Quantifying momentum, mass and energy exchange between heterogeneous terrain and the atmosphere at local- and meso-scales is the big challenge in boundary-layer meteorology. Although a theoretical treatment seems impossible due to the level of complexity, various statistical and empirical approaches have been proposed that may allow partial solutions. Proper validation and improvement of such techniques require high quality data collected at various scales. In the boreal ecosytem-atmosphere study (BOREAS 94 and 96), which was conducted to address this need, the Canadian Twin Otter aircraft team extensively mapped the fluxes of mass and energy and radiometrically observed surface conditions over two 16 km × 16 km areas from grid flights executed at 30 m height. A major finding is the rather surprising anticorrelation between radiometric "hot spots" and sensible heat flux observed particularly over black spruce at the northern study area of BOREAS, which illustrates the difficulty of predicting fluxes from surface observations in such ecosystems.
Our relative success with Artificial Neural Networks (ANNs) in sensible heat flux estimations over a grassland has encouraged us to consider ANNs as a possible solution for linking patchy landscapes to fluxes of mass and energy. The above mentioned flux maps, along with land-cover information from a variety of data bases and observation platforms, including satellites, will be used to train ANNs to associate a given flux pattern with possible matching landscape, or vice versa, at 1 km × 1 km pixel size. The potential of these associations for scaling up from local to regional scales, and for the planning and interpretation of observations in the upcoming GEWEX-MAGS (Global Energy and Water Cycle Experiment - Mackenzie Area GEWEX Study) will be outlined.