72
N.M. Mattikalli and E.T. Engman
termination of boundary locations) from original layers can frequently be acceptable with a significantly lower level of accuracy.
Errors compound and propagate through spatial processing of digital layers.
Rasterizing a vector data layer (such as soils or watershed boundary) is one obvious source of error that invariably leads to generalization and loss of accuracy.
This may be critical depending on the application. Vector-to-raster conversion can
be considered as a form of point sampling, in which a pixel is assigned a value of
that attribute which occurs at the center of pixel (see Fig. 4.3). Resolution, size
and volume of the resulting map depend on pixel dimensions. Pixel dimension
also has a strong influence on positional errors, boundary formation between map
units, and errors in calculated areas of individual map units. Frolov and Maling
(1969) and Switzer (1975) present a basis for a mathematical treatment of errors
resulting from vector-to-raster conversion. In addition to pixel dimension, map
complexity (i. e., number of boundary pixels) affects the magnitude of error.
Crapper (1980) discusses such a problem in estimating areas of land cover from
Landsat images. If a conversion procedure and its processing options can be judiciously selected to suit a particular problem, it is possible to convert data between
two formats to minimize any errors (Piwowar et aI., 1990).
Errors compound during commonly employed overlay operations (Newcomer
and Szajgin, 1984). MacDougall (1975) discusses the result of overlaying six
maps each of which on its own was considered to be of acceptable accuracy, and
find that the resulting map was not significantly different from a random map.
Although this may be an extreme case study, the implications of error propagation
in map overlay can be serious (Bailey, 1988). The magnitude of error in the end
product is directly related to the number of data layers employed in overlay operation (Walsh et aI., 1987). Therefore, such errors can be minimized in some
applications by employing a smaller number of data layers in overlay operations at
anyone time (Mattikalli, 1995).
4.3 Current Applications
The potential of integrating remotely sensed data with a GIS has been demonstrated in many areas of hydrology and water management. Although it is not
possible to cover all areas of application, a few promising examples of current
applications are presented in the following paragraphs. An exhaustive discussion
of applications is not provided here, but the cited references provide fuller details.
4.3.1
Watershed Database Development
Development of an accurate and up-to-date watershed database is the first and an
important stage of a hydrological study. Information on hydro-meteorological
variables and watershed characteristics are stored as thematic geo-registered layers. Typically, these data layers include raw data such as digital elevation data,
multi-spectral satellite imagery, soils map, watershed boundary etc. The database
also includes derived data layers such as terrain slope and aspect, upslope area,
land cover classification, soil erodibility, evapotranspiration etc.
N.M. Mattikalli and E.T. Engman
termination of boundary locations) from original layers can frequently be acceptable with a significantly lower level of accuracy.
Errors compound and propagate through spatial processing of digital layers.
Rasterizing a vector data layer (such as soils or watershed boundary) is one obvious source of error that invariably leads to generalization and loss of accuracy.
This may be critical depending on the application. Vector-to-raster conversion can
be considered as a form of point sampling, in which a pixel is assigned a value of
that attribute which occurs at the center of pixel (see Fig. 4.3). Resolution, size
and volume of the resulting map depend on pixel dimensions. Pixel dimension
also has a strong influence on positional errors, boundary formation between map
units, and errors in calculated areas of individual map units. Frolov and Maling
(1969) and Switzer (1975) present a basis for a mathematical treatment of errors
resulting from vector-to-raster conversion. In addition to pixel dimension, map
complexity (i. e., number of boundary pixels) affects the magnitude of error.
Crapper (1980) discusses such a problem in estimating areas of land cover from
Landsat images. If a conversion procedure and its processing options can be judiciously selected to suit a particular problem, it is possible to convert data between
two formats to minimize any errors (Piwowar et aI., 1990).
Errors compound during commonly employed overlay operations (Newcomer
and Szajgin, 1984). MacDougall (1975) discusses the result of overlaying six
maps each of which on its own was considered to be of acceptable accuracy, and
find that the resulting map was not significantly different from a random map.
Although this may be an extreme case study, the implications of error propagation
in map overlay can be serious (Bailey, 1988). The magnitude of error in the end
product is directly related to the number of data layers employed in overlay operation (Walsh et aI., 1987). Therefore, such errors can be minimized in some
applications by employing a smaller number of data layers in overlay operations at
anyone time (Mattikalli, 1995).
4.3 Current Applications
The potential of integrating remotely sensed data with a GIS has been demonstrated in many areas of hydrology and water management. Although it is not
possible to cover all areas of application, a few promising examples of current
applications are presented in the following paragraphs. An exhaustive discussion
of applications is not provided here, but the cited references provide fuller details.
4.3.1
Watershed Database Development
Development of an accurate and up-to-date watershed database is the first and an
important stage of a hydrological study. Information on hydro-meteorological
variables and watershed characteristics are stored as thematic geo-registered layers. Typically, these data layers include raw data such as digital elevation data,
multi-spectral satellite imagery, soils map, watershed boundary etc. The database
also includes derived data layers such as terrain slope and aspect, upslope area,
land cover classification, soil erodibility, evapotranspiration etc.