Theme: Source control measures - Monitoring & modelling
Thursday, July 2
Role of vegetation maintenance in nature-based stormwater solution: short-term plant succession in a rain garden in Estonia
Vegetation in rain gardens plays a central role in hydraulic regulation, pollutant uptake, evapotranspiration, and soil stabilization, directly influencing the long-term performance of nature-based stormwater solutions. However, the continued functioning of these systems depends strongly on routine management. To assess how quickly vegetation structure may change under unmaintained conditions, this study examines plant community dynamics in a previously well-maintained, five-year-old parking-lot rain garden following the cessation of maintenance for one growing season. Vegetation was surveyed in summer 2025 and compared to the initial planting to evaluate species persistence, spatial shifts, and emergent successional patterns. The results show a marked decline in one originally planted species out of 11, contrasting with the expansion of three species and the establishment of a total of 29 spontaneous local species. Overall, the study highlights the importance of integrating continuous vegetation management strategies into the operation of nature-based stormwater solutions to ensure resilient ecological function and sustained hydraulic and water treatment performance.
Evaluating the use of drone-based imagery to assess stormwater wetland vegetation health
Constructed wetlands rely on healthy vegetation and available storage volume to effectively treat stormwater. For wetlands with poor water quality performance, real-time control (RTC) of water levels can enhance treatment performance, but it is not clear what impact this might have on the vegetation. In Melbourne, Australia, a common issue of water levels being too high for too long has resulted in widespread vegetation loss. This study investigates drone-based multispectral imaging as a rapid, scalable alternative to labour-intensive vegetation surveys at an appropriate resolution for vegetated zones in smaller assets. High-resolution imagery from four flights was processed to generate normalized difference vegetation indices (NDVI), normalized difference water indices (NDWI), and single-band multispectral layers for both assessing cover and training Random Forest models for vegetation type and health classification. Preliminary results show high accuracy when applied to training imagery and moderate performance when transferred across survey dates. In the near future, a wetland drawdown using RTC will be undertaken to evaluate the effect on vegetation recovery. These findings demonstrate the potential of drone-based remote sensing to support adaptive wetland management and provide a foundation for expanded multi-season, multi-site analysis.
Automated Unmanned Aerial Vehicle (UAV) based condition assessment of Blue-Green Infrastructure: Application to green roofs
Blue-Green Infrastructure (BGI) is increasingly deployed to mitigate climate-change impacts in cities, yet its long-term functionality is difficult to monitor with conventional, manual inspections. This study presents an automated workflow combining UAV-based imaging, machine-learning object detection, multispectral vegetation analysis and AI-assisted interpretation for assessing the condition of green roofs. High-resolution RGB and multispectral mosaics are analysed using a Mask R-CNN model to identify roof boundaries and typical failure modes, while NDVI indicators quantify vegetation health. A large language model then synthesizes structural and vegetative findings into standardized diagnostic outputs. A preliminary demonstration with UAV imagery from Brandenburg (Germany) confirms the feasibility of the workflow and its potential for rapid, non-invasive assessments. Upcoming validation using datasets from Malmö and Berlin will evaluate detection accuracy and examine the applicability of this approach for scalable BGI asset management and maintenance planning.
Down in the flood: Mapping the impact of external factors and flood paths on green infrastructure maintenance
Green Infrastructure (GI) is increasingly used for urban stormwater management, yet municipalities lack systematic tools to assess where external stressors may compromise long-term performance. This study operationalizes a proxy-based framework for external stressors and combines it with overland flood-path mapping for 139 GIs in Trondheim, Norway. Stressor indices for leaf litter (LL), sediments from unsealed surfaces (SU) and trash accumulation (TA) were derived from available datasets to create hotspot maps. Overland flood paths were obtained for contributing areas between 0.5 and 5 ha and were used to classify GIs according to their proximity to flood corridors. Results show distinct spatial patterns for LL, SU and TA, implying that different GIs are dominated by different stressor combinations. Only a small subset of GIs lies within 5 m of the main 5 ha and 2.5 ha flood paths, making these hydraulically strategic assets. Four contrasting GIs are examined in detail to illustrate how combined stressor–flood mapping can guide inspection and maintenance activities. The approach supports GI asset management by linking spatial stressor information with flood-path exposure at city scale.