Sensitivity of Simulated Boundary-Layer Flow to the Representation of Forest Canopies in Complex Terrain

Abstract

Complex terrain covers a large fraction of the land surface of the Earth. Due to the limited uses of steeply sloped terrain a lot of this is undeveloped and often covered with a canopy of trees or shrubs. Vegetation canopies are historically modelled using a roughness length at the surface, in line with Monin-Obukhov similarity theory. While this approach has been shown to be effective for shallow canopies such as crops and grasses, for deep canopies such as mature trees this method does not replicate the within and above canopy flows accurately. New approaches were developed in the 1990s to model the canopy explicitly and with a true vertical extent. These models account for the sink of momentum and turbulent kinetic energy within the canopy. In the present work, terms representing these effects are added to these equations in the open-source Weather Research and Forecasting model so that numerical simulations can be performed. This implementation of the canopy model is then tested against a benchmark, idealised case of measurements from a wind-tunnel experiment with an artificial canopy on a ridge with a two-dimensional profile. The model is found to provide significant improvements over simulations using an identical setup but with the canopy parameterised using increased roughness at the surface. The impact of resolution on the effectiveness of the model is also explored, with higher resolutions providing a more accurate representation of the flow over the forested ridge. The model is then applied to a real-world scenario using a case study of the flow over two parallel ridges that form a valley system in the area around Perdig˜ao, Portugal. A 3-hour period during the night is simulated and in this case the canopy model still outperformed the surface roughness method. Many features of the flow were only reproduced properly by simulations using a canopy model. In particular the likelihood of recirculation in the lee of the ridges and the mean flow within the valley. Use of high resolution input data characterising properties and distribution of the canopy did not provide significantly better results than using lower resolution land use datasets with averaged canopy properties. The canopy model is shown to provide significantly better results than the surface roughness parameterisation but does require finer spatial and temporal resolution, leading to a higher computational cost. There is however scope for the implementation of the model that is used here to be improved to avoid instabilities at shorter time steps. When studying the flow over truly complex terrain covered in a canopy, it is difficult to disentangle the effect of the canopy from the effect of the terrain. Further experimental work is therefore suggested that could help to improve the understanding of canopy dynamics in complex terrain and also provide further benchmarks against which canopy models could be tested in numerical simulations. Please note that Chapters 3 and 4 are based on an article co-authored with my supervisor Charles Chemel and Chapter 4 was also co-authored by Robert Menke, who assisted with the processing of the lidar measurements. The author of this thesis carried out all simulations and analyses presented in these chapters as well as writing the text.