Megan Lewis's research while affiliated with University of Adelaide and other places

... The loss of >40 % of the original (local) seagrass cover prompted management agencies to agree on aspirational load reductions for both nitrogen (by 75 % to 600 t) and suspended solids (by 50 % to 4200 t) (Fox et al., 2007;McDowell and Pfennig, 2013). A significant reduction of nitrogen (>70 %) and solids (>50 %) in effluents from wastewater treatment plants has subsequently been accomplished (Fernandes and Dinesh, 2018), resulting in considerable water quality improvements along the Adelaide coast over the past three decades (Fernandes et al., 2022). Despite these efforts, signs of initial seagrass recovery were slow to appear, particularly in the nearshore (Bryars and Neverauskas, 2004;DEWNR, 2013). ...

... De nombreuses approches permettent d'augmenter artificiellement la quantité de données pour les techniques d'apprentissage : l'injection de bruit [163], les stratégies de relighting [164], la génération d'images hyperspectrales à partir d'images RGB [165], la génération d'images hyperspectrales synthétiques [166,167,168,169], etc... Notons également l'intérêt grandissant pour les méthodes de transfert d'apprentissage entre les premières couches de réseaux convolutifs pré-entraînés sur images RGB vers des réseaux développés pour la segmentation d'images hyperspectrales [170,171,172]. ...

... Extra drag by the presence of seagrass was estimated using Dynveg, a model that takes into account the plant properties and reconfiguration (bending) in the flow (Dijkstra and Uittenbogaard, 2010). Plant biomechanical properties were selected for Posidonia spp., which constitute around 80 % of cover in Adelaide's coastal waters (Clarke et al., 2021). Representative bottom roughness coefficients were estimated (Table S1) and applied to grid cells according to seagrass cover mapped by the South Australian Department for Environment and Water based on data collected in (EnviroDataSA, 2015. ...

... It should be noted that with this method, the global accuracies were up to 98.56%. Despite other classification methodologies being better for the identification of minerals and the generation of lithological maps such as SVM, RF, and CNN, the most used classifier is SAM which, as mentioned above, does not work very well when there are not enough training data [28,96,107]. In geological studies, it is quite difficult to obtain a lot of training data due to difficult access areas as in the case of [105] where a lithological map of an area formed by the Teide-Pico Viejo stratovolcano was made. ...

... UAVs are multipurpose platforms to which a variety of sensory equipment can be mounted. Based on user preference, UAVs can utilize passive sensors that capture reflected rays of electromagnetic radiation such as thermal, near infrared or visible spectrum, or combinations thereof in hyperspectral cameras [27,34,35]. Additionally, these systems may incorporate active sensors, such as radar or lidar which characterize the three-dimensional structure of terrain and vegetation. ...

... HSI is an emerging method that analyzes a wide range of wavelengths instead of only providing each pixel one of the three primary colors [31,32]. HSI has been used in numerous classifications fields, such as agriculture [33], cancer detection [34][35][36][37], military [38], air pollution detection [29,39], remote sensing [40], dental imaging [41], environment monitoring [42], satellite photography [43], counterfeit verification [44][45][46], forestry monitoring [47], food security [48], natural resource surveying [49], vegetation observation [50], and geological mapping [51]. ...

... Big data platforms like the GEE facilitate the comparison of images over a long period, allowing appreciating changes over time. This is apart from the time and cost-effectiveness of remotely sensed data compared to traditional statistical surveys [72]. Improved irrigated areas improve water allocation to farmers, irrigation performance and intensity assessment, and environmental impact assessment, thereby improving irrigation water use efficiency. ...

... Direct observation of photosynthetic cover is also subject to human bias, for example, decisions on whether a yellow leaf counts as photosynthetic or not. In contrast, sensor-based observations present continuous measurements of greenness that can reduce bias (Fisk et al., 2019). Despite these challenges, we believe comparisons of UAS data products and traditional field methods are still valuable, for example, to understand how contemporary UAS imagery relates to historical data. ...

... Modern data analytics technology and the advances in satellite imaging provide access to large datasets that can assist in characterising landscape features and their distribution at regional scales (e.g. Wilford et al. 2016;Jasiewicz et al. 2014Jasiewicz et al. , 2015de Caritat et al. 2017;Caruso et al. 2018;Albrecht et al. 2021). ...

... Traditional survey methods of the abundance or spatial distribution of animal populations have recently been replaced by non-invasively applied conservation drones and image processing techniques [38][39][40][41][42][43]. These aerial methods can decrease the high costs and labor requirements; overcome difficult access to large, remote areas; and increase the accuracy and precision of estimation. ...