Accurate measurements of biophysical parameters are essential for understanding the distribution and dynamics of global vegetation, which exerts an influence on the carbon cycle and atmospheric circulation. Spaceborne, large footprint lidar has been shown to be a valuable tool. It is capable of measuring denser forests than other existing remote methods. However large-footprint lidar struggles to separate ground and canopy signals over topography and in the presence of short vegetation. This prevents the physically-based measurement of forest properties (such as canopy height and cover) at an acceptable accuracy (sub 10 m root mean square error for height) without the use of external data. The necessary external datasets are not yet available at a global scale at high accuracy. In this paper the issues of measuring forests with large-footprint, monochromatic lidar are presented. A number of subtle effects, such as shadows beneath crowns, can hamper the reliable measurement of forests. It is proposed that a dual wavelength lidar will allow the separation of canopy from ground returns in these situations and so allow the physically-based measurement of forests. An initial algorithm is developed and tested with Monte-Carlo ray tracer simulations as a proof of concept. Some refinements are needed to make the method more robust, but the initial form was found to determine the start of the ground return over steep slopes and a range of forest densities, canopy heights and vertical structures with a root mean square error (RMSE) of 2.7 m and mean bias of 67 cm for canopies with covers below 99%. This resulted in canopy height RMSE of 2.88 m with a bias of -23 cm. Such a system will allow measurement of a much broader range of forests than is possible with monochromatic lidar and could form a second generation spaceborne lidar mission.