the opportunity to take
advantage of an existing resource, the Catalogue, by linking it to RS. As the Catalogue is based
on ‘elemental surfaces’ of 1 m
2 the problem has been how to aggregate up to about the 1 ha
minimum discernible area or ‘effective pixel’ on SPOT or TM [once various georeferencing
errors add up]. There are a number of possible ways of doing this.
Casenave and Valentin (1989) established a relationship between a classic classification of
‘visible features’ of the landscape by RS in terms of geomorphic features such as ‘rocky slopes’
visible on the imagery and their constituent surfaces. The surfaces typically linked to each RS
class are grouped in a manner analogous to soil associations and consist of, for example, 75%
surface type A and 25% surface type B. The relationship allowed prediction of runoff on an
unmapped catchment by calculating from the Catalogue the net runoff as a function of the
proportion of the catchment occupied by each association and the presumed proportion of surface
types within each association as previously determined from training areas. PAC was not
considered but gave a regression coefficient of 0.80 with respect to measured runoff. The other
approach would be to try to pick up the reflectance characteristics of the elemental associations
directly where they occur in large enough patches to serve as training areas. This was attempted
by Puech and Laily (1990) but only four of twelve associations could be distinguished. In
response, an effort was made to correlate the associations to vegetation and soil types, as these
are visible to RS (Puech, 1994) and this has proved to be feasible. Their current research involves
the development of quick transect based surveys based on the Catalogue, designed to furnish a
large enough sample to accurately interpret imagery on an operational basis (Lamachère and
Puech, 1995).
Yet another possibility is to start on the ground not just with Rsim but also with radiometric
reflectance; ie to relate the two in situ, and then relate this correspondance to satellite imagery. A
hand held spectoradiometer is being used in the pilot area to determine whether a relationship can
be established between propensity to generate runoff and spectral signature either directly or via
proxy measure(s).