A Tool for Classification and Regression Using Random Forest Methodology: Applications to Landslide Susceptibility Mapping and Soil Thickness Modeling

Classification and regression problems are a central issue in geosciences. In this paper, we present Classification and Regression Treebagger (ClaReT), a tool for classification and regression based on the random forest (RF) technique. ClaReT is developed in Matlab and has a simple graphic user interface (GUI) that simplifies the model implementation process, allows the standardization of the method, and makes the classification and regression process reproducible. This tool performs automatically the feature selection based on a quantitative criterion and allows testing a large number of explanatory variables. First, it ranks and displays the parameter importance; then, it selects the optimal configuration of explanatory variables; finally, it performs the classification or regression for an entire dataset. It can also provide an evaluation of the results in terms of misclassification error or root mean squared error. We tested the applicability of ClaReT in two case studies. In the first one, we used ClaReT in classification mode to identify the better subset of landslide conditioning variables (LCVs) and to obtain a landslide susceptibility map (LSM) of the Arno river basin (Italy). In the second case study, we used ClaReT in regression mode to produce a soil thickness map of the Terzona catchment, a small sub-basin of the Arno river basin. In both cases, we performed a validation of the results and a comparison with other state-of-the-art techniques. We found that ClaReT produced better results, with a more straightforward and easy application and could be used as a valuable tool to assess the importance of the variables involved in the modeling.     

Environmental and biological monitoring of methyl ethyl ketone (MEK)

This study was conducted to evaluate the usefulness of various biological parameters for monitoring of workers exposed to methyl ethyl ketone (MEK). Fifty male workers from a large magnetic videotape factory participated in this study. Personal air samples were collected using 3M organic vapor monitors and analysed for MEK by gas chromatography with flame ionisation detector (FID). 10 mL of urine; blood (1 mL) and exhaled air were also collected at the end of an 8-hour workshift. The headspace GC method was applied for measurement of urinary and blood MEK. MEK in expired air was analysed directly by using a GC/FID.The correlation coefficients (r) between environmental MEK and all other biological parameters measured show significant positive relationships. The r for environmental MEK and urine MEK was 0.84; for blood 0.73 and for breath 0.64. The correlation coefficients between blood and urine was 0.72; blood and breath was 0.88 and urine and breath 0.60. These findings suggest that measurements of unmetabolised MEK in blood, exhaled air and urine can be used for biological monitoring of MEK exposure. Nevertheless, laboratory methodological assessment is in favour of measuring urinary MEK as it is non-invasive and does not have to be analysed immediately after collection.     

Cadmium in naturally-occurring water hyacinths

Water hyacinth [$\text{Eichhornia}$$\text{crassipes}$ (Mart.) Solms] is an effective fertillizer, mulch, soil amendment and fodder for cattle and sheep. The consequences of the ability of the plant to accumulate heavy metals needs consideration. Small and large water hyacinth ecotypes collected from two “rural” rivers in Florida, the Peace and Withalachoochee respectively, were examined for an accumulation of cadmium from the natural environment. Less than trace amounts (below limits of detection, 2.5ppb) were present. Water hyacinths tested for cadmium accumulation under laboratory conditions exhibited a pronounced ability to absorb and accumulate the metal. The roots contained 0.982 mg of the 2.0 mg cadmium presented to the plant. The distribution of the metal within the plant was 76.7% in the roots, 22.4% in the stems with the remainder in the leaves. Of the total cadmium presented to the plant, 80.8 ± 3.0% was recovered by the plant after 7 days, the mean generation time of the plant. With increasing urbanization, accumulation of cadmium by water hyacinths exists. It is either an asset or a potential hazard and needs to be appreciated.     

Further studies using X-ray fluorescence to sample lead contaminated carpeted surfaces

The sampling of carpeted surfaces to test for lead contamination primarily focuses upon vacuum techniques. Vacuum sampling techniques, however, require time-consuming, expensive laboratory analysis of the dusts obtained and are unable to determine total lead load on the carpet. X-ray fluorescence (XRF) analysis is an on-site, inexpensive, non-destructive, quick technique for predicting metals levels in a variety of media, such as water, soil, filter paper and painted surfaces. A 1992 study of the feasibility of XRF to analyze for lead and soil loadings on carpeted surfaces indicated that XRF can detect lead at a low enough level to warrant further study. This paper expands this earlier study and developes lead and soil loading calibration curves for three different carpet types based upon XRF lead L-beta peak areas and XRF iron and barium K-alpha peak and background areas. Results indicate that variation in the data can be reduced through modifications of the XRF analysis technique, thus reducing the statistically determined detection level, and that carpet type does affect the calibration. Detection levels of approximately 70 mg/m² for lead and 5 g/m² for soil were obtained. Overall, good agreement was found between results of this study and the earlier one. XRF shows excellent potential for quantitative analysis of lead on carpeted surfaces.     

Microbial degradation of mefenoxam in rhizosphere of Zinnia angustifolia

The fate of the fungicide mefenoxam was studied in a containerized rhizosphere system. The rhizosphere system used Zinnia angustifolia (Tropic Snow) in a bark/sand potting mix and was compared to bulk potting mix (no plants). Rhizosphere microbial populations were allowed to establish for 3 weeks prior to fungicide addition (20 microg per g mix). Mefenoxam and degradation product concentrations were determined by High HPLC or capillary electrophoresis after extraction. Seventy eight percent of the fungicide originally applied to the rhizosphere was degraded after 21 days compared to 44% in bulk system (no plant). The primary degradation product was the free acid N-(2,6-dimethylphenyl)-N-(methoxyacetyl)-DL-alanine, which accounted for 71% of the applied parent chemical after 30 days. N-(2,6-dimethylphenyl)-acetamide was also detected, but in lesser amounts. Bacterial populations in the rhizosphere increased during the 30-day period, which correlated with an increase in degradation of the parent compound. Pure cultures of Pseudomonas fluorescens and Chrysobacterium indologenes isolated from the rhizosphere system could degrade the applied fungicide (10 microg/ml) almost completely to the free acid within 54 h.     

Biological Effects of Carbon Monoxide vs. Whole Emissions from Engine with Catalyst

The introduction of oxidizing catalytic converters in the exhaust system of automobiles has been found to be effective in reducing carbon monoxide, total hydrocarbons, and some other components of tailpipe emissions. The utilization of these converters, however, has also caused further oxidation of sulfur compounds producing an increased amount of sulfur trioxide (SO₃) which yields sulfuric acid and sulfates.¹ Studies on the biological effects of these sulfur compounds are in progress in our laboratory and preliminary findings are reported separately. This paper compares the biological effects of exposure to whole emissions from engines with and without converter and exposure to carbon monoxide alone.     

Runoff and sediment losses from 27 upland catchments in Southeast Asia: Impact of rapid land use changes and conservation practices

Rapid changes in upland farming systems in Southeast Asia generated predominantly by increased population pressure and ‘market forces’ have resulted in widespread land degradation that has been well documented at the plot scale. Yet, the links between agricultural activities in the uplands and downstream off-site effects remain largely unknown because of the difficulties in transferring results from plots to a larger scale. Many authors have thus pointed out the need for long-term catchment studies. The objective of this paper is to summarize the results obtained by the Management of Soil Erosion Consortium (MSEC) over the last 5 years from 27 catchments in five countries (Indonesia, Laos, Philippines, Thailand, and Vietnam). The purpose of the study was to assess the impacts of cultivation practices on annual runoff and erosion rates. Initial surveys in each catchment included topography, soils and land use. Monitoring included climatic, hydrologic and erosion (total sediment yield including bed load and suspended sediment load) data, land use and crop yields, and farmers’ income. In addition, new land management options were introduced through consultations with farmers and evaluated in terms of runoff and erosion. These included tree plantations, fruit trees, improved fallow with legumes, maize intercropped with legumes, planted fodder, native grass strips and agro-ecological practices (direct sowing and mulch-based conservation agriculture). Regressions analyses showed that runoff during the rainy season, and normalized runoff flow coefficient based on erosive rainfall during the rainy season (rainfall with intensity exceeding 25 mm h⁻¹) increase with the percentage of the catchment covered by maize. Both variables decrease with increasing soil depth, standard deviation of catchment slope (that reflects terrain roughness), and the percentages of the catchment covered by fallow (regular and improved), tree plantations and planted fodder. The best predictors of sediment yield were the surface percentages of maize, Job's tears, cassava and footpaths. The main conclusions generated from this study were: (i) soil erosion is predominantly influenced by land use rather than environmental characteristics not only at the plot scale but also at the catchment scale; (ii) slash-and-burn shifting cultivation with sufficiently long rotations (1 year of cultivation, 8 years of fallow) is too often unjustly blamed for degradation; (iii) in its place, continuous cropping of maize and cassava promotes high rates of soil erosion at the catchment scale; (iv) conservation technologies are efficient in reducing runoff and total sediment yield at the catchment scale; (v) the adoption of improved soil management technologies by upland farmers is not a function of the degree of intensification of their farming system and/or of their incomes. The results suggest that if expansion of maize and cassava into already degraded upland systems were to occur due to increased demand for biofuels, there is a risk of higher runoff and sediment generation. A failure to adopt appropriate land use management strategies will result in further rapid resource degradation with negative impacts to downstream communities.     

Measuring and assessing indicators of sustainable development for Scotland: a pilot survey

In this paper, we review three broad classes of indicators of sustainable development. These are economic, ecological and sociopolitical indicators. We then report the results of a pilot study which measured five different indicators for Scotland: an approximate environmentally-adjusted Net National Product; the Pearce-Atkinson Measure; Net Primary Production; Appropriated Carrying Capacity; and the Index of Sustainable Economic Welfare. We find that not all indicators point in the same direction, but a majority suggest that Scotland is at best only marginally sustainable. Data limitations, however, suggest a degree of caution in the interpretation of these results.     

Exclusion of citrullinaemia in the first trimester of pregnancy by direct assay of argininosuccinate synthetase in chorionic villi

Citrullinaemia was presumed to be excluded in a fetus at risk by the direct assay of argininosuccinate synthetase in chorionic villi. The diagnosis was confirmed after amniocentesis by normal argininosuccinate synthetase activity in the cultured amniotic fluid cells and by a normal citrulline concentration in the amniotic fluid. The prediction of a normal fetus was confirmed at term by the birth of a non-citrullinaemic boy.     

First-trimester prenatal diagnosis of Sanfilippo C disease

At 10 weeks' gestation, chorionic villus biopsy was obtained from a woman at risk for Sanfilippo type C disease. Acetyl-CoA: α-glucosaminide N-acetyltransferase activity was markedly deficient. The pregnancy was terminated at 12 weeks' gestation and follow-up study on fetal fibroblasts confirmed the diagnosis.