Stæhr PA, Groom GB, Krause-Jensen D, Hansen, LB, Huber S, Jensen, LØ, Rasmussen MB, Upadhyay, S, Ørberg, SB. 2019. Use of remote sensing technologies for monitoring chlorophyll a and submerged aquatic vegetation in Danish coastal waters. Part of the RESTEK project. Aarhus University, DCE – Danish Centre for Environment and Energy, 62 pp. Scientific Report No. 139. 

Summary 

This project investigated the abilities of different remote sensing (RS) techniques for monitoring water quality in the nearshore, shallow Danish coastal waters. Our analysis was divided into two parts: 1) monitoring of the cover of submerged aquatic vegetation (SAV: composed of eelgrass, other rooted macrophytes and macroalgae) and 2) monitoring of the water quality parameter chlorophyll a (Chl). The overall aim was to evaluate the potential of using different RS technologies as a replacement or supplement to the conventional Danish national NOVANA (National Monitoring and Assessment Programme for the Aquatic and Terrestrial Environment) monitoring programme for assessment of environmental targets (Good Ecological Status: GES) in the 119 bodies of water (Vandområder) included in the EU Water Framework Directive (WFD). The investigated RS techniques were Sentinel 2 and 3 satellites (S2, S3), summer orthophotos (SOPs) and unmanned aerial vehicles (UAVs, “drones”). Strengths, weaknesses and knowledge gaps were evaluated based on an investigation in nine shallow coastal systems, covering a total of 17 bodies of water from 2016 to 2018. In all systems, data collected with RS techniques were compared with in situ data. Based on the gained experiences, we analysed the costs and benefits of the different RS tools and provide suggestions on how to implement RS into the environmental monitoring framework of the coastal zone. Finally we present future work needed to facilitate integration of the different RS techniques into monitoring of Danish coastal waters. The main findings are presented below.

Part I: SAV monitoring with S2, SOP and drones:

• S2 is the most realistic approach for large-scale (national) annual SAV mapping in the coastal zone. SOPs are a good supplement providing valuable detailed information on patch development. The SOP archive from the 1950s furthermore allows for long-term change assessments in areal distribution, which can be used as a GES indicator.
• Both S2 and SOPs have limitations in estimating the maximum depth of SAV distribution, which for eelgrass is a key parameter in the assessment of coastal water quality and GES.
• The in situ drone technique is currently the only RS method capable of distinguishing between specific vegetation types. Moreover, first results show that the technique is able to provide SAV depth limits, but this needs further tests and validation/intercalibration against existing methods before any conclusions can be made. On the other hand, the in situ drone approach only provides discontinuous measurements (points) and is not feasible for large area national monitoring.
• Among the different RS techniques investigated in this report, the in situ application of drones provides the most obvious and easiest technology to implement into the current vegetation monitoring in the coastal zone.
• Overall, each of the presented techniques – S2, SOP, drones – has its advantages and disadvantages and the way forward is a smart combination of the RS techniques into an efficient monitoring framework to enhance the benefits of SAV monitoring, in particular, if SAV areal cover will be used as a supplementing indicator of ecological status. 

Part II: Chl monitoring with S2 and S3 satellites:

• The Sentinel satellites provide unprecedented free data in terms of spatial, spectral and temporal resolutions. Never before has satellite imagery for Denmark been available on a daily basis and consequently collecting great amounts of data which among other use can be used to estimate Chl in water bodies. However, the satellites are relatively new (launched between 2015-2018) and method development for the processing of the data is continuously improving. The results presented in this report should therefore be seen as a snapshot.
• Two different Sentinel satellites were investigated: S2 mainly designed for land applications providing imagery at 10 metre pixel resolution but only every 3-5 days, while S3 specifically designed for water applications providing imagery on a daily basis but only at 300 metre pixel resolution. The three main advantages of using satellite imagery are 1) to get frequently insights into the spatial distribution of Chl for the entire water body and large areas, 2) to get information for water bodies not included in the ground surveys at all and 3) to get data in between in situ sampling.
• In general, the analyses revealed a higher variability in satellite-derived Chl than in situ data, with S2-derived Chl particularly noisy. The results from S3 turned out to be smoother and generally closer to the NOVANA data. Despite the coarser pixels of S3, even for small water bodies, we achieved good results, most likely due to the water-specific sensor design of the S3 satellite. In order to get the most comprehensive GES assessment, the combination of S2 and S3 seems like a cost-effective way to supplement in situ monitoring, in particular in the currently non-monitored water areas.
• Temporally aggregated Chl (monthly means) showed good agreement between all three approaches, still with S2 having the highest variability.
• In order to get a quality estimate of the satellite-derived Chl concentrations, they are usually compared with in situ measurements, like NOVANA. These comparisons bring together very different spatial scales, with ground surveys providing point information and the satellites providing spatially aggregated information per pixel, like 300 × 300 metres for S3. Moreover, it is often forgotten that also in situ measurements can be erroneous. A thorough quantitative statistical analysis is mostly not feasible because of the time difference between ground sampling and satellite overpass. A proper match-up useful for statistics should optimally fall within a 30 minutes to 2 hours window from the satellite overpass to avoid moving of water masses. Still it is useful to contrast Chl from different sources by looking into time series and their seasonality and how well and when they compare.
• As the S2 and S3 satellites only retrieve information on water surface properties, they cannot replace in situ sampling in deeper stratified waters where important Chl peaks often occur. In general, in situ observations are always needed for comparison purposes as mentioned above.
• Proper validation of S2 and S3 Chl estimates has not been possible under this activity due to lack of suitable match-up data. For both S2 and S3, the inclusion of a longer time series and expansion of the geographic scope – e.g. entire Denmark – would expand the validation data basis significantly. As a supplement to NOVANA activities, we recommend that the EPA invest in the AERONET system where one or more stations could contribute with important data for an optimization of the Chl retrieval algorithms to Danish coastal conditions and contribute to a proper validation.  

Overall, for both Chl and eelgrass monitoring it must be stressed that collection of in situ data should continue to be an important part of the national assessments, as these data are essential for evaluating RS data and information on e.g. eelgrass depth limits is an important part of existing monitoring of marine vegetation.