Laboratory for
Climatology and Remote Sensing

Abstract:

Fog is a weather phenomenon that affects all of us, both in the business and private sectors. On the one hand it is vital for entire ecosystems worldwide, on the other hand it can lead to immense economic damage (airport etc.) and even life-threatening situations, for example in road traffic. The positive as well as the negative implications demonstrate the urgency of a good understanding of fog events and their prediction. Fog forecasting has been a subject of research for many years. However, due to the complex interactions of a wide variety of variables relevant for fog formation, accurate prediction still remains very difficult. One reason for this is that the simulation of parameter interactions which are difficult to capture by modelling is affected by errors due to the complexity of the system. Another reason is the neglected but most likely important combination of temporal and spatial variables, which is currently not taken into account by the models. This is mainly due to the lack of suitable approaches for the combination of temporally high-resolution station measurements of fog-relevant variables and area-wide high-resolution satellite data.

Abstract:

Fog forecasting still shows large inaccuracies in accurately predicting fog formation, dissipation and duration. Since a few years, Machine learning (ML) algorithms are increasingly used in addition to numerical fog forecasts because of their computational speed and ability to learn non-linear interactions between the variables. Due to their black-box nature, precise and accurate training and evaluation is vital to prevent insufficient training or meaningless scores. Three main points important for fog prediction are explained in the following. 1. Fog forecasting datasets consist of autocorrelated variables. In most cases, there is an information leakage between the training and test data sets which are used to evaluate the model performance. This information leakage can have an impact on the performance scores because the stronger the information flow, the easier it is for the model to memorize. 2. Fog forecasting datasets have a temporal order. To be able to make statements about the performance of an operational model this temporal order should already be simulated during model training and evaluation. This is because for an operational model, the training data points are always older than the data points to be predicted. Commonly used training methods neglect this fact. 3. Time series used for fog forecasting usually have a large imbalance between the frequency of the fog class and non-fog class. This imbalance can have an unfavorable interaction with the confusion matrix based meteorological scores that are widely used for evaluation. All of the aforementioned points, if not considered, can lead to an insufficient forecast without even being noticed. Therefore, the negative influence on the model score of two commonly used training methods that neglect the points named above will be shown using an XGBoost model and a logistic regression model. In comparison, a training and evaluation method was evaluated that maintains the temporal order and thus simulates the performance of an operational model. It will also be shown that common meteorological scores, since they are computed based on a confusion matrix, share a weakness when the data set is unbalanced: Persistence behavior remains undetected. The study is funded by the DFG research project “FOG-ML FOrecasting radiation foG by combining station and satellite data using Machine Learning”.

Abstract:

Large inaccuracies still exist in accurately predicting fog formation, dissipation, and duration. To improve these deficiencies, machine learning (ML) algorithms are increasingly used in nowcasting in addition to numerical fog forecasts because of their computational speed and their ability to learn the nonlinear interactions between the variables. Although a powerful tool, ML models require precise training and thoroughly evaluation to prevent misinterpretation of the scores. In addition, a fog dataset’s temporal order and the autocorrelation of the variables must be considered. Therefore, classification-based ML related pitfalls in fog forecasting will be demonstrated in this study by using an XGBoost fog forecasting model. By also using two baseline models that simulate guessing and persistence behavior, we have established two independent evaluation thresholds allowing for a more assessable grading of the ML model’s performance. It will be shown that, despite high validation scores, the model could still fail in operational application. If persistence behavior is simulated, commonly used scores are insufficient to measure the performance. That will be demonstrated through a separate analysis of fog formation and dissipation, because these are crucial for a good fog forecast. We also show that commonly used blockwise and leave-many-out cross-validation methods might inflate the validation scores and are therefore less suitable than a temporally ordered expanding window split. The presented approach provides an evaluation score that closely mimics not only the performance on the training and test dataset but also the operational model’s fog forecasting abilities.

Abstract:

This paper describes a method of low-altitude remote sensing in combination with in situ measurements (leaf area, spectroscopy, and position) to monitor the postfire canopy recovery of two competing grassland species. The method was developed in the Andes of Ecuador, where a tethered balloon with a digital camera was deployed to record a time series of very high spatial resolution imagery (nominal resolution=2 cm ) of an experimental plot covered by two competing species: 1) the pasture grass, Setaria sphacelata; and 2) the invasive southern bracken, Pteridium arachnoideum. Image processing techniques were combined to solve geometric issues and construct high-quality mosaics for image classification. The semiautomatic and object-oriented classification method was based on geometrical and textural attributes of image segments and showed promising results for detecting the invasive bracken fern in Setaria pastures (performance by area under the curve, AUC = 0.88). Valuable insights are given into vegetation monitoring applications using unmanned aerial vehicles, which produces a time series of species-specific maps, including foliage projective cover (FPC) and leaf area index (LAI). This new method constitutes an important and accessible tool for ecological investigations of competing species in pastures and validation of remote sensing information on mountain environments.

Abstract:

The aim of this study was to proof applicability of hyperspectral imaging for the analysis and classification of human mucosal surfaces in vivo. The larynx as a prototypical anatomically well-defined surgical test area was analyzed by microlaryngoscopy with a polychromatic lightsource and a synchronous triggered monochromatic CCD-camera. Image stacks (5 benign, 7 malignant tumors) were analyzed by established software (principal component analysis PCA, hyperspectral classification, spectral profiles). Hyperspectral image datacubes were analyzed and classified by conventional software. In PCA, images at 590–680 nm loaded most onto the first PC which typically contained 95% of the total information. Hyperspectral classification clustered the data highlighting altered mucosa. The spectral profiles clearly differed between the different groups. Hyperspectral imaging can be applied to mucosal surfaces. This approach opens the way to analyze spectral characteristics of histologically different lesions in order to build up a spectral library and to allow non-touch optical biopsy.