Recent technological advances in remote sensing sensors and platforms, such as high-resolution satellite imagers or unmanned aerial vehicles (UAV), facilitate the availability of fine-grained earth observation data. Such data reveal vegetation canopies in high spatial detail. Efficient methods are needed to fully harness this unpreceded source of information for vegetation mapping. Deep learning algorithms such as Convolutional Neural Networks (CNN) are currently paving new avenues in the field of image analysis and computer vision. Using multiple datasets, we test a CNN-based segmentation approach (U-net) in combination with training data directly derived from visual interpretation of UAV-based high-resolution RGB imagery for fine-grained mapping of vegetation species and communities. We demonstrate that this approach indeed accurately segments and maps vegetation species and communities (at least 84% accuracy). The fact that we only used RGB imagery suggests that plant identification at very high spatial resolutions is facilitated through spatial patterns rather than spectral information. Accordingly, the presented approach is compatible with low-cost UAV systems that are easy to operate and thus applicable to a wide range of users.

/pdf/convolutional-neural-networks-enable-efficient-accurate-and-muyhgqniwd.pdf

Convolutional Neural Networks enable efficient, accurate and fine-grained segmentation of plant species and communities from high-resolution UAV imagery

Unmanned aerial vehicles (UAVs) and satellite constellations are both essential Earth Observation (EO) systems for monitoring land surface dynamics. The former is frequently used for its acquisition flexibility and its ability to supply imagery with very high spatial resolution (VHSR); the latter is interesting for supplying time-series data over large areas. However, each of these data sources is generally used separately even though they are complementary and have strong and promising potential synergies. Data fusion is a well-known technique to exploit this multi-source synergy, but in practice, UAV and satellite synergies are more specific, less well known and need to be formalized. In this article, we review remote sensing studies that addressed both data sources. Current approaches were categorized to distinguish four strategies: “data comparison”, “multiscale explanation”, “model calibration” and “data fusion”. Analysis of the literature revealed emerging trends, the supply of these distinct strategies for several applications and allowed to identify key contributions of UAV data. Finally, the high potential of this synergy seems currently under-exploited; therefore a discussion is proposed about the related implications for data interoperability, machine learning and data sharing to reinforce synergies between UAVs and satellites.

https://hal.archives-ouvertes.fr/hal-03137315/document

UAV & satellite synergies for optical remote sensing applications: A literature review

Highly accurate, rapid forest inventory techniques are needed to enable forest managers to address the increasing demand for sustainable forestry. In the last two decades, Airborne Laser Scanning (...

Comparison of ALS- and UAV(SfM)-derived high-density point clouds for individual tree detection in Eucalyptus plantations

Unmanned aerial vehicles (UAV) are increasingly used for spatiotemporal monitoring of invasive plants in coastal wetlands. Early identification of invasive species is necessary in planning, restoring, and managing wetlands. This study assessed the effectiveness of UAV technology to identify invasive Phragmites australis in the Old Woman Creek (OWC) estuary using machine learning (ML) algorithms: Neural network (NN), support vector machine (SVM), and k-nearest neighbor (kNN). The ML algorithms were compared with the parametric maximum likelihood classifier (MLC) using pixel- and object-based methods. Pixel-based NN was identified as the best classifier with an overall accuracy of 94.80% and the lowest error of omission of 1.59%, the outcome desirable for effective eradication of Phragmites. The results were reached combining Sequoia multispectral imagery (green, red, red edge, and near-infrared bands) combined with the canopy height model (CHM) acquired in the mid-growing season and normalized difference vegetation index (NDVI) acquired later in the season. The sensitivity analysis, using various vegetation indices, image texture, CHM, and principal components (PC), demonstrated the impact of various feature layers on the classifiers. The study emphasizes the necessity of a suitable sampling and cross-validation methods, as well as the importance of optimum classification parameters.

https://www.mdpi.com/2072-4292/11/11/1380/pdf

Mapping Invasive Phragmites australis in the Old Woman Creek Estuary Using UAV Remote Sensing and Machine Learning Classifiers

The identification and counting of plant individuals is essential for environmental monitoring. UAV based imagery offer ultra-fine spatial resolution and flexibility in data acquisition, and so provide a great opportunity to enhance current plant and in-situ field surveying. However, accurate mapping of individual plants from UAV imagery remains challenging, given the great variation in the sizes and geometries of individual plants and in their distribution. This is true even for deep learning based semantic segmentation and classification methods. In this research, a novel Scale Sequence Residual U-Net (SS Res U-Net) deep learning method was proposed, which integrates a set of Residual U-Nets with a sequence of input scales that can be derived automatically. The SS Res U-Net classifies individual plants by continuously increasing the patch scale, with features learned at small scales passing gradually to larger scales, thus, achieving multi-scale information fusion while retaining fine spatial details of interest. The SS Res U-Net was tested to identify and map frailejones (all plant species of the subtribe Espeletiinae), the dominant plants in one of the world’s most biodiverse high-elevation ecosystems (i.e. the paramos) from UAV imagery. Results demonstrate that the SS Res U-Net has the ability to self-adapt to variation in objects, and consistently achieved the highest classification accuracy (91.67% on average) compared with four state-of-the-art benchmark approaches. In addition, SS Res U-Net produced the best performances in terms of both robustness to training sample size reduction and computational efficiency compared with the benchmarks. Thus, SS Res U-Net shows great promise for solving remotely sensed semantic segmentation and classification tasks, and more general machine intelligence. The prospective implementation of this method to identify and map frailejones in the paramos will benefit immensely the monitoring of their populations for conservation assessments and management, among many other applications.

Identifying and mapping individual plants in a highly diverse high-elevation ecosystem using UAV imagery and deep learning

Due to their low light conditions, shadows reduce the accuracy of feature extraction and change detection in remote-sensing images. Unmanned aerial vehicles UAVs are capable of acquiring images that have a resolution of several centimetres and removing shadows is a challenge. In this study, the Maximum Likelihood ML and Support Vector Machine SVM classifiers were used to classify a UAV image acquired using a red–green–blue RGB camera over the Old Woman Creek National Estuarine Research Reserve in Ohio, USA. The impact of shadows on the classification process was explored for different pixel sizes ranging from 0.03 to 1.00 m. The SVM generated higher overall accuracy OA at finer spatial resolution 0.25–0.50 m, while the optimal spatial resolution for the ML classifier was 1.00 m. The percentage of shadow coverage increased with spatial resolution for both classifiers 1.71% for ML and 6.63% for SVM. Shadows were detected and extracted using two approaches: a as a separate class using regions of interests ROIs observed in the image, and b by applying a segmentation threshold of 0.3 to visible atmospherically resistant index VARI. The extracted shadows were separately classified using ROIs selected from shaded surfaces, and then removed using the fusion of RGB reflectance, VARI, and digital surface model DSM images. The OA of classified shadows reached 91.50%. OAs of merged sunlit and shadow classified images improved for 18.48% for SVM, and 17.62% for the ML classifier. VARI accurately captures shadows, and when fused with RGB reflectance and DSM, it intensifies their low signal and enhances classification. Whether used to capture or to remove shadows, VARI serves as an effective ‘shadow index’. Shadows create obstacles to remote-sensing processing; however, their spectral information should not be neglected as both shadows and sunlit areas are important for ecological processes such as photosynthesis, carbon balance, evapotranspiration, fish abundance, and more.

Different colours of shadows: classification of UAV images

Mapping land and aquatic vegetation of coastal areas using remote sensing for better management and conservation has been a long-standing interest in many parts of the world. Due to natural complex...

Classification of shoreline vegetation in the Western Basin of Lake Erie using airborne hyperspectral imager HSI2, Pleiades and UAV data

Common Carp Cyprinus carpio and Bigmouth Buffalo Ictiobus cyprinellus are widespread across the Midwest, where they can achieve high population densities in shallow lakes. Despite their contrasting statuses as invasive and native fishes, these species are commonly managed together due to their capacity to have deleterious effects on water quality. However, intra-annual sampling variability is poorly understood for both species, making it difficult to track population changes. We compared boat electrofishing CPUE and proportional size distribution (PSD) of Common Carp and Bigmouth Buffalo during May–October 2018–2020 in seven shallow northwestern Iowa lakes. We also estimated sampling effort necessary to reach targets of fish captured (100) and precision (relative standard error < 25%) and to detect 25% and 50% differences in CPUE between samples. Common Carp CPUE peaked in June (29 fish/h), declined across the summer, and was lowest in October (9 fish/h). Bigmouth Buffalo CPUE exhibited bimodal peaks in May (16 fish/h) and October (12 fish/h), with low catches during August (2 fish/h). Common Carp size structure was largest during October (PSD of preferred-size fish [PSD-P] = 92) and smallest in July (PSD-P = 58). Bigmouth Buffalo size structure was smaller in June (PSD-P = 20) than in all other months (PSD-P = 34). Our targets for number of fish captured and precision were typically reached within twenty-five 15-min electrofishing transects for Common Carp during periods of high catch rates but were not reached after 100 transects for Bigmouth Buffalo in most cases. Our results suggest that timing of sampling is important for reliably and efficiently assessing Common Carp and Bigmouth Buffalo populations. We recommend the sampling of Common Carp with electrofishing during June and the sampling of Bigmouth Buffalo during May in Midwestern shallow lakes to maximize catch rates while minimizing the influence of seasonal sampling variability. 

https://onlinelibrary.wiley.com/doi/pdfdirect/10.1002/nafm.10877

Intra‐annual variability of common carp and bigmouth buffalo electrofishing data in shallow lakes

Abstract Martin A. Simonson, Michael J. Weber, Grace M. Wilkinson and Andrew R. Annear. 2023. Annual changes in water quality and sportfish community structure following commercial harvest of common carp and bigmouth buffalo. Lake Reserv Manage. 39:174–189. Commercial harvest of common carp (Cyprinus carpio; hereafter carp) and bigmouth buffalo (Ictiobus cyprinellus; hereafter buffalo) populations had little detectable effect on shallow lake ecosystems. We tested whether carp and buffalo biomass removal affects limnological variables and fish community metrics across 6 shallow, natural lakes of northwestern Iowa using mixed effects models. Annual commercial harvest of carp ranged from 0 to 71 kg/ha; annual harvest of buffalo ranged from 0 to 356 kg/ha. Biomanipulation (i.e., commercial harvest) of carp was associated with decreases in soluble phosphorus and nitrogen concentrations, but not total phosphorus or nitrogen. Buffalo harvest was unrelated to annual changes in nutrient concentrations but was associated with reductions in chlorophyll a and phycocyanin concentrations. Secchi disk transparency and total suspended solids were unrelated to carp and buffalo harvest. Carp harvest was associated with reduced biomass of large cladocerans but no other zooplankton biomass densities; buffalo harvest was unrelated to zooplankton biomass. Species richness and rake density of aquatic macrophytes were unrelated to carp and buffalo harvest. Carp and buffalo harvest was unrelated to changes in most indices of sportfish abundance, condition, and size distribution. Our results suggest harvest of carp and buffalo <71 kg/ha has little effect on abiotic and biotic ecosystem components on short time scales and highlights the challenges associated with shallow lake restoration.

Annual changes in water quality and sportfish community structure following commercial harvest of common carp and bigmouth buffalo

15 Bigmouth buffalo ( Ictiobus cyprinellus ) is a large-bodied planktivore inhabiting shallow 16 waterways in North America and subjected to unregulated harvest throughout much of their 17 native range. Despite high harvest pressure on some populations, we know little about the 18 ecosystem-level effects of lowering bigmouth buffalo densities. To evaluate the effect of 19 bigmouth buffalo density on lower trophic levels we added fish in ambient and harvested 20 densities to a set of ponds and compared plankton dynamics to reference, fishless ponds. 21 Zooplankton biomass declined in ambient density ponds after the fish addition while 22 chlorophyll-a concentrations increased. In the harvested density treatment chlorophyll-a 23 concentrations were similar to the reference ponds despite a decline in zooplankton biomass, 24 likely due to larger zooplankton community size in the harvested ponds exerting greater grazing 25 pressure. This experimental manipulation revealed density-dependent top-down control on lower 26 trophic levels by bigmouth buffalo and the potential for cascading trophic interactions with the 27 harvest of this species. These results point to the need for ecosystem-level studies of the effects 28 of intensive harvest of bigmouth buffalo on water quality. 29

Martin Albert Simonson

Papers

Different colours of shadows: classification of UAV images

Classification of shoreline vegetation in the Western Basin of Lake Erie using airborne hyperspectral imager HSI2, Pleiades and UAV data

Intra‐annual variability of common carp and bigmouth buffalo electrofishing data in shallow lakes

Annual changes in water quality and sportfish community structure following commercial harvest of common carp and bigmouth buffalo

Experimental evaluation of the effects of bigmouth buffalo ( Ictiobus 1 cyprinellus ) density on shallow lake ecosystems 2