1. INTRODUCTION

The work done at CSIRO Division of Water Resources (DWR), based on the Itres (Canada) CASI airborne spectral imaging instrument, is an Applications Based systems development. This development is a package consisting of an Integrated Airborne (spectral) Sensing (IAS) instrument, ancillary instrumentation and processing modules. Together, these can provide high resolution information for environmental resource assessment in a form immediately useful to end users of the products. The Applications area defined for this IAS involves integrating models and methods based on CSIRO DWR core knowledge and skills with a general system comprising data collection instruments, processing methodologies and processing hardware and software systems.

2. APPLICATIONS AREA MODEL
The outputs of the Applications system are final products that users want and are prepared to contract commercially to a processing service bureau. The necessary Application template for an imaging spectrometer such as (but not limited to) the CASI, takes into account the special circumstances provided by a fully calibrated spectral instrument. It integrates its output with a range of other airborne and ground based measurements and has nine main components.

Figure 1. Simplified Applications Area Structure Schematic

The subcomponents (primary processes) of this system have been studied in a detailed form as a cost/time and work/duration processing model through applications carried out by the established processing group at CSIRO DWR. That model and possible extensions based on queuing theory have been investigated at CSIRO DWR and by CSIRO DWR's commercial product developers as part of the Applications model. This model has been set up to help manage and monitor the progress of actual Applications projects. Figure 1 shows the main structure of the model:

2.1 Flight Mission (Planning and Execution)
Every application will have special considerations with respect to channels chosen for targeting user needs, environmental conditions, data scale and resolution. CSIRO DWR uses a set of software items which implement the flight planning and measurement models. These can be used to plan prior to flights and during the mission.

2.2 Lab and Field Data Collection & Processing
Some field data are collected at the time of each mission and some at other times. Data which must be taken at the time of flights is environmental (such as atmospheric data). Data such as characteristic spectra can be collected at other times. Separation of field and laboratory from flying missions is a key development in the measurement model approach being discussed here. Software to analyse the Lab and Field data for each measurement model for the analyses have been developed to implement this area.

2.3 Base Data Processing
CASI Suppliers (Itres, Canada) software is used to take image, attitude (roll and pitch) and GPS data from the aircraft and process them to calibrated and (by single run) geometrically resampled images in a form ready for image processing. There is a future need to develop software to use the GPS data to more effectively and accurately model the attitude of the aircraft for IAS applications. This is under development by CSIRO DWR in combination with Itres and the CSIRO DWR commercial product developers.

2.4 Mosaicking and Normalisation
The Base Processed airborne sensor data are assembled into mosaicks and atmospherically corrected in this component. Atmospheric correction means the data from different runs and times become normalised to surface reflectance. Only real differences in surface reflectance are left. At the heart of this process is an integrated spectral and geometric model for the data. This model is also the basis for fusion of the data with other data types - such as satellite data, other airborne data and GIS data at various scales.

2.5 Image Processing
The necessary functions of the image processing system can be fully described by function rather than by options in any single image processing system. At present, microBRIAN (MPA and CSIRO developed software) is being used for the CASI based prototype Projects. The use of this specific software for an advanced IAS is sufficient but not necessary. CSIRO DWR supervising software exists for the current system which makes processing tasks quicker and minimises operator action.

2.6 Measurement Models
The primary taxonomy of 'applications' from the Application system point of view is by measurement model. At present there are an atmospheric and six main surface measurement models:
n water bodies and water covered targets (eg sea grass, reefs, water quality)
n dense woody vegetation (eg forests & mangroves)
n open forests & woodlands (eg woody weeds)
n short grass and crops (eg crops and grasslands)
n bare targets (eg soils)
n (modulated by) flat and sloping ground effects
The measurement models provide the link between the parameters of interest and concern to the user and those of direct significance in forming the signal measured by the instrument. The CASI measures radiance in the visible and near infrared so that CASI measurement models involve primarily radiative transfer and scattering in the surface layer and atmosphere. However, users want to know parameters of the surface such as canopy cover and structure, soil moisture, water use, levels of phytoplankton in water bodies, benthic community type or suspended sediments concentration. The measurement models provide the scale and physical basis for these transfers from CASI data to the information required by the end user.

2.7 Measurement Model Knowledge Base
This area includes a Spectral Library and associated environmental information. The data structure and methodology for the Spectral Data Base has been developed at CSIRO DWR. The Measurement Model Knowledge Base, together with software called the 'Data Calculator', form base tools for the measurement models and applications. In particular, a spectral library is a primary tool for template matching and data modelling.

2.8 GIS Integration
The system needs two-way access to GIS data and tools. The image processing component must also have adequate data integration tools to do this. Currently, CASI based Projects interface to ArcInfo and GRASS. This component can, however, be any adequate commercial GIS system. There is benefit in it being off-the-shelf. In land operations, this area of the Applications Model must also provide the best current DTM system data as well as cadastral information.

2.9 Interpretation
Interpretation is the action of merging the information and products of modelling and image processing to derive the final products in the terms the end user needs. It involves skilled and experienced scientists labelling classes, attaching numerical levels and statistics and generating keys and summaries. The interpretation depends on the Application and on the type of models that have been employed. This area is a key component in the Applications based approach and the accumulated experience here will grow as an asset of any commercial implementation of the model. Final results will normally be included in a developing GIS and build a base of decision support for future missions to use.

2.10 Applications Products
This area involves Report Generation, product generation and annotation and product delivery and presentation. This area can use existing hardware and software, the tools in the Image Processing System and the GIS system as well as new software. In the current CASI based prototype, most of these functions are provided by the microBRIAN image processing system, CorelDraw and standard publication tools such as Microsoft Word, Chartist and MathType.
We can itemise the current set of specialised product areas for the Applications under five headings. These are:

• Coastal zone monitoring and mapping
• Reef and seagrass mapping and monitoring
• Water quality mapping and monitoring
• Forest monitoring and mapping
• Plant community determination and monitoring.

3. IMPLEMENTATION
In projects undertaken at CSIRO DWR, the system described above has been implemented by a well equipped and trained team of remote sensing specialists. The cost structures have been analysed as well as processing management and potential efficiencies identified. There is now a very good processing management and planning model which can be keyed into a commercial business plan to provide effective commercial planning and management. This established team and its commercial product developers are equipped to work together with other service groups throughout our region.
The current systems implementation and structure is under review with the commercial product developers and has a number of efficiency and throughput goals which will improve commercial operations.