| In this thesis, a landcover simulator module is developed to incorporate non-stationary landcover into the hydrological model WATFLOOD. Objectives are to quantify the uncertainty inherent in assuming landcover stationarity in the Winnipeg River basin (WRB), and to improve the projections of future streamflow. Forest fires commonplace in the WRB are modelled through logistic regression and a generalized extreme value distribution for occurrence and extent respectively, fit from historical data. Fire regeneration and natural changes in landcover are modelled though a first order Markov chain, with transition probabilities derived from satellite imagery. Using satellite imagery directly into historical simulations in a sub-basin with substantial forest fire activity improved WATFLOOD results. With climate change, incorporating non-stationary landcover results in lower flows than assuming stationarity, albeit still greater than baseline (1971--2000) flows. Projected streamflow uncertainty under climate change also increases as a result of introducing non-stationary landcover in the WRB. |
