Stochastic nonparametric methods for multi-site weather generation and flood frequency estimation: Applications to construction delay, hydrology and agricultural modeling

Stochastic nonparametric approaches for functional estimation and simulation are gaining wide recognition. These methods are data-driven, in that, they do not make any prior assumption of the underlying functional form, unlike traditional methods. This thesis has two main components - development of (i) an improved weather generator and, (ii) a conditional flood frequency estimation method. The weather generator is integrated with three diverse applications (i) quantifying construction delay, (ii) crop simulation and (iii) watershed streamflow simulation. The ability to generate realistic weather scenarios is important for improved resources management. Nonparametric stochastic weather generators based on k–nearest neighbor time series resampling tend to under-simulate the wet and dry spell statistics, which are important for a variety of applications. To alleviate this, a semiparametric multivariate weather generator (SWG) model is proposed. The proposed approach has two steps: (1) A Markov Chain for generating the precipitation state (i.e. , no rain, rain, or heavy rain), and (2) A k-Nearest Neighbor ( k-NN) bootstrap re-sampler for generating the multivariate weather variables. The Markov Chain captures the spell statistics while the k-NN bootstrap captures the distributional and lag-dependence statistics of the weather variables. We demonstrated the utility of the proposed approach and its improvement over the traditional k-NN approach, through an application to daily weather data from Pergamino in the Pampas region of Argentina. We show the applicability of the proposed framework in simulating weather scenarios conditional on the seasonal climate forecast and also at multiple sites in the Pampas region. The SWG model was coupled to three varied applications. First, this was used to generate weather sequences and consequently, attributes of weather important in better planning for construction delays. With applications to three different locations we demonstrated potential improvements in construction delay management. This comprehensive integrated framework is a major contribution in construction management. Second, the SWG model was coupled to crop simulation system to generate crop yield scenarios based on seasonal climate forecast and also future decadal climate variability. This is an important framework for efficient and sustainable agriculture management. Thirdly, the SWG model was integrated with a watershed model to generate streamflow scenarios, useful for flood and drought management. These three diverse applications make a unique and significant contribution. A fully nonparametric conditional flood frequency estimation approach based on local polynomials was developed. This approach has the ability to generate flood frequency and risk estimates based on large scale climate predictors—important for flood and drought management. The stochastic methods for weather generation and flood frequency estimation and their various applications make a useful and significant contribution.