Many landscape models require extensive computational effort using a l
arge array of grid cells that represent the landscape. The number of s
patial cells may be in the thousands and millions, while the ecologica
l component run in each of the cells to account for landscape dynamics
is often process based and fairly complex. To compensate for the incr
eased computational complexity of the model there is a tendency to sim
plify the hydrologic component that fluxes material horizontally acros
s the landscape. Instead of full scale hydrologic models based on stab
le implicit schemes, computationally simpler explicit algorithms are i
ncorporated and run with quite large time steps. As a result some fair
ly inadequate behavior may be observed, especially when the temporal a
nd spatial steps are modified without due care. We illustrate these pr
oblems with a series of runs performed using the Everglades Landscape
Model (Southern Florida, USA), that covers an area of more than 10000
km(2). Several algorithms for hydrologic fluxing are compared in terms
of their computational complexity and stability. We argue that a comp
romise can be drawn by supplementing the explicit modeling scheme with
a series of additional checks and conditions that provide for model s
tability, and with some empirical assumptions that allow the model to
operate over a sufficiently large range of temporal and spatial scales
. (C) 1998 Elsevier Science B.V. All rights reserved.