Approximating SWAT Model Using Artificial Neural Network and Support Vector Machine1

Abstract: With the popularity of complex, physically based hydrologic models, the time consumed for running these models is increasing substantially. Using surrogate models to approximate the computationally intensive models is a promising method to save huge amounts of time for parameter estimation. In this study, two learning machines [Artificial Neural Network (ANN) and support vector machine (SVM)] were evaluated and compared for approximating the Soil and Water Assessment Tool (SWAT) model. These two learning machines were tested in two watersheds (Little River Experimental Watershed in Georgia and Mahatango Creek Experimental Watershed in Pennsylvania). The results show that SVM in general exhibited better generalization ability than ANN. In order to effectively and efficiently apply SVM to approximate SWAT, the effect of cross‐validation schemes, parameter dimensions, and training sample sizes on the performance of SVM was evaluated and discussed. It is suggested that 3‐fold cross‐validation is adequate for training the SVM model, and reducing the parameter dimension through determining the parameter values from field data and the sensitivity analysis is an effective means of improving the performance of SVM. As far as the training sample size, it is difficult to determine the appropriate number of samples for training SVM based on the test results obtained in this study. Simple examples were used to illustrate the potential applicability of combining the SVM model with uncertainty analysis algorithm to save efforts for parameter uncertainty of SWAT. In the future, evaluating the applicability of SVM for approximating SWAT in other watersheds and combining SVM with different parameter uncertainty analysis algorithms and evolutionary optimization algorithms deserve further research.     

Analysis of the cost-effectiveness for ecosystem service provision and rural income generation: a comparison of three different programs in Southern Ecuador

In recent years, new tools for funding nature conservation have been designed. Because poverty is often significant in areas with high biodiversity, the improvement of local livelihoods is frequently considered as a secondary goal of new financing mechanisms besides nature conservation. The buffer zone of the Podocarpus National Park in Ecuador is such a high biodiversity zone. In this paper, we compare the cost-effectiveness and development potential of three different mechanisms to finance nature conservation implemented in this buffer zone, namely (a) an organic coffee label, (b) the Socio Bosque Program, a nationwide payment scheme for private forest conservation, and (c) FORAGUA, a regional water fund. This paper describes the functioning and the scope of the mechanisms and analyses their environmental and socio-economic impacts which are compared to the total costs. Results show that the water fund has the highest additionality in ecosystem service provision, while the payment scheme is the most cost-effective both for current as for increased ecosystem service provision and for extra rural job creation. Organic coffee certification has the highest positive impact on rural income creation.     

Association patterns among male and female spotted hyenas (Crocuta crocuta) reflect male mate choice

Although female animals tend to be choosier than males in selecting mates, sexual selection theory predicts that males should also be choosy when female fecundity varies. Reproductive success among female spotted hyenas varies greatly with social rank. Our goals were therefore to determine whether male hyenas preferentially associate with high-ranking females, and whether male preferences are affected by female reproductive state. Interactions between adult males and females were observed intensively, and association indices calculated for all male-female pairs, over a 7-year period in one population of free-living hyenas. Males initiated most affiliative interactions with females, and males associated most closely with females that were likeliest to be fertile. High- and middle-ranking males associated most closely with high-ranking females, but low-ranking males associated equally closely with females in all rank categories. We used molecular markers to determine the paternity of cubs born during the study period, and found that sires associated more closely with the mothers of those cubs than did non-sires, particularly during the last months before conception. These association data indicate that male spotted hyenas do indeed exhibit selective mate choice, and that they prefer females likeliest to maximize male reproductive success.     

Dynamics of calling by tree swallow (Tachycineta bicolor) nestmates

The begging display of nestling passerine birds has become a model for examining the evolution of animal signals. A particular problem for nestlings when transmitting begging signals to parents may be interference from nestmates. The strategies used by nestlings to reduce signal interference have not been studied, yet potentially contribute to the design of these complex displays. In this study, we recorded the begging calls of nestling tree swallows (Tachycineta bicolor) when alone and with a begging nestmate, to determine whether nestlings changed the output, structure or timing of their calls in ways that would reduce acoustic interference. We found that nestlings increased their call rate in the presence of a begging nestmate, but did not alter the length, amplitude or frequency of their calls. They also appeared not to adjust the timing of their calls to avoid those of nestmates. Contrary to expectation, nestling calls became more similar in some aspects when nestmates called together. An increase in call rate in the presence of a begging nestmate should increase the likelihood that a parent detects an individual's calls. However, if all nestlings increase their calling rate in response to competitors, then the overall level of acoustic interference across the brood is potentially increased, an effect exacerbated by the tendency for call similarity to increase when calling together. We discuss how increasing call rate may improve detectability despite this effect and we also examine how an increase in rate and call similarity may serve to produce a strong brood signal.     

Frequency-dependent male mate harassment and intra-specific variation in its avoidance by females of the damselfly Ischnura elegans

We focused on male harassment on different female color morphs of the damselfly Ischnura elegans and on variation in morph-specific mating avoidance tactics by females. In I. elegans, one of the female morphs is colored like the conspecific male (andromorphs) while the other morphs are not (gynomorphs). Our first goal was to quantify morph-specific male mating attempts, hence male harassment, in populations with manipulated population parameters (densities, sex ratios, and proportion of andromorphs). Second, we examined the female's perspective by looking for potential differences in morph-specific mating avoidance tactics and success of those tactics in a natural population. Differences in population conditions did influence the number of male mating attempts per morph. The less frequent female morph was always subject to fewer mating attempts, which contradicts earlier hypotheses on mimicry, but supports those that assume that males learn to recognize female morphs. Gynomorphs occupy less open habitat and often fly away when a male approaches, while andromorphs use more open habitat, do not fly large distances and directly face approaching males. Female morphs did not differ in the proportion of successful mating-avoidance attempts. Our results suggest that the maintenance of the color polymorphism is most probably the result of interactive selective forces depending on variation in all population conditions, instead of solely density- or frequency-dependent selection within populations.     

Are Nitrogen-Fertilized Forest Soils Sinks or Sources of Carbon?

We developed a simple conceptual model that tracks nitrogen and carbon jointly through an N fertilized forest ecosystem. The stimulation of growth increases the litterfall and imports substrate for soil microorganisms. Microbial biomass forms according to the supply of C and N. The formation of microbial biomass is accompanied by respiratory C losses. The quantity of CO₂ efflux depends on the C use efficiency of microbes. When excess N is available, the microbial activity is accelerated and the demand for substrate is high. Litterfall supplies an insufficient amount of C to the soil. In such a case, labile soil C is mineralized and the net effect of N fertilization is a loss of soil C. A strong N fertilization effect on the aboveground biomass can offset the soil C loss. In the case of a low N dosage or high N losses due to leaching or emission of nitrogen oxides, the soil C loss is small. The conceptual model was applied to a case study. The field data, collected over a time span of several decades, could not support sound conclusions on the temporal trend of soil C because the spatial and temporal variability of the chemical data was high. The conceptual model allowed to give an evaluation of the fertilization effect on soil C based on reproducible principles.     

The potential of biogas production from municipal solid waste in a tropical climate

The objective of this study was to estimate the potential of organic municipal solid waste generated in an urban setting in a tropical climate to produce biogas. Five different categories of wastes were considered: fruit waste, food waste, yard waste, paper waste, and mixed waste. These fractions were assessed for their efficiency for biogas production in a laboratory-scale batch digester for a total period of 8 weeks at a temperature of 15–30 °C. During this period, fruit waste, food waste, yard waste, paper waste, and mixed waste were observed to produce 0.15, 0.17, 0.10, 0.08, and 0.15 m³ of biogas per kilogram of volatile solids, respectively. The biogas produced and caloric value of each feedstock was in the range of 1.25?×?10^?3 m³ (17 kWh)/cap/day (paper waste) to 15?×?10^?3 m³ (170 kWh)/cap/day (mixed waste). Paper waste produced the least (<1×10^?3(<17.8 kWh)/cap/day), and mixed waste produced the highest methane yield (10?×?10^?3 m³ (178 kWh)/cap/day). Thus, mixed waste was found to be more efficient than other feedstocks for biogas and methane production; this was mainly related to the better C/N ratio in mixed waste. Taking the total waste production in Jimma into account, the total mixed organic solid waste could produce 865?×?10³ m³ (5.4 m³/capita) of biogas or 537?×?10³ m³ (3.4 m³/capita) of methane per year. The total caloric value of methane production potential from mixed organic municipal solid waste was many times higher than the total energy requirement of the area.     

Sampling and analytical methodology development for the determination of primary and secondary low molecular weight amines in ambient air

In this project, a sampling device and an analytical method have been developed to simultaneously analyse the most frequently found low molecular weight amines, including aliphatic, aromatic and alcohol amines. These amines are diethanolamine, ethanolamine, methylamine, isopropylamine, morpholine, dimethylamine, and aniline. A sampling device was developed using a 37 mm cassette with glass fibre filters impregnated with sulfuric acid. Immediately after sampling, the filter was transferred to vials containing a solution of dansyl chloride. Dansyl chloride was used for derivatisation because it forms aromatic sulfonamides that are fluorescent and easy to protonate for MS detection. The effect of using an internal standard made with the dansylated derivative of 1-(2-methoxyphenyl)piperazine (MOPIP) on the uncertainty and efficiency of the method was also evaluated. This internal standard was spiked directly onto filters. The coupling of HPLC/ESI-MS was used for the simultaneous analysis of all the derivatives. This method showed detection limits of about 0.03 microg mL(-1) to 0.3 microg mL(-1) of amine with an average expanded uncertainty of 3% to 6% depending on the amine. The methodology recoveries are close to 100% for all the amines, and the overall estimated expanded uncertainties vary between 10% and 13% depending on the amine. This new strategy will be useful in evaluating workplace air since a unique sampling system will be used, independent of the amine to be quantified.     

Metal speciation in sediment and bioaccumulation in <Emphasis Type="Italic">Meretrix lyrata</Emphasis> in the Tien Estuary in Vietnam

The concentrations of seven toxic metals (cadmium (Cd), nickel (Ni), chromium (Cr), arsenic (As), lead (Pb), copper (Cu), and zinc (Zn)) were determined in sediments and the soft tissues of a bivalve species (Meretrix lyrata) collected from the Tien Estuary in Tien Giang Province, South Vietnam. The total metal concentrations in sediments (mg/kg dry weight) increased as Cd (0.06) < Cu (5.0) < Pb (13.9) < As (16.3) < Ni (24) < Cr (50) < Zn (62). Speciation analysis revealed that these metals existed mainly in the residual fraction (43–94%), followed by the Fe-Mn oxide-bound (5–35%) and organic/sulfide-bound (0.6–9.2%) fractions. The metal concentrations in M. lyrata (mg/kg dry weight) were in the ranges of 1.3–1.9 (Cd), 1.5–2.8 (Ni), 1.8–3.4 (Cr), 11–16 (As), 0.3–0.6 (Pb), 6.9–8.7 (Cu), and 95–128 (Zn), which are safe for human consumption. The order of the mean biota-sediment accumulation factor (BSAF) of the metals in the non-residual fractions of the sediment for M. lyrata was Cd > Cu > As > Zn > Cr > Ni > Pb. The Risk Assessment Codes (RACs) suggest that the highest mobility of Cd (with RAC = 37%) poses greater environmental risk to aquatic biota. Correlation analysis results show that M. lyrata can be used as a biomonitor of Cd and Cu pollution in the exchangeable, acid-soluble, and non-residual sediment fractions.     

Assessment of human exposure to benzene through foods from the Belgian market

Benzene is a volatile organic compound known to be carcinogenic to humans (Group 1) and may be present in food. In the present study, 455 food samples from the Belgian market were analyzed for benzene contents and some possible sources of its occurrence in the foodstuffs were evaluated. Benzene was found above the level of detection in 58% of analyzed samples with the highest contents found in processed foods such as smoked and canned fish, and foods which contained these as ingredients (up to 76.21 μg kg⁻¹). Unprocessed foods such as raw meat, fish, and eggs contained much lower concentrations of benzene. Using the benzene concentrations in food, a quantitative dietary exposure assessment of benzene intake was conducted on a national representative sample of the Belgian population over 15 years of age. The mean benzene intake for all foods was 0.020 μg kg bw d⁻¹ according to a probabilistic analysis. These values are below the minimum risk level for oral chronic exposure to benzene (0.5 μg kg bw d⁻¹).