4.5 Program Input Data
FlowCast uses the concept of projects to store data, with currently only one project able to be open at any time. Projects contain input data linkages and metadata for map files, predictor definition files, predictand files, and rule-set definition files. In order to reduce startup times and memory loadings, FlowCast does not automatically preload data, rather it retrieves individual data items when accessed by the user, or though saved batch configuration.
FlowCast stores map data in a binary TeeChart format, although support is currently being developed for the XML-based GML (Geographic Markup Language) and KML (as used in “Google Maps”) formats. Currently, maps can be exported from ArcView in shapefile format and converted into the TeeChart binary file using an external converting program.
Predictor information is currently imported into FlowCast through predefined external predictor definition files (*.pdr), and linked time-series data files. The predictor files are XML-based containing calculation type and averaging period, and time-series metadata including data description, local file path, and external URI for Internet based updating. Stratification rules are defined here if applicable, including formulas, names, and colours, and more than one time-series data can be included for discriminant analysis systems. Time-series are stored externally in a variety of supported monthly ASCII formats.
The project file contains the metadata and file-paths (linkages) of predictand time-series data stored in either daily or monthly format. A wide range of data formats are supported including a monthly Microsoft Excel based format, Rainman monthly format, SILO daily (.p51) and monthly formats (http://www.apsru-silo.gov.au), IQQM daily and monthly formats , limited netCDF formats, and a range of simple ASCII formats. There is no limit on what type of time-series data that can be included as a predictand (e.g. rainfall, temperature, etc), although the user must specify whether the data can be summated (like rainfall) or averaged (like temperatures). Associated metadata includes data ranges, missing data statistics, external URIs, and geographical information.
A range of forecasting rules and configurations are imported through predefined external rule-set definition files (*.rul). These are XML-based files containing information about training and testing periods, cross-validation, and threshold (tercile, median, etc) calculation methodology. Files are named according to their effect such as “Full Training Data Range with Cross Validation”.
4.6 Program Outputs and Analyses
Program outputs are grouped according to their functionality type. This includes a range of ‘Browser’, ‘station’ and ‘spatial’ analyses. Only one analysis can be presented at any time, except during spatial analysis when individual browser and station based chart outputs can be displayed as geographical overlays. Each program output will now be summarized in more detail.