Laboratory for
Climatology and Remote Sensing

Abstract:

There is a growing demand to train Earth Observation (EO) data users in how to access and use existing and upcoming data. A promising tool for data-related training is computational notebooks, which are interactive web applications that combine text, code and computational output. Here, we present the Learning Tool for Python (LTPy), which is a training course (based on Jupyter notebooks) on atmospheric composition data. LTPy consists of more than 70 notebooks and has taught over 1000 EO data users so far, whose feedback is overall positive. We adapted five guiding principles from different fields (mainly scientific computing and Jupyter notebook research) to make the Jupyter notebooks more educational and reusable. The Jupyter notebooks developed (i) follow the literate programming paradigm by a text/code ratio of 3, (ii) use instructional design elements to improve navigation and user experience, (iii) modularize functions to follow best practices for scientific computing, (iv) leverage the wider Jupyter ecosystem to make content accessible and (v) aim for being reproducible. We see two areas for future developments: first, to collect feedback and evaluate whether the instructional design elements proposed meet their objective; and second, to develop tools that automatize the implementation of best practices.

Abstract:

The current goal of our research is to improve the machine learning (ML) based fog forecast for a forecasting period up to four hours for the station in Linden-Leihgestern. The prediction of radiation fog is still subject to large uncertainties. In particular, the precise prediction of fog start and dissipation, i.e. the transitions, is very difficult. The high-frequency fluctuations of the variables in the formation and dissipation phases pose a particular challenge to ML models. These strong fluctuations make it difficult to extract the necessary information about the past, namely increasing or decreasing trend. However, the temporal evolution in the past is determining for the development of radiation fog. Thus, these dynamics must be prepared in such a way that they can be learned during model training. Therefore, different smoothing levels were tested using a Gaussian moving average filter. Furthermore, additional trend variables for model training were generated that carry information about the temporal evolution of previous data points. Training and evaluation have been carried out with the Expanding Window Approach (Vorndran et al. 2022) that has recently been accepted as a training and validation method for radiation fog prediction. Building on this scheme with the tree-based algorithm XGBoost, the newly preprocessed data led to an improvement in the prediction of radiation fog for the station in Linden-Leihgestern. The results from this research will be presented in the poster session. 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:

Bioturbators shape their environment with considerable consequences for ecosystem processes. However, both the composition and the impact of bioturbator communities may change along climatic gradients. For burrowing animals, their abundance and composition depend on climatic and other abiotic components, with ants and mammals dominating in arid and semiarid areas, and earthworms in humid areas. Moreover, the activity of burrowing animals is often positively associated with vegetation cover (biotic component). These observations highlight the need to understand the relative contributions of abiotic and biotic components in bioturbation in order to predict soil-shaping processes along broad climatic gradients. In this study, we estimated the activity of animal bioturbation by counting the density of holes and the quantity of bioturbation based on the volume of soil excavated by bioturbators along a gradient ranging from arid to humid in Chile. We distinguished between invertebrates and vertebrates. Overall, hole density (no/ 100 m2) decreased from arid (raw mean and standard deviation for invertebrates: 14 ± 7.8, vertebrates: 2.8 ± 2.9) to humid (invertebrates: 2.8 ± 3.1, vertebrates: 2.2 ± 2.1) environments. However, excavated soil volume did not follow the same clear geographic trend and was 300-fold larger for vertebrates than for invertebrates. The relationship between bioturbating invertebrates and vegetation cover was consistently negative whereas for vertebrates both, positive and negative relationships were determined along the gradient. Our study demonstrates complex relationships between climate, vegetation and the contribution of bioturbating invertebrates and vertebrates, which will be reflected in their impact on ecosystem functions.

Abstract:

Pastoralism is the main land use in the humid tropical Andes of South America. Wide areas of mountain rainforest have been cleared for gaining pastureland. Due to the lack of indigenous useful grasses in the pristine forests, mainly exotic grass species have been used for establishing the pastures. In the Ecuadorian Andes, Axonopus compressus, Melinis minutiflora, Pennisetum clandestinum and Holcus lanatus are common pasture grass species. Their preference for certain microsites resulted in a mosaic of different pasture types, which reflect the differing ecological conditions on the undulating terrain of the mountain slopes. During the last decades, however, another exotic grass species, Setaria sphacelata has widely been introduced which, because of its fast growth on some of the sites could successfully suppress the formerly dominant plant species. With respect to the changing microclimate and cattle stocking rates the present study explored, whether planting Setaria is the best option for the common low-input type of pasture farming in these tropical mountains. In a study over twenty years, the development of four main pasture types, dominated by the above-mentioned grass species was investigated in areas with and without Setaria, and their topographical occurrence on the sloping terrain was analyzed. On forty-eight plots a pairwise (with or without Setaria) comparison of species composition and diversity, biomass production, forage quality and soil properties was performed. Although Setaria grows faster than the other grass species, its productivity was only higher on flat terrain. The nutritive value of the Setaria plots was at best equivalent to that of the former pastures, while species richness was consistently lower. Our results suggest the maintenance of a terrain-adapted diversification of the pastures and in particular the use of Setaria only on flat terrain.

Abstract:

Heatwaves exert disproportionately strong and sometimes irreversible impacts on forest ecosystems. These impacts remain poorly understood at the tree and species level and across large spatial scales. Here, we investigate the effects of the record-breaking 2018 European heatwave on tree growth and tree water status using a collection of high-temporal resolution dendrometer data from 21 species across 53 sites. Relative to the two preceding years, annual stem growth was not consistently reduced by the 2018 heatwave but stems experienced twice the temporary shrinkage due to depletion of water reserves. Conifer species were less capable of rehydrating overnight than broadleaves across gradients of soil and atmospheric drought, suggesting less resilience toward transient stress. In particular, Norway spruce and Scots pine experienced extensive stem dehydration. Our high-resolution dendrometer network was suitable to disentangle the effects of a severe heatwave on tree growth and desiccation at large-spatial scales in situ, and provided insights on which species may be more vulnerable to climate extremes.

Abstract:

EMO-5 is a free and open European high-resolution (5 km), sub-daily, multi-variable (precipitation, temperatures, wind speed, solar radiation, vapour pressure), multi-decadal meteorological dataset based on quality-controlled observations coming from almost 30 000 stations across Europe, and is produced in near real-time. EMO-5 (v1) covers the time period from 1990 to 2019. In this paper, we have provided insight into the source data, the applied methods, and the quality assessment of EMO-5.

Abstract:

Biodiversity and ecosystem functions are highly threatened by global change. It has been proposed that geodiversity can be used as an easy-to-measure surrogate of biodiversity to guide conservation management. However, so far, there is mixed evidence to what extent geodiversity can predict biodiversity and ecosystem functions at the regional scale relevant for conservation planning. Here, we analyse how geodiversity computed as a compound index is suited to predict the diversity of four taxa and associated ecosystem functions in a tropical mountain hotspot of biodiversity and compare the results with the predictive power of environmental conditions and resources (climate, habitat, soil). We show that combinations of these environmental variables better explain species diversity and ecosystem functions than a geodiversity index and identified climate variables as more important predictors than habitat and soil variables, although the best predictors differ between taxa and functions. We conclude that a compound geodiversity index cannot be used as a single surrogate predictor for species diversity and ecosystem functions in tropical mountain rain forest ecosystems and is thus little suited to facilitate conservation management at the regional scale. Instead, both the selection and the combination of environmental variables are essential to guide conservation efforts to safeguard biodiversity and ecosystem functions.

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:

Land degradation is a major environmental and social issue in temperate steppes. It is commonly determined from vegetation cover using remote sensing techniques. Steppes in eastern Mongolia are subject to resource extraction activities, such as mining and oil extraction, which affect land degradation. Recent technological progress in remote sensing has facilitated the acquirement of high-resolution data by, for example, the CubeSat satellite or unmanned aerial vehicles (UAV), providing data for detailed maps of vegetation cover and plant functional groups (PFGs). Traditional methods for monitoring vegetation cover often face typical scale issues, such as the upscaling of vegetation parameters if plot-scale field measurements are integrated to satellite data. Here, we studied the spatial distribution of PFG using machine learning and a combination of field measurements, UAV imagery (spatial resolution: 2 cm), and PlanetScope multi-temporal imagery. We provide two products at two spatial resolutions: one for UAV data, which is restricted to comparatively small areas around field measurements, and one for PlanetScope, which covers large parts of northeastern Mongolia. The results showed that the overall accuracies of UAV classification were 91–95%, whereas those of PlanetScope were 78–95%. In integrating the classified UAV data to the PlaneScope data, our proposed model minimized the scale issue that often impedes classification. Importantly, our findings revealed that the ecological effects of dirt road and railroad extended up to 60–120 m into the adjacent, otherwise less degraded steppe vegetation. A comparison between burned and unburned areas in different years indicates that wildfires affect the composition of PFG in reducing the fractional cover of graminoids and forbs, and that increasing cover of bare ground leads to a distinct and patchy mosaic of different vegetation types.

Abstract:

The broad dissemination of open data policies for Big Earth data leads to a diversification of users. The literacy in data and data handling differs for each user group, resulting in different needs and requirements. In addition, the development of cloud-based data systems challenges traditional workflows of all users. In order to tailor new cloud-based data systems for Earth data users, it is of utmost importance to obtain a better understanding of users in terms of the type of data they explore, the applications they need the data for, the way they access and process data, and the challenges they face. This is an indispensable prerequisite to ensure an increased uptake of Big Earth data in the future. In order to get a better insight in the requirements and challenges of users coping with Big Earth data, we run a comprehensive web-based user survey. Our results, inclined to users of Big Earth data in Europe and the North American continent, reveal that a majority of survey respondents still download copies of data onto their local machine and handle and process data locally with a combination of programming and desktop-based software. However, survey respondents are facing severe problems related to the growing data volumes, the data heterogeneity and the limited processing capacities for their demanding applications. Thus, they show a specific interest in using cloud-based data services in the near future but express the need for an easier data discovery and the interoperability of data systems. Based on the survey findings, we draw a set of recommendations to make Big Earth data more FAIR (findable, accessible, interoperable and re-useable).

Abstract:

Systematic, practicable, and global solutions are required for insect monitoring to address species decline and pest management concerns. Compact frequency-modulated continuous-wave (FMCW) radar can facilitate these processes. In this work, we evaluate a 60 GHz low-range FMCW radar device for its applicability to insect monitoring. Initial tests showed that radar parameters should be carefully selected. We defined optimal radar configuration during the first experiment and developed a methodology for individual target observation. In the second experiment, we tried various individual-insect targets, including small ones. The third experiment was devoted to mass-insect-target detection. All experiments were intentionally conducted in very uncertain conditions to make them closer to a real field situation. A novel parameter, the Sum of Sequential Absolute Magnitude Differences (SSAMD), has been proposed for uncertainty reduction and noisy data processing. SSAMD enables insect target presence detection and biomass estimation. We have defined ranges of SSAMD for distinguishing noise, insects, and other larger targets (e.g., bats, birds, or other larger objects). We have provided evidence of the high correlation between insect numbers and the average of SSAMD values proving the biomass estimation possibility. This work confirms that such radar devices can be used for insect monitoring. We plan to use the evaluated system assembled with a light trap for real fieldwork in the future.

Abstract:

In the past two decades, the Amazon-exposed, tropical montane rain forests in south Ecuador experienced increasing deposition of reactive N mainly from Amazonian forest fires, episodic Ca and Mg inputs from Saharan dust, and a low but constant P deposition from unknown sources. To explore the response of this tropical, perhumid ecosystem to nutrient inputs, we established in 2007 a Nutrient Manipulation Experiment (NUMEX). Since 2008, we have applied 50 kg ha−1 year−1 of N as urea, 10 kg ha−1 year−1 of P as NaH2PO4·H2O, 50 kg ha−1 year−1 of N + 10 kg ha−1 year−1 of P and 10 kg ha−1 year−1 of Ca as CaCl2·H2O in a randomized block design at 2000 m a.s.l. in a natural forest of the south Ecuadorian Andes. Previous studies have shown that alkali and alkaline earth metals had beneficial effects on the functioning of N and P co-limited tropical forests occurring on acidic soils. Therefore, we determined the response of all major aqueous ecosystem fluxes of K, Ca, Mg and Na to nutrient amendments, to understand how increasing atmospheric deposition would affect their cycling in the future. Additions of N and P decreased K leaching from the organic layer and in the mineral soil, thus tightening K cycling. This suggests that increasing future N and P availability may result in K limitation in the long term. The leaching of Ca and Mg from the canopy increased in response to amendments of N and P and we observed an enhanced uptake of these nutrients also if Ca was amended alone. Although N was applied as urea, acidity of soil solutions and leaching of K, Ca, Mg and Na did not increase following separate N amendments. In spite of the acid soils and of its low cation-exchange competitivity, Na included in the P fertilizer was only partly leached from the organic layer. We suggest that it was probably required to cover an unmet Na demand of the soil fauna. Our results demonstrate the major role in the functioning of the tropical montane forests played by K, Ca and Mg as potential future growth-limiting elements and increasingly required nutrients in response to rising N and P availability, while they also support the importance of Na as a functional element in these ecosystems.