One of the most important issues facing mesoscale modelling and atmospheric energy budget studies in complex terrain, is addressing the complicated pattern of energy fluxes that occur at the boundary of the Earth's surface. This study was designed to improve understanding of the characteristics and mechanisms of spatial and temporal variability in surface energy fluxes within a large alpine catchment. An observational dataset of time series and space series of surface radiation and energy fluxes in a 2500 km-2 catchment in the Southern Alps of New Zealand was generated. Observations were made for a period of fourteen months at a single site in the MacKenzie Basin and at nine locations in the Tekapo catchment on clear-sky summer days, during a month-long intensive observation period. The seasonal range in Q* was found to be 14.2 MJ m-2 dy-1, resulting primarily from the annual cycle of K*, which provided greater variability than L* by a factor of three. The range in Q* due to synoptic influences was of a similar magnitude and was found to be most significantly a function of cloudiness, which was seen to vary significantly from one end of the study area to the other. Daily Bowen ratio (β) values ranged from 6.85 during the 1997/98 summer to -0.13 for the 1998 winter and 9.38 for the 1998/99 summer. A consistent logarithmic relationship was found between β and soil moisture content for all data, irrespective of season or synoptic situation. Overall, the largest impact on surface energy fluxes on the temporal scale was found to result from synoptic controls on the magnitude of net radiation, and secondly from seasonal control on the partitioning of turbulent fluxes. Spatial series of observations focused on the five dominant surface classes in the catchment including, rock, lake water and three classes of vegetation. From this dataset, energy balances were derived for each surface class, to assess the spatial variability associated with heterogeneity of surface properties. A large range in surface flux density was found, related to radiative, thermal, hydrological and vegetative properties of the surface. The lake surface recorded the highest daily total Q* values of the five surface types, rock surfaces recorded the lowest and the three vegetated surfaces recorded very similar flux densities. The most significant influences were the mean diurnal albedo and the ratio between L↑and absorbed allwave radiation. Significant differences also existed between the surface types in the diurnal range of albedo. Modelling of surface radiation flux components was conducted using SRAD, a topographically based radiation model. Output from the model compared well with observations, with closer agreement found for daily mean than instantaneous fluxes. In both observational and modelled results, K↓ was found to contribute most significantly to the radiation budget both in terms of the mean and variability of flux magnitudes. By contrast, longwave radiation fluxes contributed little variability to the spatial distribution of Q*. The variability in K↓, was associated most strongly with slope aspect, secondly with slope angle, thirdly with shading, and lastly with elevation. The largest range associated with slope throughout the year was found during at the equinox when a range of 20.3 MJ m-2 dy-1 existed between south-facing 60° slopes and northfacing 40° slopes. Turbulent heat flux maps were constructed using modelled Q* maps and observed turbulent flux densities non-dimensionalised by Q*. Maps of the latter were derived from observations over each surface type and surface cover maps, generated using supervised classification of IRS-1C imagery. Large spatial variability in surface energy flux density was found to exist at any given time, although these were greatest near mid-day. Spatial variability was found to be controlled equally by spatial variability of Q* and surface heterogeneity. The role of topographic complexity on the spatial distribution of fluxes was investigated by comparing three. sub-areas of the Tekapo catchment that showed strongly contrasting topographical properties. Increase in topographic complexity was associated with decrease in mean Q*, but a large increase in spatial variability. Similar, but less extreme results were found for turbulent fluxes. However, despite the large ranges in flux densities found throughout the catchment, spatial averages of turbulent fluxes for the three sub-areas were relatively consistent, providing some confidence for the use of spatial averaging for sub-grid areas of numerical models. It is hoped that future research can build on the evidence, methodologies and hypotheses generated by this research. In particular, it is envisioned that future observations in other mountain settings will help determine the extent to which the findings of this study can be applied generally to areas of complex terrain.