Discrete principal‐monotonicity inference for hydro‐system analysis under irregular nonlinearities,data uncertainties,and multivariate dependencies. Part I: methodology development

Hydrological system analyses are challenged by complexities of irregular nonlinearities, data uncertainties, and multivariate dependencies. Among them, the irregular nonlinearities mainly represent inexistence of regular functions for robustly simulating highly complicated relationships between variables. Few existing studies can enable reliable simulation of hydrological processes under these complexities. This may lead to decreased robustness of the constructed models, unfeasibility of suggestions for human activities, and damages to socio‐economy and eco‐environment. In the first of two companion papers, a discrete principal‐monotonicity inference (DPMI) method is proposed for hydrological systems analysis under these complexities. Normalization of non‐normally distributed samples and invertible restoration of modelling results are enabled through a discrete distribution transformation approach. To mitigate data uncertainties, statistical inference is employed to assess the significance of differences among samples. The irregular nonlinearity between the influencing factors (i.e. predictors) and the hydrological variable of interest (i.e. the predictand) is interpreted as piecewise monotonicity. Monotonicity is further represented as principal monotonicity under multivariate dependencies. Based on stepwise classification and cluster analyses, all paired samples representing the responsive relationship between the predictors and the predictand are discretized as a series of end nodes. A prediction approach is advanced for estimating the predictand value given any combination of predictors. The DPMI method can reveal evolvement rules of hydrological systems under these complexities. Reliance of existing hydro‐system analysis methods on predefined functional forms is removed, avoiding artificial disturbances, e.g. empiricism in selecting model functions under irregular nonlinearities, on the modelling process. Both local and global significances of predictors in driving the evolution of hydrological variables are identified. An analysis of interactions among these complexities is also achieved. The understanding obtained from the DPMI process and associated results can facilitate hydrological prediction, guide water resources management, improve hydro‐system analysis methods, or support hydrological systems analysis in other cases. The effectiveness and advantages of DPMI will be demonstrated through a case study of streamflow simulation in Xingshan Watershed, China, in another paper. Copyright © 2016 John Wiley & Sons, Ltd.     

Discrete principal‐monotonicity inference for hydro‐system analysis under irregular nonlinearities,data uncertainties,and multivariate dependencies. Part II: Application to streamflow simulation in the Xingshan Watershed,China

Provision of reliable scientific support to socio‐economic development and eco‐environmental conservation is challenged by complexities of irregular nonlinearities, data uncertainties, and multivariate dependencies of hydrological systems in the Three Gorges Reservoir (TGR) region, China. Among them, the irregular nonlinearities mainly represent unreliability of regular functions for robust simulation of highly complicated relationships between variables. Based on the proposed discrete principal‐monotonicity inference (DPMI) approach, streamflow generation in the Xingshan Watershed, a representative watershed in this region, is examined. Based on system characterization, predictor identification, and streamflow distribution transformation, DPMI parameters are calibrated through a two‐stage strategy. Results indicate that the modelling efficiency of DPMI is satisfactory for streamflow simulation under these complexities. The distribution transformation method and the two‐stage calibration strategy can deal with non‐normality of streamflow and temporally unstable accuracy of hydrological models, respectively. The DPMI process and results reveal that both streamflow uncertainty and its rising tendency increase with flow levels. The dominant driving forces of streamflow generation are daily lowest temperature and daily cumulative precipitation in consideration of performances in global and local scales. The temporal heterogeneity of local significances to streamflow is insignificant for meteorological conditions. There is significant nonlinearity between meteorological conditions and streamflow and dependencies among meteorological conditions. The generation mechanism of low flows is more complicated than medium flows and high flows. The DPMI approach can facilitate improving robustness of hydro‐system analysis studies in the Xingshan Watershed or the TGR region. Copyright © 2016 John Wiley & Sons, Ltd.     

Future changes of temperature and heat waves in Ontario,Canada

Theoretical and Applied Climatology - Apparent changes in the temperature patterns in recent years brought many challenges to the province of Ontario, Canada. As the need for adapting to climate...