• Rural Digital Europe
• Publications close
• Mémoires en Sciences de l'Informati... close
• Canadian Journal of Forest Research close

High-energy diets increase the risk of Sub-Acute Ruminal Acidosis (SARA) inruminants. The behaviour of ruminants is modified under SARA however the impact onbehaviour of a high-energy diet and those of SARA per se are often confounded. Weaimed at distinguishing the effects of the diet and of SARA on cow behaviour. We fed28 Holstein cows a low-starch diet (low starch diet, 10.5% starch) or a high-starch diet(high starch diet, 31.5% starch). Control cows (n = 14) received the low starch diet for60 days, Challenge cows (n = 14) received the same diet except for a 2-week periodduring which they received the high starch diet and the 10 preceding days to ensurethe transition between diets. We monitored ruminal pH and activity of cows thanks tosensors, dry matter intake (DMI), and milk yield of each cow on a daily basis. TheSARA status was defined according to the relative decrease in ruminal pH and to pHvariability. High starch diet induced SARA more often than low starch diet (81% dayswhen cows received high starch diet vs. 8% when they received low starch diet. Highstarch diet also decreased milk yield and made cows spending less time eating buteating more quickly (Challenge vs. Control cows during the challenge period: milkyield, 20.0 vs. 18.2 L/d; % time spent eating, 22.5 vs. 27.6; eating rate, 77.1 vs. 69.6 gDMI/min; P (diet x period) 0.50). In conclusion, an increase ineating rate, especially combined with a decrease in milk yield should alert farmers tothe risk of ruminal acidosis. International audience

Authenticated key exchange protocols play a crucial role in the communication security stack of an Internet-of-Things (IoT) device: they authenticate the communicating parties and establish a shared symmetric secret between them.Following a large debate in the community, the Internet Engineering Task Force (IETF) has recently standardized a new protocol called EDHOC for authenticated key exchange targeting IoT environments.The EDHOC protocol performs a compact Diffie-Hellman key exchange handshake, requiring several times less bytes-over-the-air than the de-facto solution used in the Internet, the (D)TLS protocol.In this paper, we study how this reduction in message size correlates with the usage of other scarce resources in IoT environments: time, energy, and memory.We evaluate EDHOC and DTLS with different authentication configurations over two IoT radio technologies.First, we measure the EDHOC and DTLS handshakes on constrained hardware over an IEEE 802.15.4 radio.We observe that EDHOC achieves ×6 to ×14 reduction in packet sizes, ×1.44 improvement in handshake duration and ×2.79 reduction in energy consumed.Next, we simulate time on air on LoRaWAN networks and find that, in the most restrictive configuration (SF=12), DTLS uses at least ×7 more time on air than EDHOC.Finally, we measure flash memory and RAM usage, with the EDHOC implementation achieving a ×4 reduction in both. International audience

International audience; Weed recognition is an essential step for automatic weed control systems. Identifying weeds enables targeted control measures to be implemented, minimizing the use of chemicals and reducing the impact on the environment. Deep learning-based approaches proved to be e↵ective for addressing various complex classification problems. However, to benefit fully from their capabilities, large amounts of labeled data are required, which represents a limitation for agricultural applications, consequence of the tedious and time-consuming process of data labeling. Conversely, unlabeled data could be acquired in large quantities, with relative ease. Hence, our aim is to develop robust and precise deep learning models, to carry-out the recognition and identification of weed species, using both types of data. To this end, we propose a method, that adopts the semi-supervised learning paradigm, to optimally combine labeled and unlabeled data. The method is based on a new deep neural networks architecture, which consists of a modernized convolutional encoder belonging to the family ConvNeXt and a thoroughly designed deep decoder network. This architecture, enables a successful integration of consistency regularization. The conducted experiments on DeepWeeds and 4-Weeds, showed that the semi-supervised models trained through our proposed method provide a stable and high classification performance, compared to other state-of-the-art deep learning models, which were a↵ected negatively by the amount of labeled data available, and the presence of noise during inference. Furthermore, the e↵ectiveness of the proposed method was demonstrated in comparison to other semi-supervised learning methods. The results obtained demonstrate the benefits of adopting the semi-supervised learning paradigm, especially in scenarios with very limited labeled data.

International audience

Monitoring pollution plumes is a key issue, given the harmful effects they cause. The dynamic of these plumes, which may be important due to meteorological conditions, makes their study difficult. Real-time monitoring in order to obtain an accurate mapping of the pollution dispersion is helpful and valuable to mitigate risks. In this work, we consider a fleet of cooperative drones carrying pollution sensors and operating in order to assess a pollution plume. The latter is assumed to follow a Gaussian Process (GP) with varying parameters. For this use case, we propose an efficient approach to characterize spatially and temporarily the plume while optimizing the path planning of drones. In our approach, drones are guided by a Deep Reinforcement Learning (DRL) model called Categorical Deep Q-Network (Categorical DQN) to maximize the plume coverage while considering budget constraints. Specifically, we develop a scalable Independent Q-Learning (IQL) scheme that shares team rewards based on each drone's deployment relevance and therefore ensures cooperation. We evaluate the performance of the plume parameter estimation as well as the maps generated by the GP regression. By testing our framework on several plume scenarios, we show that it offers good results in terms of both estimation quality and run-time efficiency. International audience

Several wildlife species are thought to avoid edges of large habitat gaps, such as clear-cuts, but detailed evidence is rarely available for edges of smaller gaps. We compared the responses of nine wintering mammal species to forest edges in southern Quebec, Canada, using high-resolution spatial data from Light Detection And Ranging (LiDAR) and low-resolution photo-interpretation. We defined edges of open areas as roads, lakes, rivers, or forest open areas. We geolocated mammal snow tracks along systematic transect lines between 2009 and 2018. We compared distances of snow tracks and reference points along transects to the nearest edge with linear models. LiDAR data revealed five species avoiding forest open area edges, whereas no avoidance was shown using photo-interpretation data. Weasels (Mustela sp.) were the only species showing a positive association with forest open area edges using photo-interpreted data. No significant response was detected for river or lake edges. Four species were positively associated with road edges. We conclude that avoidance of small forest open area edges is widespread in our study area, but it can only be detected with high-resolution spatial data. Our results imply that edge effect can operate at a fine scale and using appropriate spatial resolution is crucial to detect such effects.

The economic efficiency of conventional breeding strategies for forest trees based on biparental crosses is compared with that of alternative strategies based on pedigree reconstruction using molecular markers. Analyses of economic efficiency is based on comparisons of breeding scenarios corresponding to the same total investment. The first step is the description and cost evaluation of each basic operation, from crossing to genetic selection and clonal archive establishment. Breeding scenarios are then compared by stochastic sampling with a parametric genetic model (POPSIM), the comparison criteria in this case being genetic gain in the seed orchard for a given level of genetic diversity. Additionally, the economic gain resulting from the use of improved material is estimated for different levels of breeding investment. Our analysis shows that genotyping costs account for a much smaller proportion of total investment than phenotyping costs. We also show that, in comparisons of breeding scenarios corresponding to the same total investment, the three main breeding strategies (biparental crosses, polymix crosses, and open pollination) achieve similar genetic gains provided that sufficiently large numbers of parents are considered. These results open up promising perspectives for the wider integration of molecular markers into forest tree breeding strategies.

This study explores how data from a handheld mobile laser scanning (MLS) system and quantitative structural models (QSM) can be used to estimate tree structural attributes. Four MLS acquisition scenarios were investigated in a 1-ha temperate hardwood stand, including 15 m and 35 m parallel lines, nine circular plots, and a 20 m × 20 m grid. Results were compared against terrestrial laser scanning and destructive field measurements. All acquisition scenarios yielded comparable results, except for the 35 m scenario, which showed greater variability. The 20 m × 20 m grid scenario showed the highest accuracy, with a RMSE of 0.41 m (2.07%) for tree height, 3.98 cm (14.93%) for diameter at breast height, 0.21 m³ (19.28%) for merchantable wood volume, and 0.07 m³ (10.11%) for merchantable stem volume. A bias < 5% was observed for these key attributes, except for an 11.68% bias in merchantable wood volume. Overestimation of branch volume was identified as the primary source of bias related to merchantable wood volume. This study highlights MLS's potential for accurate, non-destructive estimation of tree structural attributes, while pointing out the need to refine noise removal and to assess the most suitable acquisition scenarios for various forest types.

Accurate and fine-scale forest data are essential to improve natural resource management, particularly in the face of climate change. Here, we present SCANFI, the Spatialized CAnadian National Forest Inventory, which provides coherent, 30m resolution 2020 wall-to-wall maps of forest attributes (land cover type, canopy height, crown closure, aboveground tree biomass, main species composition). These maps were developed using the National Forest Inventory (NFI) photo-plot dataset, a systematic regular sample grid of photo-interpreted high-resolution imagery covering all of Canada’s non-arctic landmass. SCANFI was produced using temporally harmonized summer and winter Landsat imagery along with hundreds of tile-level regional models based on a multi-response k-nearest neighbours and random forest imputation method. This approach revealed the importance of radiometric variables in predicting vegetation attributes, namely winter radiometry, as the large-scale climate gradients were controlled at the tile-level. Cross-validation analyses were done, which revealed robust model performance for structural attributes (biomass R2=0.76; crown closure R2=0.82; height R2=0.78) and tree species cover. SCANFI attributes were also validated with several independent external products, ranging from ground plot-based tree species cover to satellite LiDAR height. The methodology can be used to map time series of these attributes and all other additional variables associated with the NFI photo-plots.