Adriaan van Niekerk

Predicting runoff-induced pesticide input in agricultural sub-catchment surface waters: linking catchment variables and contamination

An urgent need exists for applicable methods to predict areas at risk of pesticide contamination within agricultural catchments. As such, an attempt was made to predict and validate contamination in nine separate sub-catchments of the Lourens River, South Africa, through use of a geographic information system (GIS)-based runoff model, which incorporates geographical catchment variables and physicochemical characteristics of applied pesticides. We compared the results of the prediction with measured contamination in water and suspended sediment samples collected during runoff conditions in tributaries discharging these sub-catchments. The most common insecticides applied and detected in the catchment over a 3-year sampling period were azinphos-methyl (AZP), chlorpyrifos (CPF) and endosulfan (END). AZP was predominantly found in water samples, while CPF and END were detected at higher levels in the suspended particle samples. We found positive (p < 0.002) correlations between the predicted average loss and the concentrations of the three insecticides both in water and suspended sediments (r between 0.87 and 0.94). Two sites in the sub-catchment were identified as posing the greatest risk to the Lourens River mainstream. It is assumed that lack of buffer strips, presence of erosion rills and high slopes are the main variables responsible for the high contamination at these sites. We conclude that this approach to predict runoff-related surface water contamination may serve as a powerful tool for risk assessment and management in South African orchard areas.

An evaluation of supervised classifiers for indirectly detecting salt-affected areas at irrigation scheme level

Abstract Soil salinity often leads to reduced crop yield and quality and can render soils barren. Irrigated areas are particularly at risk due to intensive cultivation and secondary salinization caused by waterlogging. Regular monitoring of salt accumulation in irrigation schemes is needed to keep its negative effects under control. The dynamic spatial and temporal characteristics of remote sensing can provide a cost-effective solution for monitoring salt accumulation at irrigation scheme level. This study evaluated a range of pan-fused SPOT-5 derived features (spectral bands, vegetation indices, image textures and image transformations) for classifying salt-affected areas in two distinctly different irrigation schemes in South Africa, namely Vaalharts and Breede River. The relationship between the input features and electro conductivity measurements were investigated using regression modelling (stepwise linear regression, partial least squares regression, curve fit regression modelling) and supervised classification (maximum likelihood, nearest neighbour, decision tree analysis, support vector machine and random forests). Classification and regression trees and random forest were used to select the most important features for differentiating salt-affected and unaffected areas. The results showed that the regression analyses produced weak models (

Random Forest (RF) Wrappers for Waveband Selection and Classification of Hyperspectral Data

Hyperspectral data collected using a field spectroradiometer was used to model asymptomatic stress in Pinus radiata and Pinus patula seedlings infected with the pathogen Fusarium circinatum. Spectral data were analyzed using the random forest algorithm. To improve the classification accuracy of the model, subsets of wavebands were selected using three feature selection algorithms: (1) Boruta; (2) recursive feature elimination (RFE); and (3) area under the receiver operating characteristic curve of the random forest (AUC-RF). Results highlighted the robustness of the above feature selection methods when used in conjunction with the random forest algorithm for analyzing hyperspectral data. Overall, the Boruta feature selection algorithm provided the best results. When discriminating F. circinatum stress in Pinus radiata seedlings, Boruta selected wavebands (n = 69) yielded the best overall classification accuracies (training error of 17.00%, independent test error of 17.00% and an AUC value of 0.91). Classification results were, however, significantly lower for P. patula seedlings, with a training error of 24.00%, independent test error of 38.00%, and an AUC value of 0.65. A hybrid selection method that utilizes combinations of wavebands selected from the three feature selection algorithms was also tested. The hybrid method showed an improvement in classification accuracies for P. patula, and no improvement for P. radiata. The results of this study provide impetus towards implementing a hyperspectral framework for detecting stress within nursery environments.

Temporal patterns of animal-related traffic accidents in the Eastern Cape, South Africa

Abstract This study analysed road accident data for a five-year period on a route between Uitenhage and Graaff-Reinet in the Eastern Cape (South Africa) to investigate the temporal patterns of animal-related accidents (ARAs). Although a large number of ARAs was recorded during the high-traffic holiday season, a distinct period of relatively high ARA rates was identified from April to July. These months coincide with the rutting and hunting seasons of greater kudu (Tragelaphus strepsiceros) when movement is more pronounced. The observed daily danger times for ARAs are from dusk to around midnight and early mornings until dawn. Weekends, with increased motor vehicle traffic, are also hazardous times. These results should be used by traffic authorities and regional traffic managers to help initiate mitigating procedures and launch public awareness programmes to inform motorists of the most accident-prone times of the day, days of the week and seasons of the year for animal-related motor vehicle accidents.

Climate and the evolution of group-living behaviour in the armadillo lizard ( Ouroborus cataphractus)

We evaluated the hypothesis that the regular use of the southern harvester termite, Microhodotermes viator, as food source by the armadillo lizard, Ouroborus cataphractus, originated as an adaptation to survive the summer dry season in a climatic regime where rainfall is highly seasonal. To do so, we determined the most important climatic predictors of the geographical range of this species. Climatic data were obtained for 130 localities where O. cataphractus is known to occur and 168 adjacent localities where it is known to be absent. For each locality, data for 10 climatic variables were extracted from the South African Atlas of Agrohydrology and Climatology database. We constructed a forward stepwise logistic regression model of the probability of O. cataphractus occurrence, based on the set of 10 climatic variables. The best model included, in order of importance, average monthly summer rainfall, mean annual precipitation, average monthly solar radiation, and the ratio of winter rainfall over summer rainfall as most significant predictors. The selected model predicted 88.80% of the presences correctly and 85.52% of the absences. In essence, O. cataphractus is restricted to the winter rainfall zone of South Africa, excluding the high-rainfall southwestern section. We postulate that the highly predictable seasonal rainfall and the ameliorating effect of the Atlantic Ocean on climates in the Namaqualand region, in particular, have provided a unique selective regime for the origin of group-living in O. cataphractus. Dependence on M. viator as food source developed to survive the summer-autumn period of low food availability and resulted in the evolution of heavy armour and group-living behaviour. The moderate winters and early spring temperatures allowed full capitalization on high arthropod abundance during winter—spring, thereby overriding the negative impacts of armour and groupliving on foraging efficiency at the home crevice.

CLUES: A web-based land use expert system for the Western Cape

In this research, a web-based spatial decision support system (SDSS) was developed to demonstrate how the Internet can be used to deliver low-cost, user-friendly and interactive land use decision support to a wide audience. Although the resulting Cape Land Use Expert System (CLUES) was specifically developed to perform land suitability analysis in the Western Cape, the technology can also be applied to other regions and modified for other applications. CLUES consists of five components: a land unit database (LUD), knowledge base, inference engine, web map service (WMS) and graphical user interface (GUI). The LUD consists of polygons (land units) and attributes (land properties), while the knowledge base stores each users land use requirement rules. These rules are used by the inference engine to rate the suitability of each land unit in the LUD. The result is then mapped by the WMS and presented to the users as suitability maps. Users can direct the entire analysis through a user-friendly GUI. The development and demonstration of CLUES exposed several advantages and limitations of current web technology and has demonstrated that the Internet offers great opportunities for the deployment of land use suitability information to a wide audience.

Monitoring sustainable urban development using built-up area indicators: a case study of Stellenbosch, South Africa

AbstractRapid urbanisation in many developing countries causes land transformation from agricultural, rural, and natural landscapes into urban areas. Data to monitor this transformation are often out of date, unreliable, not in standard format, cumbersome and expensive to collect or simply unavailable. This inhibits local authorities and other stakeholders’ capacity to monitor and leverage resources towards sustainable urban development. This paper investigates the use of earth observation (EO) data for supporting sustainable urban development planning. The study demonstrates that EO adds value to sustainable urban development by providing area-wide and up-to-date thematic and geometric characterisation of the urban built-up area, which would be difficult to obtain from other data sources. This helps local planning authorities to monitor urban growth and sustainability, and facilitate evidence-based decision-making and an array of other practical uses.

Climate and the Presence of Generation Glands in Female Girdled Lizards: A Case Study of the cordylus-niger-oelofseni Complex

Abstract Geographic variation in the presence/absence of generation glands in female lizards of the cordylus-niger-oelofseni complex was described and correlations with climatic variables investigated. Preserved and live specimens from 95 localities in the area south of 32°30′S and west of 19°45′E in the Western Cape, South Africa, were examined for the presence or absence of generation glands. A GIS analysis was performed to determine mean annual minimum temperature, mean annual maximum temperature, potential evaporation, mean annual precipitation, mean annual fog, and mean annual cloud cover for each locality. Discriminant function and canonical analyses showed highly significant correlations between presence and absence of generation glands and five of the six climatic variables. Females from western coastal localities and from the Cape Fold Mountains generally lacked generation glands, whereas females from inland lowland localities generally possessed glands.

Value of dimensionality reduction for crop differentiation with multi-temporal imagery and machine learning

Abstract This study evaluates the use of automated and manual feature selection – prior to machine learning – for the differentiation of crops in a Mediterranean climate (Western Cape, South Africa). Five Landsat-8 images covering the different crop class phenological stages were acquired and used to generate a range of spectral and textural features within an object-based image analysis (OBIA) paradigm. The features were used as input to decision trees (DTs), k-nearest neighbour (k-NN), support vector machine (SVM), and random forest (RF) supervised classifiers. Testing was done by performing classifications (using all spatial variables) and then incrementally reducing the feature counts (based on importance allocated to features by filters), feature extraction, and manual (semantic) feature selection. Classification and regression trees (CART) and RF were used as methods to filter feature selection. Feature-extraction methods employed include principal components analysis (PCA) and Tasselled cap transformation (TCT). The classification results were analysed by comparing the overall accuracies and kappa coefficients of each scenario, while McNemar’s test was used to assess the statistical significance of differences in accuracies among classifiers. Feature selection was found to improve the overall accuracies of the DT, k-NN, and RF classifications, but reduced the accuracy of SVM. The results showed that SVM with feature extraction (PCA) on individual image dates produced the most accurate classification (96.2%). Semantic groupings of features for classification also revealed that using the image bands and indices is not sufficient for crop classification, and that additional features are needed. The accuracy differences of the classifiers were, however, not statistically significant, which suggests that, although dimensionality reduction can improve crop differentiation when multi-temporal Landsat-8 imagery is used, it had a marginal effect on the results. For operational crop-type classification in the study area (and similar regions), we conclude that the SVM algorithm can be applied to the full set of features generated.

Investigating the Utility of Oblique Tree-Based Ensembles for the Classification of Hyperspectral Data

Ensemble classifiers are being widely used for the classification of spectroscopic data. In this regard, the random forest (RF) ensemble has been successfully applied in an array of applications, and has proven to be robust in handling high dimensional data. More recently, several variants of the traditional RF algorithm including rotation forest (rotF) and oblique random forest (oRF) have been applied to classifying high dimensional data. In this study we compare the traditional RF, rotF, and oRF (using three different splitting rules, i.e., ridge regression, partial least squares, and support vector machine) for the classification of healthy and infected Pinus radiata seedlings using high dimensional spectroscopic data. We further test the robustness of these five ensemble classifiers to reduced spectral resolution by spectral resampling (binning) of the original spectral bands. The results showed that the three oblique random forest ensembles outperformed both the traditional RF and rotF ensembles. Additionally, the rotF ensemble proved to be the least robust of the five ensembles tested. Spectral resampling of the original bands provided mixed results. Nevertheless, the results demonstrate that using spectral resampled bands is a promising approach to classifying asymptomatic stress in Pinus radiata seedlings.