ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS", Bangkok, May 23-25, 2001 
ISPRS, Vol.3 
tor the identification and location of areas at high risk of change 
and subsequent increase in pesticides pollution(Stephen and 
Kyehyun, 1993; Vassilisos et. al., 1997). High-resolution digital 
remote sensing data have several characteristics that make 
them well suited for contributing to pesticides pollution 
assessment studies: Computer analysis of multispectral data 
allows rapid mapping and monitoring of land cover types and 
conditions; Terrain analysis helps to show the hydrological 
characteristics of study areas; The regional coverage of 
satellite data provides a cost-effective method of rapidly 
inventorying extensive areas; Additionally, the digital maps 
produced can be readily integrated with other information in a 
GIS or used to derive input parameters for mathematical 
models that predict the pesticides pollution potential (Mark et. 
al., 1992; Hanadi et. al. 1993; Kurt and Robert, 1993; Mark et. al. 
1994; Adamus and Bergman, 1995; Nuckols et. al., 1996). 
Fig.1 Study area: Kintore Creek in south of Ontario, Canada 
sub-watershe 
year round sc 
Branch of th 
Ontario, befoi 
heavily used | 
Table 1. Kint( 
Size of sub Wi 
Soil types 
Soil erosion p 
Area under st 
Area tile drain 
Total forest cc 
Total crop are 
In the wee 
conservation 1 
row crops, th< 
reduced till pr< 
slope stabiliza 
eastern sub-\ 
tillage pract 
corn-wheat-all 
DATA DEVELi 
At present, the efficient cooperation of nonpoint source 
(Nitrogen, Phosphorous, etc.) pollution model and GIS model is 
main research trend (Heidtke and Auer, 1993). Several studies 
have illustrated the role of GIS in supplying data and 
information for assessing nonpoint source pollution attributes 
and formulating land resource planning and management 
strategies. For instance, Newell et al. (1992) created a ranking 
of nonpoint source water pollution loads in Galveston Bay, 
Texas, using eight land-use categories derived from Landsat 
TM data incorporated with soil run-off models rainfall amounts, 
and water quality parameters. Subra and Waters (1993) 
examined an area of southwestern Louisiana to develop a 
prototype nonpoint source pollution model using 15 land-cover 
types mapped from TM imagery, watershed, hydrography, 
slope, and soil type data. A Connecticut watershed was the 
focus of research by Nelson and Arnold (1995). Six categories 
of land- cover were extracted from TM imagery and weighted by 
their percent of impervious areas to produce current and future 
runoff values. Floyd et al. (1998) examined the positive 
potential of an existing satellite-based (TM) land-cover data set 
(Coastal Change Analysis Program) in a rapidly developing 
coastal area for nonpoint source water pollution controlling and 
management. 
However, few successful studies are found in the integration of 
pesticides pollution model with GIS and remote sensing. 
Especially, real landscapes are three-dimensional and this 
three-dimensionality has a major impact on the hydrologic and 
erosional processes occurring on the landscape. Few models 
with GIS and remote sensing applications are capable of 
accounting for this kind of three-dimensionality (John, 1991). 
The objective of this paper is to demonstrate how terrain 
analysis methods could be applied in pesticides pollution 
control model to improve their prediction capabilities and 
decrease the time and effort required to assemble the input 
data sets. A Pesticides Surface Runoff model was used to 
integrate with GIS and remote sensing techniques. The terrain 
analysis approach was used to resolve the runoff flow direction 
based on Digital Elevation Model (DEM) photogrammetrically 
derived from aerial photograph. The flow direction as an 
important parameter was then input into the pesticides runoff 
model to simulate the pesticides losses. 
METHODOLOGY 
STUDY AREA 
The study area is the Kintore Creek watershed (Latitude: 
43°. 189 ~ 43°. 145, Longitude: -81°.075 ~ -80° 995), which has 
two adjacent sub-watersheds, in south Ontario, Canada (Fig.1). 
The total area of the watershed is 1,288 ha. The 
sub-watersheds have nearly equal size, similar highly erodible 
landscapes, and cropping patterns. Detailed Kintore Creek 
sub-watershed characteristics are listed in Table 1. Both 
In this paper 
Information C< 
generate a it 
grid-DEM cree 
data for the p< 
parameters of 
ArcView 3.1 C 
Slope and asf 
on the terrair 
determination 
For the study c 
the effects o 
transport to s 
Canada and tl 
some relative 
were also obta 
Table 2. Parai 
SCS curve 
Land slop« 
Slope sha| 
Field slope 
Channel si 
Channel si 
Roughnes; 
Cropping f 
38