Laboratory for
Climatology and Remote Sensing

Abstract:

Accurate forecasts of radiation fog are an objective of significant relevance due to its impact on traffic, aviation, and transportation. This study will explore the adaptation and enhancement of a previously developed Machine Learning-based nowcasting framework for radiation fog events. The objective is to explore the expansion potential to a spatial scale and model accuracy improvements through application at three distinct weather station locations that experience radiation fog. Further, the effectiveness of Numerical Weather Prediction (NWP) data as additional predictor variable source on model performance will be assessed. This will be performed through integration of datasets from German Weather Service (DWD) stations, Meteosat Second Generation (MSG) channel properties and regional reanalysis variables from COSMO NWP model. Distinct model variants based on different dataset combinations (Station, MSG+COSMO, Station+MSG+COSMO, Visibility-Only) will be evaluated. Using eXtreme Gradient Boosting (XGBoost) algorithm, the framework forecasts absolute visibility with 60-minute lead time. A persistence model serves as benchmark. Performance will be assessed using scoring metrics (Accuracy, Correlation, Percentage bias, Mean Absolute Error) across the full visibility range and three visibility threshold bounds (2 km, 1.1 km, 0.4 km). Temporal accuracy of fog formation and dissipation will be determined through evaluation of fog formation and dissipation time shifts. XGBoost models mostly outperform PM, with tendencies of Station+MSG+COSMO variant performing best and MSG+COSMO variant worst. Prediction difficulties arise in the 0.4 km threshold segment due to measurement resolution limitations and value imbalance of visibility data. The model variants reliably predict fog event transitions, with the majority forecasted with deviations < 30 minutes and only few events overseen. A consistent tendency towards delayed prediction is observed. Variability in model performances across station locations suggests that small-scale environmental characteristics contribute to different model robustness at distinct sites. The results indicate strong potential for further spatial framework extension. COSMO variables partially contribute to improved model performance. The framework marks a solid foundation for future exploitation.

Abstract:

Latent heat flux is a central element of land-atmosphere interactions under climate change. Knowledge is particularly poor in the biodiversity hotspot of the Andes, where heat flux measurements using eddy covariance stations are scarce and land surface models (LSMs) often oversimplify the complexity of the ecosystems. The main objective of this study is to perform latent heat flux simulations for the tropical South Eastern (SE) Ecuadorian Andes using a coupled LSM framework, and to test the performance with heat flux and soil moisture data collected from a tropical high-altitude pasture. Prior to testing, we applied multi-criteria model calibration of sensitive model parameters, focusing on improving simulated soil water conditions and radiation fluxes as a prerequisite for proper heat flux simulations. The most sensitive parameters to improve soil moisture and radiation flux simulations were soil porosity, saturated hydraulic conductivity, leaf area index, soil colour and NIR (Near Infrared) leaf optical properties. The best calibrated model run showed a very good performance for half-hourly latent heat flux simulations with an R2 of 0.8 and an RMSE of 34.0 W m−2, outperforming simulations with uncalibrated and uncoupled LSM simulations in comparable areas. The slight overall overestimation in the simulated latent heat flux can be related to (i) simulation uncertainties in the canopy heat budget, (ii) an imbalance in the observed flux data and (iii) slight overestimations in the simulated soil moisture. Although our study focuses on latent heat fluxes and their relation to simulated radiation fluxes and soil moisture, model outputs of sensible heat fluxes were also discussed. The systematic overestimation of sensible heat flux in the model seems to be mainly a result of overestimated canopy temperatures. The improved simulation for latent heat flux has a high translational potential to support land use strategies in the tropical Andes under climate change.

Abstract:

An understanding of sub-hourly precipitation variability in the Galapagos Islands is crucial for water resource management and effective biodiversity conservation. This study compares the diurnal cycle and event-scale precipitation characteristics (ESPC), such as mean and maximum intensity, duration and rainfall accumulation at different altitudes during El Niño-Southern Oscillation (ENSO) 2022-2024 on Santa Cruz Island. The La Niña phase was analyzed from April 2022 to January 2023 and the El Niño phase from June 2023 to April 2024. Precipitation data, recorded every 10 minutes, was collected from a recently established network of automatic weather stations, which were strategically positioned at three windward and two leeward sites. The results suggest that the diurnal cycle was influenced by altitude, with a maximum vari ability between morning and afternoon, regardless of ENSO phase. During La Niña, ESPC exhibited similarities at interme diate altitudes at both windward and leeward sides. However, rainfall events at the island’s summit were less intense and of longer duration. During El Niño, the highest intensities were observed along the coast and at intermediate altitudes of both windward and leeward locations. In contrast, at the top of the island, rainfall events were less intense and more prolonged. At all altitudes, more than half of the rainfall events corresponded to garúa events, and at the top of the island, almost all events were of this type. At this altitude, the contribution of garúa events to the total rainfall accumulation was 80% and 85% for La Niña and El Niño, respectively. This study provides a detailed analysis of how sub-hourly precipitation varies significantly at different altitudes on the windward and leeward sides as a function of ENSO phases, providing valuable baseline information for future studies in this unique and fragile ecosystem.

Abstract:

Understanding the partitioning of downward shortwave radiation into direct and diffuse components is essential for modeling ecosystem energy fluxes. Accurate partitioning functions are critical for land surface models (LSMs) coupled with climate models, yet these functions often depend on regional cloud and aerosol conditions. While data for developing semi-empirical partitioning functions are abundant in mid-latitudes, their performance in tropical regions, particularly in the high Andes, remains poorly understood due to scarce ground-based measurements. This study analyzed a unique dataset of shortwave radiation components from a tropical mountain rainforest (MRF) in southern Ecuador, developing and testing a locally adapted partitioning function using Random Forest Regression. The model achieved high accuracy in predicting the percentage of diffuse radiation (%Dif; R2=0.95, RMSE = 5.33, MAE = 3.74) and absolute diffuse radiation (R2=0.99, RMSE = 5.30, MAE = 14). When applied to simulate upward shortwave radiation, the model outperformed commonly used partitioning functions achieving the lowest RMSE (8.62) and MAE (5.82) while matching the highest R2 (0.97). These results underscore the importance of regionally adapted radiation partitioning functions for improving LSM performance, particularly in complex tropical environments. The adapted LSM will be further utilized for studies on heat fluxes and carbon sequestration.

Abstract:

Fog and low stratus clouds (FLS) form as a result of complex interactions of multiple factors in the atmosphere and at the land surface and impact both the anthropogenic and natural environments. Here, we analyze the role of synoptic conditions and aerosol loading on FLS occurrence and persistence in the Po valley in northern Italy. By applying k‐means clustering to reanalysis data, we find that FLS formation in the Po valley is either based on radiative processes or moisture advection from the Mediterranean sea. Satellite‐based data on FLS persistence shows longer persistence of radiatively formed FLS events, likely due to air mass stagnation and a temperature inversion. Ground‐based aerosol optical depth observations further reveal that FLS event duration is significantly higher under high aerosol loading. The results underline the combined effect of topography, moisture advection and aerosol loading on the FLS life cycle in the Po valley.

Abstract:

Radiation fog nowcasting remains a complex yet critical task due to its substantial impact on traffic safety and economic activity. Current numerical weather prediction models are hindered by computational intensity and knowledge gaps regarding fog-influencing processes. Machine-Learning (ML) models, particularly those employing the eXtreme Gradient Boosting (XGB) algorithm, may offer a robust alternative, given their ability to learn directly from data, swiftly generate nowcasts, and manage non-linear interrelationships among fog variables. However, unlike recurrent neural networks XGB does not inherently process temporal data, which is crucial in fog formation and dissipation. This study proposes incorporating preprocessed temporal data into the model training and applying a weighted moving-average filter to regulate the substantial fluctuations typical in fog development. Using an ML training and evaluation scheme for time series data, we conducted an extensive bootstrapped comparison of the influence of different smoothing intensities and trend information timespans on the model performance on three levels: overall performance, fog formation and fog dissipation. The performance is checked against one benchmark and two baseline models. A significant performance improvement was noted for the station in Linden-Leihgestern (Germany), where the initial F1 score of 0.75 (prior to smoothing and trend information incorporation) was improved to 0.82 after applying the smoothing technique and further increased to 0.88 when trend information was incorporated. The forecasting periods ranged from 60 to 240 min into the future. This study offers novel insights into the interplay of data smoothing, temporal preprocessing, and ML in advancing radiation fog nowcasting.

Abstract:

Floods cause significant damage to human life, infrastructure, agriculture, and the economy. Predicting peak runoffs is crucial for hazard assessment, but it is challenging in remote areas like the Andes due to limited hydrometeorological data. We utilized a 300 km2 catchment over the period 2015–2021 to develop runoff forecasting models exploiting precipitation information retrieved from an X-band weather radar. For the modeling task, we employed the Random Forest (RF) algorithm in combination with a Feature Engineering (FE) strategy applied to the radar data. The FE strategy is based on an object-based approach, which derives precipitation characteristics from radar data. These characteristics served as inputs for the models, distinguishing them as “enhanced models” compared to “referential models” that incorporate precipitation estimates from all available pixels (1210) for each hour. From 29 identified events, enhanced models achieved Nash-Sutcliffe efficiency (NSE) values ranging from 0.94 to 0.50 for lead times between 1 and 6 h. A comparative analysis between the enhanced and referential models revealed a remarkable 23% increase in NSE-values at the 3 h lead time, which marks the peak improvement. The enhanced models integrated new data into the RF models, resulting in a more accurate representation of precipitation and its temporal transformation into runoff.