Random Forest Classification Approach Research Articles

Characterization of seismic-scale petrofacies variability in the Arbuckle Group using supervised machine learning: Wellington Field, Kansas

The Arbuckle Group in southern Kansas has been investigated for carbon geosequestration-related studies. In this study, we evaluated seismic-scale petrophysically defined facies variability of the Arbuckle Group at the Wellington Field, Kansas, using quantitative seismic interpretation and a supervised Random Forest classification approach. We first defined three petrophysics-based rock types (petrofacies) from core-derived porosity and permeability measurements using the flow-zone indicator (FZI) approach. Then, using the artificial neural network (ANN), we classified these petrofacies in non-cored intervals. We observed that petrofacies 1 corresponds to medium and coarse-grained dolomitic packstone, wackestone, and dolomitic breccia with up to 8% porosity and Darcy-scale permeability values. Whereas petrofacies 2 and 3 correspond to argillaceous and fine-grained micritic dolomites and dolomitic mudstones with lower permeability values for a given porosity, with respect to petrofacies 1. Using the common reflection-point gathers, we performed pre-stack seismic inversion and calculated various amplitude-versus-offset (AVO) attribute volumes. We used these elastic properties and AVO attribute volumes as input for estimating supervised seismic-scale 3D petrofacies and petrofacies probability volumes using the Random Forest algorithm. Results reveal the complex distribution of petrofacies in the candidate injection and baffle zones in the study area, where petrofacies 1 is mainly prevalent within the lower and upper portions of the Arbuckle group, while petrofacies 2 and 3 are mainly present in the middle Arbuckle interval. The workflow we present through this study provides spatial variability of facies distribution that is reflective of actual lithology and petrophysical properties of the Arbuckle group in the study area with limited well control.

Temporal Variability in Stride Kinematics during the Application of TENS: A Machine Learning Analysis.

The purpose of our report was to use a Random Forest classification approach to predict the association between transcutaneous electrical nerve stimulation (TENS) and walking kinematics at the stride level when middle-aged and older adults performed the 6-min test of walking endurance. Data from 41 participants (aged 64.6 ± 9.7 years) acquired in two previously published studies were analyzed with a Random Forest algorithm that focused on upper and lower limb, lumbar, and trunk kinematics. The four most predictive kinematic features were identified and utilized in separate models to distinguish between three walking conditions: burst TENS, continuous TENS, and control. SHAP analysis and linear mixed models were used to characterize the differences among these conditions. Modulation of four key kinematic features - toe-out angle, toe-off angle, and lumbar range of motion (ROM) in coronal and sagittal planes - accurately predicted walking conditions for the burst (82% accuracy) and continuous (77% accuracy) TENS conditions compared with control. Linear mixed models detected a significant difference in lumbar sagittal ROM between the TENS conditions. SHAP analysis revealed that burst TENS was positively associated with greater lumbar coronal ROM, smaller toe-off angle, and less lumbar sagittal ROM. Conversely, continuous TENS was associated with less lumbar coronal ROM and greater lumbar sagittal ROM. Our approach identified four kinematic features at the stride level that could distinguish between the three walking conditions. These distinctions were not evident in average values across strides.

Comparing Pixel- and Object-Based Approaches for Classifying Multispectral Drone Imagery of a Salt Marsh Restoration and Reference Site

Monitoring salt marshes with remote sensing is necessary to evaluate their state and restoration. Determining appropriate techniques for this can be overwhelming. Our study provides insight into whether a pixel- or object-based Random Forest classification approach is best for mapping vegetation in north temperate salt marshes. We used input variables from drone images (raw reflectances, vegetation indices, and textural features) acquired in June, July, and August 2021 of a salt marsh restoration and reference site in Aulac, New Brunswick, Canada. We also investigated the importance of input variables and whether using landcover classes representing areas of change was a practical way to evaluate variation in the monthly images. Our results indicated that (1) the classifiers achieved overall validation accuracies of 91.1–95.2%; (2) pixel-based classifiers outperformed object-based classifiers by 1.3–2.0%; (3) input variables extracted from the August images were more important than those extracted from the June and July images; (4) certain raw reflectances, vegetation indices, and textural features were among the most important variables; and (5) classes that changed temporally were mapped with user’s and producer’s validation accuracies of 86.7–100.0%. Knowledge gained during this study will inform assessments of salt marsh restoration trajectories spanning multiple years.

SPATIAL-TEMPORAL ANALYSIS BETWEEN LANDCOVER CHANGE AND URBAN SURFACE TEMPERATURE OF BEKASI CITY, INDONESIA

Unregulated urban growth can result in a rise in urban population density, leading to the expansion of developed land into suburban regions. The urbanization of Bekasi City inevitably results in the conversion of vegetated land and green open spaces into built-up areas. In addition, there has been a notable rise in the exceptionally high average surface temperature of 12.66 °C during the past 25 years. It is vital to investigate the correlation between landcover change factors and surface temperature, considering these two significant occurrences. This study conducted a spatial-temporal analysis of the relationship between landcover and urban surface temperature in the years 1993, 1998, 2004, 2009, 2018, and 2023. The Random Forest classification approach was employed to acquire comprehensive landcover information, while the remote sensing/image satellite approach was utilized to obtain surface temperature data. The temperature is determined using the thermal channel of satellite photography. The research findings indicate a robust correlation between alterations in land cover, specifically high-density buildings, medium/low-density buildings, and high-density vegetation, and variations in the surface temperature of an urban area. Hence, it is imperative to closely monitor the expansion of land cover to uphold the stability of surface temperature in urban areas.

Evaluating relocation behavior of establishments: Evidence for the short-term effects of COVID-19

Due to the significant effects of establishments’ relocations on travel patterns and land-use conditions, investigating establishments’ relocation behavior is an important issue. In recent years, many establishments closed or relocated because of a downturn economy, health concerns, interrupted supply chains, and work-from-home caused by the pandemic. Hence, this study aims to propose a modeling approach to assess and compare the relocation behavior of establishments before, during, and after the pandemic. Establishments’ relocation behavior is modeled in two steps: relocation decision and relocation action. The former provides insights into behavioral factors associated with establishment relocation and the latter models likelihood of spatial relocation choice. Using the data collected from the state of Tennessee, USA, the Random Forest classification approach is incorporated to model both steps, where the model validation results showed the promising accuracy of this modeling approach. Moreover, statistical analyses are applied to evaluate the differences between the spatial relocation choices throughout the time. Results showed that in post-Covid conditions, the importance of establishment characteristics on relocation decisions was reduced by half and relocations occurred more due to office profile and accessibility. Results of modeling relocation action indicated the high importance of accessibility, even though the attractiveness of accessibility was reduced by 20.9% in post-Covid analysis. The findings of this study enrich the knowledge on establishment relocation behavior and provide valuable information regarding the effect of the pandemic, which can be used in policy development and travel behavior modeling by urban and transportation planners.

Automatic Detection of Phytophthora pluvialis Outbreaks in Radiata Pine Plantations Using Multi-Scene, Multi-Temporal Satellite Imagery

This study demonstrates a framework for using high-resolution satellite imagery to automatically map and monitor outbreaks of red needle cast (Phytophthora pluvialis) in planted pine forests. This methodology was tested on five WorldView satellite scenes collected over two sites in the Gisborne Region of New Zealand’s North Island. All scenes were acquired in September: four scenes were acquired yearly (2018–2020 and 2022) for Wharerata, while one more was obtained in 2019 for Tauwhareparae. Training areas were selected for each scene using manual delineation combined with pixel-level thresholding rules based on band reflectance values and vegetation indices (selected empirically) to produce ‘pure’ training pixels for the different classes. A leave-one-scene-out, pixel-based random forest classification approach was then used to classify all images into (i) healthy pine forest, (ii) unhealthy pine forest or (iii) background. The overall accuracy of the models on the internal validation dataset ranged between 92.1% and 93.6%. Overall accuracies calculated for the left-out scenes ranged between 76.3% and 91.1% (mean overall accuracy of 83.8%), while user’s and producer’s accuracies across the three classes were 60.2–99.0% (71.4–91.8% for unhealthy pine forest) and 54.4–100% (71.9–97.2% for unhealthy pine forest), respectively. This work demonstrates the possibility of using a random forest classifier trained on a set of satellite scenes for the classification of healthy and unhealthy pine forest in new and completely independent scenes. This paves the way for a scalable and largely autonomous forest health monitoring system based on annual acquisitions of high-resolution satellite imagery at the time of peak disease expression, while greatly reducing the need for manual interpretation and delineation.

Identifying the optimal waveband positions for autumn grassland senescence detection using the broadband multispectral remotely sensed dataset

While remote sensing of grass senescence is addressed in the literature, knowledge of optimal waveband positions that are suitable for discriminating between senescent and non-senescent grasses is still limited. Notably, detection of senescent grass is important for understanding the available forage in rangeland environments and associated ecological implications. The free provision of remote sensing data from modern broadband multispectral sensors with improved spatial and spectral properties offers prospects for reliable and wall-to-wall monitoring of grassland senescence in rangeland ecosystems. The current study tested the potential of the modern multispectral remote sensing dataset (i.e., Sentinel 2 and Landsat) in mapping the senescent grass, and to identify the optimal waveband positions that are suitable for discriminating between senescent and non-senescent grasses. Locational information for both senescent and non-senescent grasses was acquired on the field and was used to train the classification process. A Random Forest classification approach was employed using the Landsat 8 and Sentinel 2 multispectral datasets to spectrally discern between senescent and non-senescent grasses. Our analysis yielded overall classification accuracies of 0.82 and 0.78 and kappa coefficients of 0.64 and 0.56 for Sentinel 2 and Landsat 8, respectively. Using the stepwise selection approach, the study further identified that the Red Edge Position (REP), and the visible green and red bands of the electromagnetic spectrum were the optimal waveband positions for separating between senescent and non-senescent grasses based on the broadband multispectral remote sensing. This study has demonstrated the value of the broadband multispectral remote sensing data in mapping autumn grassland senescence, and this lays a foundation for effective operational scale monitoring of foraging resources at the landscape scale, particularly during dry periods.

Block Chain and Machine Learning Models to Evaluate Faults in the Smart Manufacturing System

Smart Manufacturing Systems (SMS) have revolutionized industrial processes by incorporating automation, data analytics, and real-time monitoring to improve efficiency and quality. However, ensuring the reliability and fault tolerance of SMS remains a challenge. This paper proposes an innovative approach that combines Blockchain technology with Machine Learning (ML) models to evaluate faults in SMS. By leveraging the immutability and transparency of the blockchain and the predictive capabilities of ML, this approach enhances fault detection, facilitates traceability, and ultimately contributes to the resilience of smart manufacturing. The industrial sector's increase in data creation has made monitoring systems a crucial idea for management and decision-making. The Internet of Things (IoT), which is sensor-based and one of the most advanced and potent technologies today, can process appropriate ways to monitor the manufacturing process. The research's suggested method combines IoT, machine learning (ML), and monitoring of the industrial system. Temperature, humidity, gyroscope, and accelerometer IoT sensors are used to gather environmental data. Sensor data is produced in unstructured, enormous, and real-time data forms. Many big data approaches are used to process the data further. This system's hybrid prediction model employs the Random Forest classification approach to weed out outliers in the sensor data and aid in defect identification throughout the production process. The suggested approach was examined for South Korean vehicle production. This system uses a strategy to protect and strengthen data trust in order to prevent genuine data changes with fictitious data and system interactions. The efficacy of the suggested methodology in comparison to other methods is provided in the results section. Furthermore, compared to other inputs, the hybrid prediction model offers a respectable fault prediction. The suggested technique is anticipated to improve decision-making and decrease errors during the production process.

Developing a two-decadal time-record of rice field maps using Landsat-derived multi-index image collections with a random forest classifier: A Google Earth Engine based approach

Historic maps showing the temporal distribution of rice fields are important for precision agriculture, irrigation optimisation, forecasting crop yields, land use management and formulating policies. However, mapping rice fields using traditional ground surveys is impractical when high cost, time and labour requirements are considered, and the availability of such detailed records is limited. Although satellite remote sensing appears to be a viable solution, conventional segmentation and classification methods with spectral bands are often unable to contrast the distinct characteristics between rice fields and other vegetation classes. To this end, we explored a novel, Google Earth Engine (GEE) based multi-index random forest (RF) classification approach to map rice fields over two decades. Landsat images from 2000 to 2020 of two Sri Lankan rice cultivation districts were extracted from GEE and a multi-index RF classification algorithm was applied to distinguish the rice fields. The results showed above 80% accuracy for both training and validation, when compared against high spatial resolution Google Earth imagery. In essence, multi-index sampling and RF together synergised the compelling classification accuracy by effectively capturing vegetation, water (ponding) and soil characteristics unique to the rice fields using a single-click approach. The maps developed in this study were further compared against the MODIS land cover type product (MCD12Q1) and the corresponding superior statistics on rice fields demonstrated the robustness of the proposed approach. Future work seeking effective index combinations is recommended, and this approach can potentially be extended to other crop analyses elsewhere.

Harmonizing and Extending Fragmented 100 Year Flood Hazard Maps in Canada’s Capital Region Using Random Forest Classification

With the record breaking flood experienced in Canada’s capital region in 2017 and 2019, there is an urgent need to update and harmonize existing flood hazard maps and fill in the spatial gaps between them to improve flood mitigation strategies. To achieve this goal, we aim to develop a novel approach using machine learning classification (i.e., random forest). We used existing fragmented flood hazard maps along the Ottawa River to train a random forest classification model using a range of flood conditioning factors. We then applied this classification across the Capital Region to fill in the spatial gaps between existing flood hazard maps and generate a harmonized high-resolution (1 m) 100 year flood susceptibility map. When validated against recently produced 100 year flood hazard maps across the capital region, we find that this random forest classification approach yields a highly accurate flood susceptibility map. We argue that the machine learning classification approach is a promising technique to fill in the spatial gaps between existing flood hazard maps and create harmonized high-resolution flood susceptibility maps across flood-vulnerable areas. However, caution must be taken in selecting suitable flood conditioning factors and extrapolating classification to areas with similar characteristics to the training sites. The resulted harmonized and spatially continuous flood susceptibility map has wide-reaching relevance for flood mitigation planning in the capital region. The machine learning approach and flood classification optimization method developed in this study is also a first step toward Natural Resources Canada’s aim of creating a spatially continuous flood susceptibility map across the Ottawa River watershed. Our modeling approach is transferable to harmonize flood maps and fill in spatial gaps in other regions of the world and will help mitigate flood disasters by providing accurate flood data for urban planning.

Biofluid Metabolomics and Lipidomics of Mice Exposed to External Very High-Dose Rate Radiation

High-throughput biodosimetry methods to determine exposure to ionizing radiation (IR) that can also be easily scaled to multiple testing sites in emergency situations are needed in the event of malicious attacks or nuclear accidents that may involve a substantial number of civilians. In the event of an improvised nuclear device (IND), a complex IR exposure will have a very high-dose rate (VHDR) component from an initial blast. We have previously addressed low-dose rate (LDR, ≤1 Gy/day) exposures from internal emitters on biofluid small molecule signatures, but further research on the VHDR component of the initial blast is required. Here, we exposed 8- to 10-week-old male C57BL/6 mice to an acute dose of 3 Gy using a reference dose rate of 0.7 Gy/min or a VHDR of 7 Gy/s, collected urine and serum at 1 and 7 d, then compared the metabolite signatures using either untargeted (urine) or targeted (serum) approaches with liquid chromatography mass spectrometry platforms. A Random Forest classification approach showed strikingly similar changes in urinary signatures at 1 d post-irradiation with VHDR samples grouping closer to control samples at 7 d. Identical metabolite panels (carnitine, trigonelline, xanthurenic acid, N6,N6,N6-trimethyllysine, spermine, and hexosamine-valine-isoleucine-OH) could differentiate IR exposed individuals with high sensitivity and specificity (area under the receiver operating characteristic (AUROC) curves 0.89–1.00) irrespective of dose rate at both days. For serum, the top 25 significant lipids affected by IR exposure showed slightly higher perturbations at 0.7 Gy/min vs. 7 Gy/s; however, identical panels showed excellent sensitivity and specificity at 1 d (three hexosylceramides (16:0), (18:0), (24:0), sphingomyelin [26:1], lysophosphatidylethanolamine [22:1]). Mice could not be differentiated from control samples at 7 d for a 3 Gy exposure based on serum lipid signatures. As with LDR exposures, we found that identical biofluid small molecule signatures can identify IR exposed individuals irrespective of dose rate, which shows promise for more universal applications of metabolomics for biodosimetry.

Solar park detection from publicly available satellite imagery

ABSTRACT The rapid increase in large-scale photovoltaic installations, or solar parks, causes a need to monitor their amount and allocation, and assess their impacts. While their spectral signature suggests that solar parks can be identified among other land covers, this detection is challenged by their low occurrence. Here, we develop an object-based random forest (RF) classification approach, using publicly available satellite imagery, which has the advantage of requiring relatively little training data and being easily extendable to large spatial extents and new areas. First, we segmented Sentinel-2 imagery into homogenous objects using a Simple Non-Iterative Clustering algorithm in Google Earth Engine. Thereafter, we calculated for each object the mean, standard deviation, and median for all 10- and 20-meter resolution bands of Sentinel-1 and Sentinel-2. These features are subsequently used to train and validate a range of RF models to select the most promising model setup. The training datasets consisted of subsampled presence/absence data, oversampled presence/absence data, and multiple land-cover categories. The best-performing model used an oversampled dataset trained on all 10- and 20- meter resolution spectral bands and the radar backscatter properties of one period. Independent test results show an overall classification accuracy of 99.97% (Kappa: 0.90). For this result, the producer accuracy was 85.86% for solar park objects and 99.999% for non-solar park objects. The user accuracy was 92.39% for solar park objects and 99.999% for non-solar park objects. These high classification accuracies indicate that our approach is suitable for transfer learning and is able to detect solar parks in new study areas.

Vegetation Mapping in the Permafrost Region: A Case Study on the Central Qinghai-Tibet Plateau

An accurate and detailed vegetation map is of crucial significance for understanding the spatial heterogeneity of subsurfaces, which can help to characterize the thermal state of permafrost. The absence of an alpine swamp meadow (ASM) type, or an insufficient resolution (usually km-level) to capture the spatial distribution of the ASM, greatly limits the availability of existing vegetation maps in permafrost modeling of the Qinghai-Tibet Plateau (QTP). This study generated a map of the vegetation type at a spatial resolution of 30 m on the central QTP. The random forest (RF) classification approach was employed to map the vegetation based on 319 ground-truth samples, combined with a set of input variables derived from the visible, infrared, and thermal Landsat-8 images. Validation using a train-test split (i.e., 70% of the samples were randomly selected to train the RF model, while the remaining 30% were used for validation and a total of 1000 runs) showed that the average overall accuracy and Kappa coefficient of the RF approach were 0.78 (0.68–0.85) and 0.69 (0.64–0.74), respectively. The confusion matrix showed that the overall accuracy and Kappa coefficient of the predicted vegetation map reached 0.848 (0.844–0.852) and 0.790 (0.785–0.796), respectively. The user accuracies for the ASM, alpine meadow, alpine steppe, and alpine desert were 95.0%, 83.3%, 82.4%, and 86.7%, respectively. The most important variables for vegetation type prediction were two vegetation indices, i.e., NDVI and EVI. The surface reflectance of visible and shortwave infrared bands showed a secondary contribution, and the brightness temperature and the surface temperature of the thermal infrared bands showed little contribution. The dominant vegetation in the study area is alpine steppe and alpine desert. The results of this study can provide an accurate and detailed vegetation map, especially for the distribution of the ASM, which can help to improve further permafrost studies.

Satellite-based observations of the green depressing cropping system in a farming-pastoral ecotone of northern China

The green depressing cropping system (GDCS) has existed in the farming-pastoral ecotone of northern China (FPENC) for hundreds of years. This ancient practice aims to promote an agricultural system that is both green and sustainable at the same time. However, there is a knowledge gap regarding the extent and implementation of the GDCS in China. Here we present a study that analyses GDCS fields and their related farmland use patterns in Ulanqab, China. This study takes into consideration satellite imagery, ground truth data, phenological records, and other ancillary data from 2015 to 2019 at the Ulanqab region. The integration of the former datasets, combined with visual interpretation and a random forest classification approach, successfully mapped the distribution of GDCS fields, GDCS frequencies, GDCS-crop rotation, and the determinants (terrain, irrigation) of the distribution of the GDCS practice. Our results suggest that GDCS is still prevailing and relatively widespread in Ulanqab, especially in the Houshan area, where a total of 10.29% of farmlands in the study region have adopted the practice. Cereal crops, potato, and other drought-resistant crops were the preferred previous and subsequent crops of GDCS, rather than maize and sunflowers. According to our results, 91.5% of the GDCS fields were rainfed in Ulanqab, suggesting that the practice is associated with dry farming. We also found that the practice is particularly prevalent in the fields with a steeper slope. Our findings provide empirical evidence for the importance of remote sensing towards the implementation of pilot programs that involve crop rotation and fallow systems.

Sentence polarity detection using stepwise greedy correlation based feature selection and random forests: An fMRI study

Cognitive state analysis or reading the brain was always an exciting field for researchers. Analysis of the human brain while a person is engaged in doing a particular task is an essential topic in the recent development of neuro-imaging studies. The introduction of new non-invasive methods like PET (Positron Emission Tomography) and fMRI (Functional Magnetic Resonance Imaging) is used to analyze the brain. At the same time, subjects are involved in doing diverse activities. This study aims to investigate the processing of affirmative and negative sentences in the brain. Using a greedy stepwise correlation-based feature selection technique and random forest classification approach, our model can classify the cognitive state in sentence polarity detection task with, on average, 95.41% accuracy. We have also analyzed the category-specific selected feature voxel set in determining the sentence polarity in the brain. Our result shows that CALC, RDLPFC, and LDLPFC are positively contributing areas feature selection. In contrast, RPPREC, RSGA, RFEF add very little to polarity check.

Tree Species Classification in Mixed Deciduous Forests Using Very High Spatial Resolution Satellite Imagery and Machine Learning Methods

Spatially explicit information on tree species composition is important for both the forest management and conservation sectors. In combination with machine learning algorithms, very high-resolution satellite imagery may provide an effective solution to reduce the need for labor-intensive and time-consuming field-based surveys. In this study, we evaluated the possibility of using multispectral WorldView-3 (WV-3) satellite imagery for the classification of three main tree species (Quercus robur L., Carpinus betulus L., and Alnus glutinosa (L.) Geartn.) in a lowland, mixed deciduous forest in central Croatia. The pixel-based supervised classification was performed using two machine learning algorithms: random forest (RF) and support vector machine (SVM). Additionally, the contribution of gray level cooccurrence matrix (GLCM) texture features from WV-3 imagery in tree species classification was evaluated. Principal component analysis confirmed GLCM variance to be the most significant texture feature. Of the 373 visually interpreted reference polygons, 237 were used as training polygons and 136 were used as validation polygons. The validation results show relatively high overall accuracy (85%) for tree species classification based solely on WV-3 spectral characteristics and the RF classification approach. As expected, an improvement in classification accuracy was achieved by a combination of spectral and textural features. With the additional use of GLCM variance, the overall accuracy improved by 10% and 7% for RF and SVM classification approaches, respectively.

Analysis of microplastics of a broad size range in commercially important mussels by combining FTIR and Raman spectroscopy approaches.

Microplastic (MP) contamination is present in the entire marine environment from the sediment to the water surface and down to the deep sea. This ubiquitous presence of MP particles opens the possibility for their ingestion by nearly all species in the marine ecosystem. Reports have shown that MP particles are present in local commercial seafood species leading to the possible human ingestion of these particles. However, due to a lack of harmonized methods to identify microplastics (MPs), results from different studies and locations can hardly be compared. Hence, this study was aimed to detect, quantify, and estimate MP contamination in commercially important mussels originating from 12 different countries distributed worldwide. All mussels were obtained from supermarkets and were intended for human consumption. Using a combinatorial approach of focal plane array (FPA)-based micro- Fourier-transform infrared (FTIR) spectroscopy and micro-Raman spectroscopy allowed the detection and characterization of MP down to a size of 3μm in the investigated mussels. Further, a gentle sample purification method based on enzymes has been modified in order to optimize the digestion of organic material in mussels. A random forest classification (RFC) approach, which allows a rapid discrimination between different polymer types and thus fast generation of data on MP abundance and size distributions with high accuracy, was implemented in the analytical pipeline for IR spectra. Additionally, for the first time we also applied a RFC approach for the automated characterization of Raman spectra of MPs.

Toward the probabilistic forecasting of cyclone-induced marine flooding by overtopping at Reunion Island aided by a time-varying random-forest classification approach

In 2017, Irma and Maria highlighted the vulnerability of small islands to cyclonic events and the necessity of advancing the forecast techniques for cyclone-induced marine flooding. In this context, this paper presents a generic approach to deriving time-varying inundation forecasts from ensemble track and intensity forecasts applied to the case of Reunion Island in the Indian Ocean. The challenge for volcanic islands is to account for the full complexity of wave overtopping processes while also ensuring a robustness and timeliness that are compatible with emergency requirements. The challenge is addressed by following a hybrid approach relying on the combination of process-based models with a statistical model (herein, a random-forest classifier) trained with a precalculated database. The latter enables one to translate any time series of coastal marine conditions into the time-varying probability of inundation for different sectors. The application detailed for the case of Cyclone Dumile at Sainte-Suzanne city shows that the proposed approach enables quick discrimination, in both space and time, thereby identifying safe and exposed areas and demonstrating that probabilistic forecasting of marine flooding by overtopping is feasible. The whole method can be easily adapted to other territories and scales provided that validated process-based models are available. Beyond early warning applications, the developed database and statistical models may also be used for informing risk prevention and adaptation strategies.

COMBINING ENVIRONMENTAL AND LANDSAT ANALYSIS READY DATA FOR VEGETATION MAPPING: A CASE STUDY IN THE BRAZILIAN SAVANNA BIOME

Abstract. The Cerrado biome in Brazil covers approximately 24% of the country. It is one of the richest and most diverse savannas in the world, with 23 vegetation types (physiognomies) consisting mostly of tropical savannas, grasslands, forests and dry forests. It is considered as one of the global hotspots of biodiversity because of the high level of endemism and rapid loss of its original habitat. This work aims to analyze the potential of Landsat Analysis Ready Data (ARD) in combination with different environmental data to classify the vegetation in the Cerrado in two different hierarchical levels. Here we present results of a pixel-based modelling exercise, in which field data were combined with a set of input variables using a Random Forest classification approach. On the first hierarchical level, with the three classes savanna, grasslands and forest, our model results reached f1-scores of 0.86, 0.87 and 0.85 leading to an overall accuracy of 0.86. In the second hierarchical level we differentiated a total of 12 vegetation physiognomies with an overall accuracy of 0.77.

Validating the Use of Bovine Buccal Sampling as a Proxy for the Rumen Microbiota by Using a Time Course and Random Forest Classification Approach.

Analysis of the cow microbiome, as well as host genetic influences on the establishment and colonization of the rumen microbiota, is critical for development of strategies to manipulate ruminal function toward more efficient and environmentally friendly milk production. To this end, the development and validation of noninvasive methods to sample the rumen microbiota at a large scale are required. In this study, we further optimized the analysis of buccal swab samples as a proxy for direct bacterial samples of the rumen of dairy cows. To identify an optimal time for sampling, we collected buccal swab and rumen samples at six different time points relative to animal feeding. We then evaluated several biases in these samples using a machine learning classifier (random forest) to select taxa that discriminate between buccal swab and rumen samples. Differences in the inverse Simpson's diversity, Shannon's evenness, and Bray-Curtis dissimilarities between methods were significantly less apparent when sampling was performed prior to morning feeding (P < 0.05), suggesting that this time point was optimal for representative sampling. In addition, the random forest classifier was able to accurately identify nonrumen taxa, including 10 oral and putative feed-associated taxa. Two highly prevalent (>60%) taxa in buccal and rumen samples had significant variance in relative abundances between sampling methods but could be qualitatively assessed via regular buccal swab sampling. This work not only provides new insights into the oral community of ruminants but also further validates and refines buccal swabbing as a method to assess the rumen bacterial in large herds.IMPORTANCE The gastrointestinal tracts of ruminants harbor a diverse microbial community that coevolved symbiotically with the host, influencing its nutrition, health, and performance. While the influence of environmental factors on rumen microbes is well documented, the process by which host genetics influences the establishment and colonization of the rumen microbiota still needs to be elucidated. This knowledge gap is due largely to our inability to easily sample the rumen microbiota. There are three common methods for rumen sampling but all of them present at least one disadvantage, including animal welfare, sample quality, labor, and scalability. The development and validation of noninvasive methods, such as buccal swabbing, for large-scale rumen sampling is needed to support studies that require large sample sizes to generate reliable results. The validation of buccal swabbing will also support the development of molecular tools for the early diagnosis of metabolic disorders associated with microbial changes in large herds.