Spatial Modelling of Forest Watershed Processes 
Lawrence Band, Ramakrishna Nemani, University of Toronto 
Steven Running, Joseph Coughlan, University of Montana 
David Peterson, Jennifer Dungan, NASA/Ames Research Center 
ABSTRACT 
A geographic information processing and modelling system is under construction for the 
simulation of forest watershed processes over a range of spatial scales. The system organizes 
all environmental information by a formal landscape feature model, utilizing the hillslopes and 
drainage lines of a watershed as the spatial framework for storage and manipulation of model 
parameters. Parameters are derived from a combination of remotely sensed imagery, digital 
terrain data, digitized soil maps and base station meteorologic data. Simulation is distributed 
over the watershed area by seperately parameterizing and simulating hydroecologic processes for 
each hillslope. 
INTRODUCTION 
We describe a system for automating the 
spatial data handling, model parameterization 
and simulation of forest ecosystem processes 
within watersheds of varying scales. We 
link an existing ecosystem model, FOREST- 
BGC, which simulates the cycles of water, 
carbon and nitrogen between soil, canopy 
and atmosphere (Running and Coughlan, 
1988), with a set of algorithms specifically 
designed to stratify and manipulate spatial 
data in a watershed. The system is designed 
to first identify and extract all stream 
channels and hillslopes in a watershed as 
data objects, and build a hierarchical 
database in which the data objects are 
explicitly defined in relation to all other 
objects in the natural watershed hierarchy. 
The data model used here includes all 
hillslopes and stream channel links as the 
most primitive objects, from which 
composite objects of larger hillslopes, 
subcatchments and watersheds can be easily 
formed by use of a formal geomorphic 
model of drainage basins. 
FOREST-BGC simulates the net 
primary productivity (NPP) and 
évapotranspiration (ET) of a forest stand. 
Extensive sensitivity analysis of FOREST- 
BGC has shown the most important 
parameters are net radiation (R), atmospheric 
vapor pressure deficit (VPD), canopy leaf 
area index (LAI), air temperature (T) and the 
available soil water capacity (SWC). In 
mountainous environments, the close 
coupling of these parameters with the 
topography suggests that use of hillslope 
objects to spatially aggregate the parameters 
results in lower internal parameter variance 
and higher between unit (hillslope) 
parameter variance. The relatively 
conservative range of net radiation over a 
hillslope facet compared to the range 
expected over arbitrarily located grid cells 
(such as AVHRR pixels) allows a much 
greater discrimination of well expressed and 
observable landscape patterns that otherwise 
may be filtered out. In addition, hillslopes, 
stream channels and watersheds are 
functional hydrologic units as sources and 
conveyors of water, and allow aggregation 
into larger functional units, unlike arbitrarily 
located grid cells. 
In order to discriminate parameter 
values between hillslopes, it is necessary that 
parameters be estimated at significantly 
higher spatial resolution than that of the 
hillslopes. Consequently, high resolution 
imagery and terrain data must be used to 
produce initial parameter fields, from which 
the hillslope parameterization is developed. 
In this paper, RESSYS (Regional