Jan Peters

Impact of Reducing Polarimetric SAR Input on the Uncertainty of Crop Classifications Based on the Random Forests Algorithm

Although the use of multidate polarimetric synthetic aperture radar (SAR) data for highly accurate land cover classification has been acknowledged in the literature, the high dimensionality of the data set remains a major issue. This study presents two different strategies to reduce the number of features in multidate SAR data sets: an accuracy-oriented reduction and an efficiency-oriented reduction. For both strategies, the effect of feature reduction on the quality of the land cover map is assessed. The analyzed data set consists of 20 polarimetric features derived from L-band (1.25 GHz) SAR acquired by the Danish EMISAR on four dates within the period April to July in 1998. The predictive capacity of each feature is analyzed by the importance score generated by random forests (RF). Results show that according to the variation in importance score over time, a distinction can be made between general and specific features for crop classification. Based on the importance ranking, features are gradually removed from the single-date data sets in order to construct several multidate data sets with decreasing dimensionality. In the accuracy-oriented and efficiency-oriented reduction, the input is limited to eight and three features per acquisition, respectively. On the reduced input, a multidate model is built using the RF algorithm. Results indicate a decline in the classification uncertainty when feature reduction is performed.

The potential of multitemporal Aqua and Terra MODIS apparent thermal inertia as a soil moisture indicator

Abstract Variations in surface thermal inertia—the resistance to temperature variations—can be indicative for variations in soil moisture. In this paper, we present a flexible multitemporal approach to derive an approximation of thermal inertia, called apparent thermal inertia ( ATI ), from daily Aqua and Terra MODIS observations. In a first step, a varying number of land surface temperature measurements were, together with the time of observation, fit to a sinusoidal function to obtain diurnal surface temperature amplitudes. These were subsequently combined with surface albedo to derive ATI . This was done for the southern part of the African continent for the year 2009. Apparent thermal inertia was compared both spatially and temporally to AMSR-E soil moisture, generated by the algorithm developed by the Vrije Universiteit Amsterdam and NASA. The temporal behaviour of apparent thermal inertia, derived using MODIS data only, showed a strong correspondence to that of AMSR-E soil moisture, especially in arid and semi-arid environments. The approach showed some limitations for vegetated terrains. Further post-processing is required to filter meteorologically induced noise and to transform ATI to actual soil moisture content.

Wetland vegetation distribution modelling for the identification of constraining environmental variables

Wetland ecosystems are of primary concern for nature conservation and restoration. Adequate conservation and restoration strategies emerge from a scientific comprehension of wetland properties and processes. Hereby, the understanding of plant species and vegetation patterns in relation to environmental gradients is an important issue. The modelling approaches in this study statistically relate vegetation patterns to measured environmental gradients in a lowland wetland ecosystem. Measured environmental gradients included groundwater quantity and quality aspects, soil properties and vegetation management. Among this variety, the objective was to identify the key environmental gradients constraining the vegetation, using recently developed methodologies within the modelling approaches. Comparison of results indicated that different environmental gradients were considered to be important by different methodologies.

Absence reduction in entomological surveillance data to improve niche-based distribution models for Culicoides imicola.

Data-driven models for the prediction of bluetongue vector distributions are valuable tools for the identification of areas at risk for bluetongue outbreaks. Various models have been developed during the last decade, and the majority of them use linear discriminant analysis or logistic regression to infer vector-environment relationships. This study presents a performance assessment of two established models compared to a distribution model based on a promising ensemble learning technique called Random Forests. Additionally, the impact of false absences, i.e. data records of suitable vector habitat that are, for various reasons, incorrectly labelled as absent, on the model outcome was assessed using alternative calibration-validation schemes. Three reduction methods were applied to reduce the number of false absences in the calibration data, without loss of information on the environmental gradient of suitable vector habitat: random reduction and stratified reduction based on the distance between absence and presence records in geographical (Euclidean distance) or environmental space (Mahalanobis distance). The results indicated that the predicted vector distribution by the Random Forest model was significantly more accurate than the vector distributions predicted by the two established models (McNemar test, p<0.01) when the calibration data were not reduced with respect to false absences. The performance of the established models, however, increased considerably by application of stratified false absence reductions. Model validation revealed no significant difference between the performance of the three distinct Culicoides imicola distribution models for the majority of alternative stratified reduction schemes. The main conclusion of this study is that the application of Random Forests, or linear discriminant analysis and logistic regression on the condition that calibration data were first reduced on geographical or environmental information, potentially lead toward better vector distribution models.

Synergy of very high resolution optical and radar data for object-based olive grove mapping

This study investigates the potential of very high resolution (VHR) optical and radar data for olive grove landscape mapping. VHR data were fed into a four-step processing chain performing an object-based land-use classification. The four steps included (i) image segmentation, (ii) object feature calculation, (iii) object-based classification and (iv) land-use map evaluation. First, the optical (ADS40) and radar (RAMSES SAR and TerraSAR-X) data were applied to the processing chain separately. As supported by two segmentation evaluation measures, the stand purity index (PI) and the potential mapping accuracy (PMA), the optical data thereby led to a significantly better segmentation and a more accurate olive cover map (Kruskal–Wallis test, ). Second, synergy models were developed combining data from the different sensors at different stages of the object-based classification process, namely, (1) during the segmentation step, (2) during the feature calculation step and (3) after the object classification step. The combined use of features from the different sensors resulted in a considerable improvement in mapping accuracy, with correctly classified objects supported by high probabilities. The assessment of feature importance revealed that optical data were most important for successful object-based olive grove mapping; however, features related to object shape and texture of the radar imagery added to its success. Comparison of the object-based synergy model with a pixel-based synergy model indicated a limited classification improvement. This research showed that the integrated use of VHR optical and radar data is appropriate in an object-based classification framework, leading towards more accurate olive grove landscape mapping.

Influence of topographic normalization on the vegetation index–surface temperature relationship

The estimation of surface soil moisture status and evapotranspiration from optical remote sensing using the vegetation index–surface temperature ( VI - T S ) relationship is severely hampered in regions with strong topography, due to the influence of altitude and terrain orientation on surface temperature. In our study, a new empirical approach to normalize surface temperature for terrain elevation—a stratified linear regression model—is presented and is applied on moderate-resolution imaging spectroradiometer (MODIS) data over Calabria, Italy. The method incorporates remotely sensed land surface temperature, a vegetation index, and a digital elevation model. The influence of the newly developed normalization on the VI - T S relationship and on a soil dryness index is compared to the influence of two existing normalization methods: one using a standard lapse rate of 0.65 K per 100 m and one using a lapse rate derived through simple linear regression between elevation and surface temperature. Stratified linear regression adequately corrects surface temperature while the two other normalization techniques seem to overestimate the actual temperature lapse rate during certain periods of the year. Comparison of a soil dryness index derived using the three different normalization methods with limited in situ soil moisture data results in a slightly stronger correlation for the stratified linear regression model than for the two other normalization methods. VI - T S –based soil wetness estimation in mountainous terrains remains, however, limited by other spatially varying factors, including terrain orientation and atmospheric conditions.

Modelling the Spatial Distribution of Culicoides imicola: Climatic versus Remote Sensing Data

Culicoides imicola is the main vector of the bluetongue virus in the Mediterranean Basin. Spatial distribution models for this species traditionally employ either climatic data or remotely sensed data, or a combination of both. Until now, however, no studies compared the accuracies of C. imicola distribution models based on climatic versus remote sensing data, even though remotely sensed datasets may offer advantages over climatic datasets with respect to spatial and temporal resolution. This study performs such an analysis for datasets over the peninsula of Calabria, Italy. Spatial distribution modelling based on climatic data using the random forests machine learning technique resulted in a percentage of correctly classified C. imicola trapping sites of nearly 88%, thereby outperforming the linear discriminant analysis and logistic regression modelling techniques. When replacing climatic data by remote sensing data, random forests modelling accuracies decreased only slightly. Assessment of the different variables’ importance showed that precipitation during late spring was the most important amongst 48 climatic variables. The dominant remotely sensed variables could be linked to climatic variables. Notwithstanding the slight decrease in predictive performance in this study, remotely sensed datasets could be preferred over climatic datasets for the modelling of C. imicola. Unlike climatic observations, remote sensing provides an equally high spatial resolution globally. Additionally, its high temporal resolution allows for investigating changes in species’ presence and changing environment.

POSSIBILITIES FOR ECOHYDROLOGICAL MONITORING IN NATURAL AND MANAGED ECOSYSTEMS IN SOUTHERN CHILE

La ecohidrologia se enfoca sobre las vinculaciones entre las plantas y el ambiente abiotico a partir del ciclo hidrologico. Como estas interacciones mutuas son importantes en muchos ecosistemas, la ecohidrologia tiene una amplia aplicacion. Un aspecto importante de la investigacion en ecohidrologia es la evaluacion y prediccion de la presencia de especies vegetales o tipos vegetacionales en relacion con la hidrologia o condiciones hidrogeoquimicas del habitat. Para ello, diversos modelos han sido publicados (por ejemplo, DEMNAT, ICHORS, ITORS, ITORS-VL). Todos estos modelos necesitan datos. Consecuentemente una red de monitoreo se tiene que instalar en adicion a las bases de datos existentes. Estas redes incluyen el monitoreo de propiedades importantes de los ecosistemas tales como: profundidad de la zona saturada, conductividad electrica del agua subterranea, contenido del agua del suelo, caracteristicas quimicas del agua del suelo (pH, concentraciones ionicas y nutrientes) y composicion de la vegetacion. El modelo ITORS-VL ha sido probado satisfactoriamente para predecir la ocurrencia de comunidades de plantas en los ecosistemas pantanosos de Flanders monitoring and modelling: Peters, J. et al. (Belgica), usando regresiones basadas en variables ambientales que van desde variables hidrologicas como profundidad de la zona saturada a variables quimicas que caracterizan el agua del suelo via concentracion de nitratos y conductividad electrica. Sin embargo, debido a su naturaleza empirica para un ecosistema especifico, este modelo podria no ser adecuado para los ecosistemas en Chile. El presente trabajo investiga la posibilidad de implementar un monitoreo y modelamiento en diferentes ecosistemas del sur de Chile. Los beneficios de esta actividad son variados: (i) una comprension holistica del funcionamiento del ecosistema, (ii) la habilidad para determinar la importancia relativa de las caracteristicas del sitio en varios ecosistemas, y (iii) la posibilidad de predecir estados sucesionales de la vegetacion bajo condiciones ambientales variables. Estos beneficios presentan potenciales roles en la capacidad de gestion, restauracion de ecosistemas y evaluacion del riesgo o impacto ambiental