OpenSkyLab | Research

Environmental remote sensing has advanced rapidly with the rise of drones, high-resolution sensors and expanding satellite and in-situ archives. These tools have opened new possibilities for monitoring farmland and forest biodiversity, detecting vegetation stress, mapping surface temperature and quantifying structural parameters. Yet persistent uncertainty remains a major challenge. Variable weather conditions, sensor limitations and divergent workflows can degrade data quality, hinder comparability and weaken reproducibility—especially in structurally complex landscapes such as heterogeneous farmland and mixed forests.

Experience from recent studies highlights that reliable monitoring depends on systematic metadata collection. Flight parameters, sensor settings, and weather conditions during acquisition, including solar angle, wind speed and humidity, must be documented to ensure results are interpretable, comparable and repeatable. Robust correction procedures are equally essential. Radiometric and atmospheric calibration, geometric alignment and sensor harmonisation underpin any comparison across dates, sites or platforms. Without such steps, vegetation indices, canopy parameters or surface temperature estimates can be unstable and misleading.

Another consistent finding is that there is no universal workflow. Thresholds and pipelines must be tailored to the environment and the variable of interest, while transparency about assumptions and provenance improves transferability. Larger and better-stratified samples reduce variance more effectively than post-hoc modelling fixes, and multi-sensor integration strengthens robustness when geometric and radiometric consistency are maintained. Ultimately, progress depends not only on technology but also on people: close cooperation between ecology and computer science is essential to design methods that are transferable, scalable and ecologically meaningful.

Taken together, these lessons emphasise that advancing drone-based environmental monitoring will rely less on ever-new sensors and more on smarter deployment strategies, careful correction workflows and transparent, standardised data handling. Drone observations can only fulfil their potential when accompanied by rigorous metadata, robust validation and a commitment to reproducibility, ensuring that the insights they generate are both scientifically sound and operationally useful.

Relevant papers
Komárek, J. (2025). When the Wind Blows: Exposing the Constraints of Drone‐Based Environmental Mapping. Natural Sciences, 5(1-2), e70003. Link
Chakhvashvili, E., Machwitz, M., ..., Komárek, J., Klouček, T., & Rascher, U. (2024). Crop stress detection from UAVs: best practices and lessons learned for exploiting sensor synergies. Precision Agriculture, 25(5), 2614-2642 Link
Moravec, D., Komárek, J., López-Cuervo Medina, S., & Molina, I. (2021). Effect of atmospheric corrections on NDVI: Intercomparability of Landsat 8, Sentinel-2, and UAV sensors. Remote Sensing, 13(18), 3550 Link
Komárek, J. (2020). The perspective of unmanned aerial systems in forest management: Do we really need such details?. Applied Vegetation Science, 23(4), 718-721 Link

To address these limitations, we are developing a correction framework that integrates drone thermal imagery with ground-based meteorological observations and vegetation indices. The framework incorporates relative humidity, wind speed, and NDVI-derived emissivity to improve radiometric consistency and reduce systematic errors. Atmospheric correction models compensate for transmittance loss and emissivity variability across vegetation types, while ground reference stations provide temperature and humidity benchmarks. This integration significantly enhances the accuracy of surface temperature estimates, enabling the more reliable derivation of soil and canopy moisture proxies.

The resulting workflow provides a practical, scalable method that links drone observations directly to ground conditions and extends these relationships across time series. This approach enhances our ability to monitor short-term variability in surface energy balance and water stress, offering valuable insights for biodiversity monitoring, precision agriculture, and ecosystem management.

Despite these advances, several challenges remain. Thermal drift in lightweight sensors, rapid atmospheric fluctuations during flight, and mixed-pixel effects in heterogeneous canopies can still introduce uncertainty. Ongoing work focuses on automating calibration routines, integrating radiative transfer models, and testing the framework across diverse climatic regions to assess robustness and transferability.

Ultimately, integrating drone-based thermal data with in-situ and spectral information provides a powerful tool for monitoring land–atmosphere interactions at high spatial and temporal resolutions. Potential applications range from precision irrigation scheduling and crop stress detection to biodiversity assessment, drought risk analysis, and ecosystem management. By bridging the gap between drone-scale measurements and ground observations, our framework offers a pathway toward more reliable, repeatable, and interpretable estimates of surface temperature and moisture, which making drone thermal monitoring more robust and operationally valuable for ecological and agricultural research.

Relevant papers
Rous, J., Kuželková, M., Jačka, L., Komárek, J. (2025) UAV-Based Surface Temperature Estimates in Agroforestry: Improvement Through Integration with Ground Sensors and Meteorological Observations. In Review

Our methods integrate high-resolution time series of multispectral and thermal drone imagery to capture changes in spectral reflectance and canopy surface temperature, which are key indirect indicators of stress. Drones equipped with multispectral and thermal sensors provide detailed observations at the level of individual tree crowns, while machine and deep learning algorithms accurately detect early-warning signals, classify infested trees, and model outbreak progression under variable environmental conditions.

Multispectral and thermal data capture fundamentally different aspects of vegetation stress. Multispectral imagery measures reflected light in discrete wavelength bands (typically visible, red-edge, and near-infrared), providing information about leaf pigment composition, chlorophyll content, and canopy structure. Stress factors, such as drought, nutrient deficiency, or pest infestation, alter photosynthetic activity and pigment balance, which can be detected as changes in canopy reflectance or vegetation indices (e.g., NDVI, NDRE). In contrast, thermal imagery records emitted longwave radiation, reflecting the surface temperature of the canopy. Under water stress, plants close their stomata to reduce transpiration, leading to increased leaf temperature. Thermal data thus provide a direct physical indicator of stress intensity and are particularly useful for assessing current or acute stress conditions. Combining both approaches yields a comprehensive view of canopy condition: multispectral indices reveal the onset and type of stress (symptoms), while thermal data quantify its severity and temporal dynamics (causes).

A key focus of our work lies in reproducibility and scalability. By testing our workflows across different forest types and outbreak stages, we develop transferable methods that forest managers and agencies can adapt to local contexts. Open data policies, transparent metadata, and harmonised processing pipelines ensure that our results extend beyond individual case studies and contribute to long-term, policy-relevant forest health monitoring frameworks.

Our research highlights the potential of drone-based multispectral and thermal remote sensing for forest monitoring and protection. Beyond pest detection, our methods support broader goals: preserving ecosystem services, maintaining timber production, and conserving biodiversity in forested landscapes under increasing climate pressure.

For more information and project updates, please visit https://kurovec.czu.cz.

Relevant papers
Klouček, T., Modlinger, R., Zikmundová, Štěpánová, K., Pracná, P., Rous, J., Kozhoridze, G., Štych, P., Laštovička, J., & Komárek, J. The sensitivity of UAV-borne thermal imagery for early detection of the bark beetle-infested spruce trees. Journal of Forestry Research. In Review.
Klouček, T., Modlinger, R., Zikmundová, M., Kycko, M., & Komárek, J. (2024). Early detection of bark beetle infestation using UAV-borne multispectral imagery: a case study on the spruce forest in the Czech Republic. Frontiers in Forests and Global Change, 7, 1215734 Link
Klouček, T., Komárek, J., Surový, P., Hrach, K., Janata, P., & Vašíček, B. (2019). The use of UAV mounted sensors for precise detection of bark beetle infestation. Remote Sensing, 11(13), 1561 Link

At CZU Prague, our research team develops and validates drone-based approaches for semi-automatic mapping and monitoring of vegetation with very high spatial precision at the level of individual trees or shrubs, enabling detailed assessment of vegetation structure and dynamics along railways and roads. Using multispectral and RGB sensors, our workflow captures vegetation indices (e.g., NDVI, green leaf index), canopy height models, and canopy density metrics. These data are collected through time-series drone campaigns, enabling the detection of seasonal changes, growth trends, and rapid changes associated with invasive species or safety hazards. The resulting datasets create a spatial database of individual trees, which is validated against ground-based surveys to ensure local accuracy and reliability. By combining drone-based multispectral or RGB imagery and canopy height models with machine learning and time-series analysis alongside in-situ measurements, we can support infrastructure managers in decision-making about green space maintenance.

Thanks to multidisciplinary research combining environmental science with practical management needs, our workflow delivers spatially explicit risk maps, vegetation health indices, and temporal analyses. These outputs provide infrastructure managers with actionable information to plan vegetation clearance, prioritise interventions, optimise maintenance schedules, and mitigate ecological risks. Importantly, the workflows are designed for reproducibility and scalability, allowing railway and road authorities to implement them in different regions with minimal adaptation.

This research demonstrates how drone-based remote sensing can simultaneously support ecological understanding, public safety, and economic efficiency. By integrating high-resolution drone data with operational management workflows, we contribute to safer transport networks, more sustainable vegetation management, and reduced long-term costs, while also creating a valuable open dataset for research and monitoring.

For more information and project updates, please visit https://doprava.fzp.czu.cz.

Relevant papers
Komárek, J., Lagner, O., & Klouček, T. (2024). UAV leaf-on, leaf-off and ALS-aided tree height: A case study on the trees in the vicinity of roads. Urban Forestry & Urban Greening, 93, 128229 Link
Komárek, J., Klápště, P., Hrach, K., & Klouček, T. (2022). The potential of widespread UAV cameras in the identification of conifers and the delineation of their crowns. Forests, 13(5), 710 Link
Klouček, T., Klápště, P., Marešová, J., & Komárek, J. (2022). UAV-borne imagery can supplement airborne Lidar in the precise description of dynamically changing Shrubland Woody vegetation. Remote Sensing, 14(9), 2287 Link