Search

Block Kriging (a spatial interpolation method) and log10 transformation were compared for their effectiveness in reducing relative variance (coefficient of variance: CV) and estimate mean values in all harvested maize plants grown in three randomly taken field plots and for harvested plants after removing the "edge or margin" ones. The results showed that log10 transformation reduced CVs of all harvested original fresh weight (FW) plant data in the three plots from 35.6–41.6% (original data) to 6.0–7.5%, while the respective CVs due to Block Kriging were reduced to 14.5–19.9%. The back-log10-transformed means of all harvested FW plant data were reduced by 6.8–9.4%, while the respective reduction for plants excluding the margin ones was 1.3–8.3%. The Block Kriging means for all harvested FW plant data were reduced only by 0.3–0.4%, while the respective means of the harvested plants excluding margin ones were increased by 0.4–4.3%. These findings strongly suggest that Block Kriging should be preferred over the log10 transformation method (used so far by agroscientists) as it managed to effectively reduce variability in crop data and estimate missing values that provide more precise and reliable estimates of corn yield for farmers. [ABSTRACT FROM AUTHOR]

A novel Self-Organizing Neuro-Fuzzy Multilayered Classifier, the GA-SONeFMUC model, is proposed in this paper for land cover classification of multispectral images. The model is composed of generic fuzzy neuron classifiers (FNCs) arranged in layers, which are implemented by fuzzy rule-based systems. At each layer, parent FNCs are combined to generate a descendant FNC at the next layer with higher classification accuracy. To exploit the information acquired by the parent FNCs, their decision supports are combined using a fusion operator. As a result, a data splitting is devised within each FNC, distinguishing those pixels that are currently correctly classified to a high certainty grade from the ambiguous ones. The former are handled by the fuser, while the ambiguous pixels are further processed to enhance their classification confidence. The GA-SONeFMUC structure is determined in a self-constructing way via a structure-learning algorithm with feature selection capabilities. The parameters of the models obtained after structure learning are optimized using a real-coded genetic algorithm. For effective classification, we formulated three input sets containing spectral and textural feature types. To explore information coming from different feature sources, we apply a classifier fusion approach at the final stage. The outputs of individual classifiers constructed from each input set are combined to provide the final assignments. Our approach is tested on a lake-wetland ecosystem of international importance using an IKONOS image. A high-classification performance of 92.02% and of 75.55% for the wetland zone and the surrounding agricultural zone is achieved, respectively. Index Terms--Classifier fusion, fusion operators, genetic algorithms (GAs), image classification, neuro-fuzzy classifiers, remote sensing.

Several organizations provide satellite Leaf Area Index (LAI) data regularly, at various scales, at high frequency, but at low spatial resolution. This study attempted to enhance the spatial resolution of the MODIS LAI product to the Landsat resolution level. Four climatically diverse sites in Europe and Africa were selected as study areas. Regression analysis was applied between MODIS Enhanced Vegetation Index (EVI) and LAI data. The regression equations were used as input in a downscaling model, along with Landsat EVI images and land-cover maps. The estimated LAI values showed high correlation with field-measured LAI during the dry period. The model validation gave statistically significant results, with correlation coefficient values ranging from relatively low (0.25–0.32), to moderate (0.48–0.64) and high (0.72–0.94). Limited samples per vegetation type, the diversity of species within the same vegetation type, land-use/land-cover changes and saturated EVI values affected the accuracy of the downscaling model. [ABSTRACT FROM AUTHOR]

Remote sensing techniques are routinely used in plant species discrimination and of weed mapping. In the presented work, successful Silybum marianum detection and mapping using multilayer neural networks is demonstrated. A multispectral camera (green-red-near infrared) attached on a fixed wing unmanned aerial vehicle (UAV) was utilized for the acquisition of high-resolution images (0.1 m resolution). The Multilayer Perceptron with Automatic Relevance Determination (MLP-ARD) was used to identify the S. marianum among other vegetation, mostly Avena sterilis L. The three spectral bands of Red, Green, Near Infrared (NIR) and the texture layer resulting from local variance were used as input. The S. marianum identification rates using MLP-ARD reached an accuracy of 99.54%. The study had an one year duration, meaning that the results are specific, although the accuracy shows the interesting potential of S. marianum mapping with MLP-ARD on multispectral UAV imagery. [ABSTRACT FROM AUTHOR]

In the present study, the detection and mapping of Silybum marianum (L.) Gaertn. weed using novelty detection classifiers is reported. A multispectral camera (green-red-NIR) on board a fixed wing unmanned aerial vehicle (UAV) was employed for obtaining high-resolution images. Four novelty detection classifiers were used to identify S. marianum between other vegetation in a field. The classifiers were One Class Support Vector Machine (OC-SVM), One Class Self-Organizing Maps (OC-SOM), Autoencoders and One Class Principal Component Analysis (OC-PCA). As input features to the novelty detection classifiers, the three spectral bands and texture were used. The S. marianum identification accuracy using OC-SVM reached an overall accuracy of 96%. The results show the feasibility of effective S. marianum mapping by means of novelty detection classifiers acting on multispectral UAV imagery. [ABSTRACT FROM AUTHOR]

Despite playing a critical role in the division of precipitation between runoff and infiltration, soil moisture (SM) is difficult to estimate at the catchment scale and at frequent time steps, as is required by many hydrological, erosion and flood simulation models. In this work, an integrated methodology is described to estimate SM at the root zone, based on the remotely-sensed evaporative fraction (L) and ancillary information on soil and meteorology. A time series of Terra MODIS satellite images was used to estimate SM maps with an eight-day time step at a 250-m spatial resolution for three diverse catchments in Europe. The study of the resulting SM maps shows that their spatial variability follows the pattern of land cover types and the main geomorphological features of the catchment, and their temporal pattern follows the distribution of rain events, with the exception of irrigated land. Field surveys provided in situ measurements to validate the SM maps' accuracy, which proved to be variable according to site and season. In addition, several factors were analyzed in order to explain the variation in the accuracy, and it was shown that the land cover type, the soil texture class, the temporal difference between the datasets' acquisition and the presence of rain events during the measurements played a significant role, rather than the often referred to scale difference between in situ and satellite observations. Therefore, the proposed methodology can be used operationally to estimate SM maps at the catchment scale, with a 250-m spatial resolution and an eight-day time step. [ABSTRACT FROM AUTHOR]

Monitoring soil erosion risk is an important part of soil conservation practices. It is usually estimated with the Universal Soil Loss Equation, and the C-factor (vegetation cover) is derived from optical satellite images. However, because of lack of data and resources, or in rapid assessments, C-factor is estimated using one or a few satellite observations, despite being temporally variable according to plants' phenology. The aim of this work was to study the effect of seasonality in estimating C-factor. This was achieved by demonstrating first that there is a difference when estimating soil erosion with Universal Soil Loss Equation at variable time steps in a year, namely once, seasonally and monthly. Using Moderate Resolution Imaging Spectroradiometer normalized difference vegetation index images and statistical analysis at subcatchment scale, it was shown that there is a significant difference when estimating mean annual soil loss with the aforementioned temporal options. The highest differences were observed between monthly and annual time steps. The second objective was to identify which is the optimum time to estimate C-factor in a year. The results show that November, October and March are the optimum months for single image estimation of annual soil erosion. Statistical analysis with a random point dataset suggested that the spatial variability of the results was influenced by the land cover type, especially in areas with variable leaf cover where a single date estimation of C-factor was not representative of the whole year, such as annual crops and deciduous trees. Copyright © 2013 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]