Monthly suspended sediment load prediction using artificial intelligence: testing of a new random subspace method

ABSTRACT

The predictive capability of a new artificial intelligence method, random subspace (RS), for the prediction of suspended sediment load in rivers was compared with commonly used methods: random forest (RF) and two support vector machine (SVM) models using a radial basis function kernel (SVM-RBF) and a normalized polynomial kernel (SVM-NPK). Using river discharge, rainfall and river stage data from the Haraz River, Iran, the results revealed: (a) the RS model provided a superior predictive accuracy (NSE = 0.83) to SVM-RBF (NSE = 0.80), SVM-NPK (NSE = 0.78) and RF (NSE = 0.68), corresponding to very good, good, satisfactory and unsatisfactory accuracies in load prediction; (b) the RBF kernel outperformed the NPK kernel; (c) the predictive capability was most sensitive to gamma and epsilon in SVM models, maximum depth of a tree and the number of features in RF models, classifier type, number of trees and subspace size in RS models; and (d) suspended sediment loads were most closely correlated with river discharge (PCC = 0.76). Overall, the results show that RS models have great potential in data poor watersheds, such as that studied here, to produce strong predictions of suspended load based on monthly records of river discharge, rainfall depth and river stage alone.     

Calibration of diatoms along a nutrient gradient in Florida Everglades Water Conservation Area-2A, USA

The relationship between diatom taxa preserved in surface soils and environmental variables at 31 sites in Water Conservation Area 2A (WCA-2A) of the Florida Everglades was explored using multivariate analyses. Surface soils were collected along a phosphorus (P) gradient and analyzed for diatoms, total P, % nitrogen (N), %carbon (C), calcium (Ca), and biogenic silica (BSi). Phosphorus varied from 315-1781 g g^-1, and was not found to be correlated with the other geochemical variables. Canonical correspondence analysis (CCA) was used to examine which environmental variables correlated most closely with the distributions in diatom taxa. Canonical correspondence analysis with forward selection, constrained and partial CCA, and Monte Carlo permutation tests of significance show the most significant changes in diatom assemblages along the P gradient (p < 0.01), with additional species differences correlated with soil C, N, Ca, and BSi.Weighted-averaging (WA) regression and calibration models of diatom assemblages to P and BSi were developed. The diatom-based inference model for soil [P] had a high apparent r² (0.86) with RMSE_boot = 218 g g^-1. Indicator diatom species identified by assessing species WA optima and WA tolerance to [P], such as Nitzschia amphibia and N. palea for high [P] (~1300-1400 g^-1) and Achnanthes minutissima var. scotica and Mastogloia smithii for low [P] (~400-600 g g^-1), may be useful as monitoring tools for eutrophication in WCA-2A as well as other areas of the Everglades. Diatom assemblages analyzed by cluster analysis were related to location within WCA-2A, and dominant taxa within clusters are discussed in relation to the geochemical variables measured as well as hydrology and pH. Diversity of diatom assemblages and a Disturbance Index based on diatom data are discussed in relation to the historically P-limited Everglades ecosystem. Diatom assemblages should be very useful for reconstructions of [P] through time in the Florida Everglades, provided diatoms are well preserved in soil cores.     

Marine policy and disarmament

Disarmament has again become a major public concern in many nations. The main manifestations of this development have been demonstrations and marches and mass media attention to the possible consequences of a nuclear war. Less publicized has been governmental work, including that at the United Nations, in negotiating disarmament treaties.     

Generation of complete theoretical seismograms for SH—

Summary. We report the initial results of our attempts to obtain theoretical seismograms for direct comparison with the experimental time series obtained with the long-period instruments of the WWSSN. The entire theoretical seismogram, including both body waves and surface waves, can be generated for a spherical, anelastic earth by simple inverse Fourier transformation of the sum of the propagating fundamental and higher-mode surface waves. The key to success in reproducing the WWSSN records involves the number of modes, and the minimum period used in these computations; here we use eight modes and a minimum period of 2 s. Efficient computational algorithms make it possible to handle up to 2000 frequency points for each mode; approximately 200 layers are used to model the radial heterogeneity of the earth; attenuation is treated exactly. Examples are given of the SH theoretical seismograms resulting from dislocation sources buried at various depths in the Earth.     

Influence of uncertain boundary conditions and model structure on flood inundation predictions

In this study, the GLUE methodology is applied to establish the sensitivity of flood inundation predictions to uncertainty of the upstream boundary condition and bridges within the modelled region. An understanding of such uncertainties is essential to improve flood forecasting and floodplain mapping. The model has been evaluated on a large data set. This paper shows uncertainty of the upstream boundary can have significant impact on the model results, exceeding the importance of model parameter uncertainty in some areas. However, this depends on the hydraulic conditions in the reach e.g. internal boundary conditions and, for example, the amount of backwater within the modelled region. The type of bridge implementation can have local effects, which is strongly influenced by the bridge geometry (in this case the area of the culvert). However, the type of bridge will not merely influence the model performance within the region of the structure, but also other evaluation criteria such as the travel time. This also highlights the difficulties in establishing which parameters have to be more closely examined in order to achieve better fits. In this study no parameter set or model implementation that fulfils all evaluation criteria could be established. We propose four different approaches to this problem: closer investigation of anomalies; introduction of local parameters; increasing the size of acceptable error bounds; and resorting to local model evaluation. Moreover, we show that it can be advantageous to decouple the classification into behavioural and non-behavioural model data/parameter sets from the calculation of uncertainty bounds.     

Towards medium-term (order of months) morphodynamic modelling of the Teign estuary, UK

The main objective of this paper is to address the principal mechanisms involved in the medium-term (order of months to years) morphodynamic evolution of estuaries through the application of a process-based numerical modelling. The Teign estuary (Teignmouth, UK) is the selected site. The system is forced by the macrotidal semi-diurnal tide in the English Channel and is perturbed to a minor extent by high river discharge events (freshets). Although waves have a definite influence on the adjacent coastal area, Wells (Teignmouth Quay Development Environmental Statement: Changes to Physical Processes. Report R.984c:140. ABP Marine Environmental Research Ltd., Southampton, 2002b) suggested that swell waves do not enter the estuary. Hence, wave effects are neglected in this study, as only tides and the river discharge are taken into account. The sediment grain size is highly variable, but mainly sandy. Within the frame of the COAST3D project (), four bathymetric surveys of the adjacent coastal area were carried out at a nearly weekly intervals. The outer estuary and the adjacent coastal area were also surveyed every 6 months as part of the COASTVIEW project (). Based on these data and on continuously measured parameters, such as water level, waves, wind and river discharge, numerical modelling of the morphodynamic processes can be tested. To replicate the morphological changes in the medium-term within a feasible simulation time, forcing conditions are reduced through the use of an input reduction method (called ensemble technique). In this study, simulations are based on the coupling between Telemac-2D and its non-cohesive sediment transport module, Sisyphe (version 5.3 for both modules). Three different sediment transport formulae were tested: (1) Engelund and Hansen (A monograph on sediment transport in alluvial streams, 3rd edn. Technological University of Denmark, Copenhagen, 1967) including the modifications proposed by Chollet and Cunge (J Hydraul Eng 17(1):1–13, 1979); (2) Bijker (Mechanics of sediment transport by the combination of waves and current. In: Design and reliability of coastal structures. 23rd international conference on Coastal Engineering, pp 147–173, 1968) and (3) Soulsby (Dynamics of Marine Sands. A manual for practical applications. HR Wallingford, Wallingford, p 142, 1997) modified version of van Rijn [J Hydraul Eng 110(10):1431–1456, 1984a, J Hydraul Eng 110(11):1613–1641, 1984b] formulation. Both a qualitative (i.e. visual comparison) and a quantitative tool [Brier Skill Score (BSS); described in Sutherland et al. in Coast Eng 51:917–939, 2004b] are applied to assess the similarity of simulations when compared to model predictions and observations. Tests confirmed the reliability and time efficiency of the ensemble technique, since it reproduced very well the results of a reference run, a computation based on the observed boundary conditions. For the spring-neap cycle modelled, the BSS was of 0.91 (a perfect modelling would have a BSS of 1), with a reduction in the simulation time on the order of 80%. For the 6-month-period simulation, results were also excellent: BSS=0.92 and a computer time reduction of 85%. In principle, this method has the advantage of being applied to any process-based numerical model.     

Surface simulation studies of the hydration of white rust Fe(OH)2, goethite α-FeO(OH) and hematite α-Fe2O3

Computer modelling techniques were used to elucidate the hydration behaviour of three iron (hydr)oxide minerals at the atomic level: white rust, goethite and hematite. A potential model was first adapted and tested against the bulk structures and properties of eight different iron oxides, oxyhydroxides and hydroxides, followed by surface simulations of Fe(OH)₂, α-FeO(OH) and α-Fe₂O₃. The major interaction between the adsorbing water molecules and the surface is through interaction of their oxygen ions with surface iron ions, followed by hydrogen-bonding to surface oxygen ions. The energies released upon the associative adsorption of water range from 1 to 17 kJ mol⁻¹ for Fe(OH)₂, 26 to 80 kJ mol⁻¹ for goethite and 40 to 85 kJ mol⁻¹ for hematite, reflecting the increasing oxidation of the iron mineral. Dissociative adsorption at goethite and hematite surfaces releases larger hydration energies, ranging from 120 to 208 kJ mol⁻¹ for goethite and 76 to 190 kJ mol⁻¹ for hematite.The thermodynamic morphologies of the minerals, based on the calculated surface energies, agree well with experimental morphologies, where these are available. When the partial pressures required for adsorption of water from the gas phase are plotted against temperature for the goethite and hematite surfaces, taking into account experimental entropies for water, it appears that these minerals may well be instrumental in the retention of water during the cyclic variations in the atmosphere of Mars.     

Analysis of I in groundwater samples: Direct and quantitative results below the drinking water standard

This paper highlights an analytical method based on mass measurement that can be used to directly quantify ¹²⁹I in groundwater samples at concentrations below the maximum contaminant level (MCL) without the need for sample pre-concentration or extraction. Samples were analyzed on a Perkin Elmer ELAN DRC II ICP-MS after minimal dilution using O₂ as the reaction gas. Analysis of continuing calibration verification standards indicated that the dynamic reaction cell (DRC) mode could be used for quantitative analysis of ¹²⁹I in samples below the MCL (0.0057 ng/mL or 1 pCi/L). The low analytical detection limit of ¹²⁹I analysis in the DRC mode coupled with minimal sample dilution (1.02x) resulted in a final estimated quantification limit of 0.0051 ng/mL. Subsequent analysis of three groundwater samples containing ¹²⁹I resulted in fully quantitative results in the DRC mode, and spike recovery analyses performed on all three samples confirmed that the groundwater matrix did not adversely impact the analysis of ¹²⁹I in the DRC mode. This analytical approach has been proven to be a cost-effective, high-throughput technique for the direct, quantitative analysis of ¹²⁹I in groundwater samples at environmentally relevant concentrations that reach below the current MCL.     

Risk analysis of possible impacts of climate change on South Australian wheat production

This study presents a model-based risk assessment of wheat production under projected climate change by 2080 in eight locations of South Australia. The vulnerability of wheat production under future climate change was quantitatively evaluated via a risk analysis in which the identification of critical yield thresholds applies. Research results show that risk (conditional probability of not exceeding the critical yield thresholds) increased more or less across all locations under the most likely climate change. Wheat production in drier areas such as Minnipa, Orroroo and Wanbi will not be economically viable under the most likely climate change. Intensive studies on adaptation are now required.