90 
cation, only a subset of functionality is needed. Loose integration excels at mixing and matching 
the right pieces from available tools to deliver the desired functionality for a given application. 
What is needed to accomplish this is an integration framework, and modular components that can 
be plugged into it [Kleinfeldt et al. 1996a]. 
4.1 A DATA FLOW-BASED FRAMEWORK 
The data flow paradigm, as used by some scientific visualization systems, is a very intuitive way 
to build and to represent complex applications. A flow consists of modules ("boxes”) which are 
connected by channels through which the data "flows” (Figure 2). A module is an encapsulated 
function and performs a specific piece of the overall process. 
The flow concept has also been implemented in the world of computer aided design (CAD). We 
believe that some of the integration techniques originally developed for CAD applications, such as 
CAD frameworks, could be usefully adopted for environmental modeling applications. CAD frame 
works are software infrastructures for building integrated environments out of uncooperative, often 
proprietary CAD tools. CAD frameworks have a common data repository (the design database) 
that promotes tool interoperability. They maintain information about the structure and status of 
the design to provide management support and to enforce constraints on the design process (the 
design data manager), and they support the user in correctly executing the required tools for a 
given task (the design process manager) [van der Wolf 1994]. The different steps of the process 
are logically connected in a modular flow, called the design flow. Design flow management is a set 
of CAD framework functions concerned with assisting designers during the (design) process. The 
functions of (design) flow management are: tracking (to determine the progress made and to decide 
on consecutive operations to be performed): constraint enforcement (to keep users from making 
certain errors); assistance (to help users understand the process): and automation (to relieve users 
of straightforward, repetitive operations). 
The data flow paradigm in CAD systems, based on CAD frameworks as an integration platform, 
contains features that enable developers and users to address complex processes in an intuitive way. 
5.0 TWO EXAMPLES 
The examples described here are being developed as part of a larger research project combining 
hydrologic modeling and field experiments at four European basins to evaluate the use of satellite 
microwave data for soil moisture estimation [Troch et al. 1995]. The first example is a pragmatic 
approach perhaps best suited to very simple integration problems. For more complex problems 
we will need some of the concepts described in the previous section and illustrated in the second 
example. 
5.1 Tcl/Tk INTERFACE 
A first, basic, example of integration for the hydrologic modeling process consisrs in the design of 
scripts which automate the pre- and post-processing of the simulation task. The integration cost of 
this operation is minimal, since it can be carried out with basic utilities, such as Unix shell scripts, 
that should be familiar to knowledgeable computer users. A primitive "process flow” scheme is