Multiobjective analysis of chaotic dynamic systems with sparse learning machines

Sparse learning machines provide a viable framework for modeling chaotic time-series systems. A powerful state-space reconstruction methodology using both support vector machines (SVM) and relevance vector machines (RVM) within a multiobjective optimization framework is presented in this paper. The utility and practicality of the proposed approaches have been demonstrated on the time series of the Great Salt Lake (GSL) biweekly volumes from 1848 to 2004. A comparison of the two methods is made based on their predictive power and robustness. The reconstruction of the dynamics of the Great Salt Lake volume time series is attained using the most relevant feature subset of the training data. In this paper, efforts are also made to assess the uncertainty and robustness of the machines in learning and forecasting as a function of model structure, model parameters, and bootstrapping samples. The resulting model will normally have a structure, including parameterization, that suits the information content of the available data, and can be used to develop time series forecasts for multiple lead times ranging from two weeks to several months.     

A proposed probabilistic seismic tsunami hazard analysis methodology

Natural Hazards - We expand on existing probabilistic tsunami hazard analysis (PTHA) approaches by presenting a methodology wherein the epistemic uncertainties associated with tsunami timing,...     

Environmental changes affecting light climate in oligotrophic mountain lakes: the deep chlorophyll maxima as a sensitive variable

The North-Patagonian Andean lakes of Argentina are high light, low nutrient environments that exhibit development of deep chlorophyll maxima (DCM) at the metalimnetic layer during summer stratification, at approximately 1 % of surface PAR irradiance. We examined whether the position of DCM changes as a consequence of long-time (global warming: glacial clay input) and short-time (eruption: volcanic ashes) events. We performed different field studies: (1) an interlacustrine analysis of six lakes from different basins, including data of the 2011 volcanic eruption, which caused an unexpected variation in water transparency; and (2) an intralacustrine analysis in which we compared different stations along a transparency gradient in Lake Mascardi caused by glacial clay input at one end of the gradient. In these analyses, we documented changes in DCM depth and its relationship with different parameters. DCM development was not related with thermocline depth or nutrient distribution. In all cases, the only significant variables were Kd 320 nm and Kd PAR. Our study showed that suspended particles (glacial clay and volcanic ashes) can play a crucial role in transparent lakes, affecting lake features such as the phototrophic biomass distribution along the water column. Suspended solid inputs from either glacial clay or volcanic ashes produce a comparable effect, provoking a decrease in light and, consequently, an upper location of the DCM. Thus, the DCM position is highly sensitive to global changes, such as increased temperatures causing glacier recession or to regional changes caused by volcanic eruptions.     

Support vector machines (SVMs) for monitoring network design

In this paper we present a hydrologic application of a new statistical learning methodology called support vector machines (SVMs). SVMs are based on minimization of a bound on the generalized error (risk) model, rather than just the mean square error over a training set. Due to Mercer's conditions on the kernels, the corresponding optimization problems are convex and hence have no local minima. In this paper, SVMs are illustratively used to reproduce the behavior of Monte Carlo-based flow and transport models that are in turn used in the design of a ground water contamination detection monitoring system. The traditional approach, which is based on solving transient transport equations for each new configuration of a conductivity field, is too time consuming in practical applications. Thus, there is a need to capture the behavior of the transport phenomenon in random media in a relatively simple manner. The objective of the exercise is to maximize the probability of detecting contaminants that exceed some regulatory standard before they reach a compliance boundary, while minimizing cost (i.e., number of monitoring wells). Application of the method at a generic site showed a rather promising performance, which leads us to believe that SVMs could be successfully employed in other areas of hydrology. The SVM was trained using 510 monitoring configuration samples generated from 200 Monte Carlo flow and transport realizations. The best configurations of well networks selected by the SVM were identical with the ones obtained from the physical model, but the reliabilities provided by the respective networks differ slightly.     

Anthropogenic influences on heavy metals across marine habitats in the western coast of Venezuela

Concentrations of ten metals were measured in waters and sediments at 14 sites during four sampling periods (1996–1997). These sites include various marine ecosystems that are highly influenced by industry, tourism and river discharges, nine of which are within the Morrocoy National Park. Spatially, metal concentrations in water were homogenous, whereas in sediments their distributions were related to grain size. Maximum concentrations of cadmium (Cd) and mean concentrations of copper (Cu), nickel (Ni), lead (Pb) and zinc (Zn) in water were above the guideline values proposed by NOAA, indicating the potential of these metals for producing chronic effects in marine biota. Sheltered sites showed the highest metal concentrations in sediments; with Cd and Zn above these guidelines. Enrichment factors and geo-accumulation indexes suggested that metals in sediments were largely of natural origin except for Pb, Cd and vanadium (V), which were apparently associated with industrial effluents. A disruption of the spatial distribution of metals after heavy rainfall, when exposed sites reached concentrations as high as those in sheltered sites, showed the influence of nearby rivers. The potential increase of such climatic events could represent additional stress for natural protected areas in the Caribbean.