Citizen science for monitoring seasonal-scale beach erosion and behaviour with aerial drones

Sandy beaches are highly dynamic systems which provide natural protection from the impact of waves to coastal communities. With coastal erosion hazards predicted to increase globally, data to inform decision making on erosion mitigation and adaptation strategies is becoming critical. However, multi-temporal topographic data over wide geographical areas is expensive and time consuming and often requires highly trained professionals. In this study we demonstrate a novel approach combining citizen science with low-cost unmanned aerial vehicles that reliably produces survey-grade morphological data able to model sediment dynamics from event to annual scales. The high-energy wave-dominated coast of south-eastern Australia, in Victoria, is used as a field laboratory to test the reliability of our protocol and develop a set of indices to study multi-scale erosional dynamics. We found that citizen scientists provide unbiased data as accurate as professional researchers. We then observed that open-ocean beaches mobilise three times as much sediment as embayed beaches and distinguished between slowed and accelerated erosional modes. The data was also able to assess the efficiency of sand nourishment for shore protection. Our citizen science protocol provides high quality monitoring capabilities, which although subject to important legislative preconditions, it is applicable in other parts of the world and transferable to other landscape systems where the understanding of sediment dynamics is critical for management of natural or anthropogenic processes.


Supplementary Method "Citizen scientists, UAV surveys and photogrammetric details"
For this work, citizen scientists are defined as members of the public who volunteer in the acquisition of aerial image data and ground control points for photogrammetric purposes. They have been invited to volunteer through social and traditional media channels or directly approaching local land managers and established community groups. Citizen scientists in our project can have two main roles: (1) Unmanned Aerial Vehicles (UAV) pilots, who are in charge of the UAV flight operations and (2) survey assistants, who assist in the Ground Control Points (GCP) survey while also acting as hazard spotters (e.g. air and pedestrian traffic). As a result, citizen scientists range from primaryschool students to retirees.  Squared Error (rmse) being the most used error metric in the literature 9 , its validity is robust only when a normal distribution of absolute errors with no outliers is assumed, which is seldom occurring due to filtering and interpolation errors introduced by the digital photogrammetric procedure 10,11 . The normalised median absolute deviation (nmad) is reported to be a more robust estimator for elevation precision of photogrammetric products, in case the above mentioned assumptions are not met 10,12,13 .
Accordingly, in addition to performing statistical tests (Shapiro-Whilk and D'Agostino-Pearson tests), we evaluated the normality of the absolute error distribution by visually assessing their Q-Q plots, as

Supplementary Discussion "Citizen scientists' accuracy"
To validate the vertical accuracy of citizen scientists' digital surface models in an operational scenario we used the independent checkpoint method, proved the non-normality and presence of outliers in the absolute error distribution and obtained an nmad of 0.048 m and an rmse of 0.089 m.
In the UAV-SfM beach monitoring literature, although other methods and metrics have been used, the great majority of researchers evaluated vertical accuracy with independent checkpoints and root mean squared error (rmse), with only two studies using robust statistics due to prior error distribution evaluation (Supplementary Table 2). For this reason, in addition to the more appropriate and robust normalised median absolute deviation (nmad), we also compared our rmse value with those reported in the literature (Supplementary Figure 5). However, these studies assumed or reported normal distribution of the checkpoint errors and no outliers, which is not our case and in general rarely occurs when dealing with photogrammetric datasets 10 . Therefore, using the more appropriate and robust nmad as error metrics, citizen scientists obtained the best vertical accuracy reported in the literature so far, although nmad has currently been used in only two other studies 12,14 . Given the aforementioned observations, we conclude that citizen scientists under our protocol can provide beach topographic data as accurate as professional researchers.

Supplementary Discussion "Role of legislation"
While all members of the public are able to participate in this study as survey assistants, those who wish to operate the UAVs must register with the Australia Civil Aviation Safety Authority (CASA).
Thus, differently from other coastal citizen science projects where anyone can take part as a citizen scientist 15,16 , in our approach, legislation plays a central role in defining who can be a pilot,.
In Australia, scientific operations with UAVs < 2 kg are part of the 'sub-2 kg excluded category", which allows individuals to fly small UAVs within standard operating conditions without the need of a remote pilot licence. CASA requires them to be older than 16 years, obtain an aviation reference number and to become accredited operators by passing a free online assessment. No formal risk assessment is required by CASA.
Nevertheless, citizen scientists employed in this work must complete an additional 2-days training program in the theory and practical operation of UAVs for photogrammetric surveys in sandy beaches, organised by Deakin University. This includes knowledge of the regulations, weather, documentation, record keeping, the safe set-up and pack down of the UAV and a flight competency test in both normal operating conditions and abnormal conditions (i.e. simulating a failure in the UAV