This study aims to simplify the current flood mapping methodology, so it can be done in quicker time, easy to learn, but still produce accurate flood information to support the emergency responses. This study is divided into two activities: Survey development and flood mapping methods; Field survey and mapping to test the developed method. The method development was then conducted by examining the simple methods in the flood mapping survey and then testing it in the field. The results of these trials were then evaluated to determine the most effective and efficient methods. The utilization of free android mapping application to conduct flood survey shows satisfactory results. Based on two trials, it was known that it takes only 4 hours to conduct a 15 km survey of Citarum River segment. The mapping shows that the flood areas in Baleendah and its surrounding reached 763 Ha and 794 Ha respectively on 25 December 2014 and 13 March 2016. In addition, the method developed is also relatively easy to use, so it is expected to trigger the local communities to play an active role in disaster prevention efforts, especially in emergency response by providing accurate information about the flood inundation areas.

Anthropogenic changes of the biota and human hyper-dominance are modulating the evolution of life on our planet. Humankind has spread worldwide supported by cultural and technological knowledge, and has already modified uncountable biological interactions. While numerous species have been extinguished by human actions, others are directly favored, such as alien species, hybrids, and genetically modified organisms. These biodiversity shifts have generated new interactions among all living organisms in anthropized or anthropogenic ecosystems, with the consequent establishment of novel evolutionary pathways. Thus, humans have created a strong evolutionary bias on Earth, leading to unexpected and irreversible outcomes. Anthropogenic changes and novelty organisms are shifting the evolutionary paths of all organisms towards the Biological Anthropocene, a new concept of our imprint on biodiversity and evolution.

Rapid damage assessment is of crucial importance to emergency responders during hurricane events, however, the evaluation process is often slow, labor-intensive, costly, and error-prone. New advances in computer vision and remote sensing open possibilities to observe the Earth at a different scale. However, substantial pre-processing work is still required in order to apply state-of-the-art methodology for emergency response. To enable the comparison of methods for automatic detection of damaged buildings from post-hurricane remote sensing imagery taken from both airborne and satellite sensors, this paper presents the development of benchmark datasets from publicly available data. The major contributions of this work include (1) a scalable framework for creating benchmark datasets of hurricane-damaged buildings and (2) public sharing of the resulting benchmark datasets for Greater Houston area after Hurricane Harvey in 2017. The proposed approach can be used to build other hurricane-damaged building datasets on which researchers can train and test object detection models to automatically identify damaged buildings.

Pixel-wise classification in remote sensing identifies entities in large-scale satellite-based images at the pixel level. Few fully annotated large-scale datasets for pixel-wise classification exist due to the challenges of annotating individual pixels. Training data scarcity inevitably ensues from the annotation challenge, leading to overfitting classifiers and degraded classification performance. The lack of annotated pixels also necessarily results in few hard examples of various entities critical for generating a robust classification hyperplane. To overcome the problem of the data scarcity and lack of hard examples in training, we introduce a two-step hard example generation (HEG) approach that first generates hard example candidates and then mines actual hard examples. In the first step, a generator that creates hard example candidates is learned via the adversarial learning framework by fooling a discriminator and a pixel-wise classification model at the same time. In the second step, mining is performed to build a fixed number of hard examples from a large pool of real and artificially generated examples. To evaluate the effectiveness of the proposed HEG approach, we design a 9-layer fully convolutional network suitable for pixel-wise classification. Experiments show that using generated hard examples from the proposed HEG approach improves the pixel-wise classification model's accuracy on red tide detection and hyperspectral image classification tasks.

In this work, we describe our approach to pneumonia classification and localization in chest radiographs. This method uses only \emph{open-source} deep learning object detection and is based on CoupleNet, a fully convolutional network which incorporates global and local features for object detection. Our approach achieves robustness through critical modifications of the training process and a novel ensembling algorithm which merges bounding boxes from several models. We tested our detection algorithm tested on a dataset of 3000 chest radiographs as part of the 2018 RSNA Pneumonia Challenge; our solution was recognized as a winning entry in a contest which attracted more than 1400 participants worldwide.

This paper focuses on YOLO-LITE, a real-time object detection model developed to run on portable devices such as a laptop or cellphone lacking a Graphics Processing Unit (GPU). The model was first trained on the PASCAL VOC dataset then on the COCO dataset, achieving a mAP of 33.81% and 12.26% respectively. YOLO-LITE runs at about 21 FPS on a non-GPU computer and 10 FPS after implemented onto a website with only 7 layers and 482 million FLOPS. This speed is 3.8x faster than the fastest state of art model, SSD MobilenetvI. Based on the original object detection algorithm YOLOV2, YOLO- LITE was designed to create a smaller, faster, and more efficient model increasing the accessibility of real-time object detection to a variety of devices.

Feature pyramids are widely exploited by both the state-of-the-art one-stage object detectors (e.g., DSSD, RetinaNet, RefineDet) and the two-stage object detectors (e.g., Mask R-CNN, DetNet) to alleviate the problem arising from scale variation across object instances. Although these object detectors with feature pyramids achieve encouraging results, they have some limitations due to that they only simply construct the feature pyramid according to the inherent multi-scale, pyramidal architecture of the backbones which are actually designed for object classification task. Newly, in this work, we present a method called Multi-Level Feature Pyramid Network (MLFPN) to construct more effective feature pyramids for detecting objects of different scales. First, we fuse multi-level features (i.e. multiple layers) extracted by backbone as the base feature. Second, we feed the base feature into a block of alternating joint Thinned U-shape Modules and Feature Fusion Modules and exploit the decoder layers of each u-shape module as the features for detecting objects. Finally, we gather up the decoder layers with equivalent scales (sizes) to develop a feature pyramid for object detection, in which every feature map consists of the layers (features) from multiple levels. To evaluate the effectiveness of the proposed MLFPN, we design and train a powerful end-to-end one-stage object detector we call M2Det by integrating it into the architecture of SSD, which gets better detection performance than state-of-the-art one-stage detectors. Specifically, on MS-COCO benchmark, M2Det achieves AP of 41.0 at speed of 11.8 FPS with single-scale inference strategy and AP of 44.2 with multi-scale inference strategy, which is the new state-of-the-art results among one-stage detectors. The code will be made available on \url{https://github.com/qijiezhao/M2Det.

The ice phase in clouds contributes largely to uncertainties in global climate models partly due to a lack of atmospheric observations. At moderate supercooling ice nucleating particles (INP) and ice particles (IP) are present in small concentrations and a large volume of air is necessary for observation. Here, we report on initial observations of IP in supercooled fog with a new setup. We use a 0.3 m wide, vertical curtain of light and a camera pointing perpendicularly at it to record light scattered by IP formed in radiation fog near the ground at temperatures between -3 {\deg}C and -9 {\deg}C. Deposition rates of IP were several times larger than expected from number concentrations of INP found on $PM_{10}$-filters at a nearby air quality monitoring station. The discrepancy might be explained by secondary ice formation through the fragmentation of freezing water droplets, the entrainment of INP from above the fog layer through settling, the loss or deactivation of INP on $PM_{10}$-filters prior to analysis, or radiative cooling of INP below the temperature of the surrounding air. In summary, radiation fog constitutes an easily accessible form of a supercooled cloud, in which observations can be made for long enough to quantify the deposition rate of rare IP produced in a natural environment.

The work in this paper is motivated by the increasing use of electrical and electromagnetic methods in geoscience problems where steel-cased wells are present. Applications of interest include monitoring carbon capture and storage and hydraulic fracturing operations, as well as detecting flaws or breaks in degrading steel-casings -- such wells pose serious environmental hazards. The general principles of electrical methods with steel-cased wells are understood, and several authors have demonstrated that the presence of steel-cased wells can be beneficial for detecting signal due to targets at depth. However, the success of a DC resistivity survey lies in the details. Secondary signals might only be a few percent of the primary signal. In designing a survey, the geometry of the source and receivers, and whether the source is at the top of the casing, inside of it, or beneath the casing will impact measured responses. Also the physical properties and geometry of the background geology, target, and casing will have a large impact on the measured data. Because of the small values of the diagnostic signals, it is important to understand the detailed physics of the problem and also to be able to carry out accurate simulations. This latter task is computationally challenging because of the extreme geometry of the wells, which extend kilometers in depth but have millimeter variations in the radial direction, and the extreme variation in the electrical conductivity (typically 5-7 orders of magnitude between the casing and the background geology).

Models that envisage successful subduction channel transport of upper crustal materials below 300 km depth, past a critical phase transition in buoyant crustal lithologies, are capable of accumulating and assembling these materials into so-called "second continents" that are gravitationally stabilized at the base of the Transition Zone, at some 600 to 700 km depth. Global scale, Pacific-type subduction (ocean-ocean and ocean-continent convergence), which lead to super continent assembly, were hypothesized to produce second continents that scale to about the size of Australia, with continental upper crustal concentration levels of radiogenic power. Seismological techniques are incapable of imaging these second continents because of their negligible difference in seismic wave velocities with the surrounding mantle. We can image the geoneutrino flux linked to the radioactive decays in these second continents with land and/or ocean-based detectors. We present predictions of the geoneutrino flux of second continents, assuming different scaled models and we discuss the potential of current and future neutrino experiments to discover or constrain second continents. The power emissions from second continents were proposed to be drivers of super continental cycles. Thus, testing models for the existence of second continents will place constraints on mantle and plate dynamics when using land and ocean-based geoneutrino detectors deployed at strategic locations.

We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models, BERT is designed to pre-train deep bidirectional representations from unlabeled text by jointly conditioning on both left and right context in all layers. As a result, the pre-trained BERT model can be fine-tuned with just one additional output layer to create state-of-the-art models for a wide range of tasks, such as question answering and language inference, without substantial task-specific architecture modifications. BERT is conceptually simple and empirically powerful. It obtains new state-of-the-art results on eleven natural language processing tasks, including pushing the GLUE score to 80.5% (7.7% point absolute improvement), MultiNLI accuracy to 86.7% (4.6% absolute improvement), SQuAD v1.1 question answering Test F1 to 93.2 (1.5 point absolute improvement) and SQuAD v2.0 Test F1 to 83.1 (5.1 point absolute improvement).

One of the defining features of both recent and historical cases of global climate change is Arctic Amplification (AA). This is the more rapid change in the surface air temperature (SAT) in the Arctic compared to some wider reference region, such as the Northern Hemisphere (NH) mean. Many different metrics have been developed to quantify the degree of AA based on SAT anomalies, trends and variability. The use of different metrics, as well as the choice of dataset to use can affect conclusions about the magnitude and temporal variability of AA. Here we review the established metrics of AA to see how well they agree upon the temporal signature of AA, such as the multi-decadal variability, and assess the consistency in these metrics across different commonly-used datasets which cover both the early and late 20th century warming in the Arctic. We find the NOAA 20th Century Reanalysis most closely matches the observations when using metrics based upon SAT trends (A2), variability (A3) and regression (A4) of the SAT anomalies, and the ERA 20th Century Reanalysis is closest to the observations in the SAT anomalies (A1) and variability of SAT anomalies (A3). However, there are large seasonal differences in the consistency between datasets. Moreover, the largest differences between the century-long reanalysis products and observations are during the early warming period, likely due to the sparseness of the observations in the Arctic at that time. In the modern warming period, the high density of observations strongly constrains all the reanalysis products, whether they include satellite observations or only surface observations. Thus, all the reanalysis and observation products produce very similar magnitudes and temporal variability in the degree of AA during the recent warming period.

Fishing activities have broad impacts that affect, although not exclusively, the targeted stocks. These impacts affect predators and prey of the harvested species, as well as the whole ecosystem it inhabits. Ecosystem models can be used to study the interactions that occur within a system, including those between different organisms and those between fisheries and targeted species. Trophic web models like Ecopath with Ecosim (EwE) can handle fishing fleets as a top predator, with top-down impact on harvested organisms. The aim of this study was to better understand the Icelandic marine ecosystem and the interactions within. This was done by constructing an EwE model of Icelandic waters. The model was run from 1984 to 2013 and was fitted to time series of biomass estimates, landings data and mean annual temperature. The final model was chosen by selecting the model with the lowest Akaike information criterion. A skill assessment was performed using the Pearson's correlation coefficient, the coefficient of determination, the modelling efficiency and the reliability index to evaluate the model performance. The model performed satisfactorily when simulating previously estimated biomass and known landings. Most of the groups with time series were estimated to have top-down control over their prey. These are harvested species with direct and/or indirect links to lower trophic levels and future fishing policies should take this into account. This model could be used as a tool to investigate how such policies could impact the marine ecosystem in Icelandic waters.

Being able to effectively identify clouds and monitor their evolution is one important step toward more accurate quantitative precipitation estimation and forecast. In this study, a new gradient-based cloud-image segmentation technique is developed using tools from image processing techniques. This method integrates morphological image gradient magnitudes to separable cloud systems and patches boundaries. A varying scale-kernel is implemented to reduce the sensitivity of image segmentation to noise and capture objects with various finenesses of the edges in remote-sensing images. The proposed method is flexible and extendable from single- to multi-spectral imagery. Case studies were carried out to validate the algorithm by applying the proposed segmentation algorithm to synthetic radiances for channels of the Geostationary Operational Environmental Satellites (GOES-R) simulated by a high-resolution weather prediction model. The proposed method compares favorably with the existing cloud-patch-based segmentation technique implemented in the PERSIANN-CCS (Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network - Cloud Classification System) rainfall retrieval algorithm. Evaluation of event-based images indicates that the proposed algorithm has potential to improve rain detection and estimation skills with an average of more than 45% gain comparing to the segmentation technique used in PERSIANN-CCS and identifying cloud regions as objects with accuracy rates up to 98%.

Detecting small objects over large areas remains a significant challenge in satellite imagery analytics. Among the challenges is the sheer number of pixels and geographical extent per image: a single DigitalGlobe satellite image encompasses over 64 km2 and over 250 million pixels. Another challenge is that objects of interest are often minuscule (~pixels in extent even for the highest resolution imagery), which complicates traditional computer vision techniques. To address these issues, we propose a pipeline (SIMRDWN) that evaluates satellite images of arbitrarily large size at native resolution at a rate of > 0.2 km2/s. Building upon the tensorflow object detection API paper, this pipeline offers a unified approach to multiple object detection frameworks that can run inference on images of arbitrary size. The SIMRDWN pipeline includes a modified version of YOLO (known as YOLT), along with the models of the tensorflow object detection API: SSD, Faster R-CNN, and R-FCN. The proposed approach allows comparison of the performance of these four frameworks, and can rapidly detect objects of vastly different scales with relatively little training data over multiple sensors. For objects of very different scales (e.g. airplanes versus airports) we find that using two different detectors at different scales is very effective with negligible runtime cost.We evaluate large test images at native resolution and find mAP scores of 0.2 to 0.8 for vehicle localization, with the YOLT architecture achieving both the highest mAP and fastest inference speed.

Proximal gamma-ray spectroscopy supported by adequate calibration and correction for growing biomass is an effective field scale technique for a continuous monitoring of top soil water content dynamics to be potentially employed as a decision support tool for automatic irrigation scheduling. This study demonstrates that this approach has the potential to be one of the best space-time trade-off methods, representing a joining link between punctual and satellite fields of view. The inverse proportionality between soil moisture and gamma signal is theoretically derived taking into account a non-constant correction due to the presence of growing vegetation beneath the detector position. The gamma signal attenuation due to biomass is modelled with a Monte Carlo-based approach in terms of an equivalent water layer which thickness varies in time as the crop evolves during its life-cycle. The reliability and effectiveness of this approach is proved through a 7 months continuous acquisition of terrestrial gamma radiation in a 0.4 ha tomato (Solanum lycopersicum) test field. We demonstrate that a permanent gamma station installed at an agricultural field can reliably probe the water content of the top soil only if systematic effects due to the biomass shielding are properly accounted for. Biomass corrected experimental values of soil water content inferred from radiometric measurements are compared with gravimetric data acquired under different soil moisture levels, resulting in an average percentage relative discrepancy of about 3% in bare soil condition and of 4% during the vegetated period. The temporal evolution of corrected soil water content values exhibits a dynamic range coherent with the soil hydraulic properties in terms of wilting point, field capacity and saturation.

Object detection-the computer vision task dealing with detecting instances of objects of a certain class (e.g., 'car', 'plane', etc.) in images-attracted a lot of attention from the community during the last 5 years. This strong interest can be explained not only by the importance this task has for many applications but also by the phenomenal advances in this area since the arrival of deep convolutional neural networks (DCNN). This article reviews the recent literature on object detection with deep CNN, in a comprehensive way, and provides an in-depth view of these recent advances. The survey covers not only the typical architectures (SSD, YOLO, Faster-RCNN) but also discusses the challenges currently met by the community and goes on to show how the problem of object detection can be extended. This survey also reviews the public datasets and associated state-of-the-art algorithms.

The accurate and automated determination of earthquake locations is still a challenging endeavor. However, such information is critical for monitoring seismic activity and assessing potential hazards in real time. Recently, a convolutional neural network was applied to detect earthquakes from single-station waveforms and approximately map events across several large surface areas. In this study, we locate 194 earthquakes induced during oil and gas operations in Oklahoma, USA, within an error range of approximately 4.9 km on average to the epicenter and 1.0 km to the depth in catalogs with data from 30 network stations by applying the fully convolutional network. The network is trained by 1,013 historic events, and the output is a 3D volume of the event location probability in the Earth. The trained system requires approximately one hundredth of a second to locate an event without the need for any velocity model or human interference.

We are increasingly using nature for tourism and recreation, an economic sector now generating more than 10% of the global GDP and 10% of global total employment. This growth though has come at a cost and we now have 5930 species for which tourism and recreation are conservation threats. For the first time we use global social media data to estimate where people go to experience nature and determine how this tourism and recreation pressure overlap with the distribution of threatened species. The more people seek interactions with nature in an area, the larger the number of species threatened by those interactions is. Clear crisis areas emerge where many species sensitive to tourism are exposed to high tourism pressures and those are mainly coastal marine regions. Our current tourism management approaches are not achieving biodiversity conservation. The global increase in nature tourism and recreation is set to continue and we need a global consistent response to mitigate its biodiversity impact. Like with other extractive industries, we must prioritise our efforts to diverge tourism away from crisis areas.

Hydrological models of karst aquifers are often semi-distributed, and physical processes such as infiltration and spring discharge generation are described in a lumped way. Several works have previously addressed the problems associated with the calibration of such models, highlighting in particular the issue of model parameter estimation and model equifinality. In this work, we investigate the problem of model calibration using the active subspace (AS) method, a novel tool for model parameter dimension reduction. We apply the method to a newly proposed hydrological model for karst aquifers, LuKARS, to investigate if the AS framework identifies catchment-specific characteristics or if the results only depend on the chosen model structure. Therefore, we consider four different case studies, three synthetic and one real case (Kerschbaum springshed in Waidhofen a.d. Ybbs, Austria), with varying hydrotope distributions and properties. We find that both the hydrotope area coverage and the catchment characteristics have major impacts on parameter sensitivities. While model parameters are similarly informed in scenarios with less varying catchment characteristics, we find significant differences in parameter sensitivities when the applied hydrotopes were different from each other. Our results show that the AS method can be used to investigate the relation between the model structure, the area of a hydrotope, the physical properties of a catchment and the discharge data. Finally, we successfully effectively reduce the parameter dimensions of the LuKARS model for the Kerschbaum case study using the AS method. The model with reduced parameter dimensions is able to reproduce the observed impacts of land use changes in the Kerschbaum springshed, highlighting the robustness of the hydrotope-based modeling approach of LuKARS and its applicability for land use change impact studies in karstic systems.

Resolving the relationship between biodiversity and ecosystem functioning has been one of the central goals of modern ecology. Early debates about the relationship were finally resolved with the advent of a statistical partitioning scheme that decomposed the biodiversity effect into a "selection" effect and a "complementarity" effect. We prove that both the biodiversity effect and its statistical decomposition into selection and complementarity are fundamentally flawed because these methods use a na\"ive null expectation based on neutrality, likely leading to an overestimate of the net biodiversity effect, and they fail to account for the nonlinear abundance-ecosystem functioning relationships observed in nature. Furthermore, under such nonlinearity no statistical scheme can be devised to partition the biodiversity effects. We also present an alternative metric providing a more reasonable estimate of biodiversity effect. Our results suggest that all studies conducted since the early 1990s likely overestimated the positive effects of biodiversity on ecosystem functioning.

Statistical analyses have shown that the sunward component of the interplanetary magnetic field, $\mathit{B}_{x}$ (Geocentric Solar Magnetospheric), moderately but significantly affects the auroral intensity. These observations have been interpreted as signatures of a similar interplanetary magnetic field $\mathit{B}_{x}$ control on Birkeland currents yet to be observed directly. Such a control, attributed to differences in magnetic tension on newly opened magnetic field lines, would lead to stronger region 1 (R1) Birkeland currents for $\mathit{B}_{x}$ negative (positive) conditions in the Northern (Southern) Hemispheres than when $\mathit{B}_{x}$ is positive (negative). In this paper we perform a detailed investigation of three different sets of magnetic field measurements, from the Challenging Minisatellite Payload and Swarm low Earth orbit satellites, from the Active Magnetosphere and Planetary Electrodynamics Response Experiment products derived from the Iridium satellite constellation, and from the SuperMAG ground magnetometer network, each analyzed using different techniques, to test these predictions. The results show that a change in sign of $\mathit{B}_{x}$ changes the Birkeland currents by no more than $\approx$10%. The current patterns show little support for an interhemispheric asymmetry of the kind proposed to explain auroral observations. Instead, we propose an alternative interpretation, which is consistent with most of the auroral observations and with the current observations in the present paper, except for those based on Active Magnetosphere and Planetary Electrodynamics Response Experiment: The solar wind-magnetosphere coupling is more efficient when the dipole tilt angle and $\mathit{B}_{x}$ have the same sign than when they are different... (continued on https://agupubs.onlinelibrary.wiley.com/doi/abs/10.1002/2017JA024864)

Change detection in heterogeneous multitemporal satellite images is an emerging topic in remote sensing. In this paper we propose a framework, based on image regression, to perform change detection in heterogeneous multitemporal satellite images, which has become a main topic in remote sensing. Our method learns a transformation to map the first image to the domain of the other image, and vice versa. Four regression methods are selected to carry out the transformation: Gaussian processes, support vector machines, random forests, and a recently proposed kernel regression method called homogeneous pixel transformation. To evaluate not only potentials and limitations of our framework, but also the pros and cons of each regression method, we perform experiments on two data sets. The results indicates that random forests achieve good performance, are fast and robust to hyperparameters, whereas the homogeneous pixel transformation method can achieve better accuracy at the cost of a higher complexity.

This short paper reports the algorithms we used and the evaluation performances for ISIC Challenge 2018. Our team participates in all the tasks in this challenge. In lesion segmentation task, the pyramid scene parsing network (PSPNet) is modified to segment the lesions. In lesion attribute detection task, the modified PSPNet is also adopted in a multi-label way. In disease classification task, the DenseNet-169 is adopted for multi-class classification.

As melanoma diagnoses increase across the US, automated efforts to identify malignant lesions become increasingly of interest to the research community. Segmentation of dermoscopic images is the first step in this process, thus accuracy is crucial. Although techniques utilizing convolutional neural networks have been used in the past for lesion segmentation, we present a solution employing the recently published DeepLab 3, an atrous convolution method for image segmentation. Although the results produced by this run are not ideal, with a mean Jaccard index of 0.498, we believe that with further adjustments and modifications to the compatibility with the DeepLab code and with training on more powerful processing units, this method may achieve better results in future trials.

This paper addresses the problem of photometric stereo for non-Lambertian surfaces. Existing approaches often adopt simplified reflectance models to make the problem more tractable, but this greatly hinders their applications on real-world objects. In this paper, we propose a deep fully convolutional network, called PS-FCN, that takes an arbitrary number of images of a static object captured under different light directions with a fixed camera as input, and predicts a normal map of the object in a fast feed-forward pass. Unlike the recently proposed learning based method, PS-FCN does not require a pre-defined set of light directions during training and testing, and can handle multiple images and light directions in an order-agnostic manner. Although we train PS-FCN on synthetic data, it can generalize well on real datasets. We further show that PS-FCN can be easily extended to handle the problem of uncalibrated photometric stereo.Extensive experiments on public real datasets show that PS-FCN outperforms existing approaches in calibrated photometric stereo, and promising results are achieved in uncalibrated scenario, clearly demonstrating its effectiveness.

Ecological systems are emergent features of ecological and adaptive dynamics of a community of interacting species. By natural selection through the abiotic environment and by co-adaptation within the community, species evolve, thereby giving rise to the ecological networks we regard as ecosystems. This reductionist perspective can be contrasted with the view that as species have to fit in the surrounding system, the system itself exerts selection pressure on the evolutionary pathways of the species. This interplay of bottom-up and top-down control in the development and growth of ecological systems has long been discussed, however empirical ecosystem data is scarce and a comprehensive mathematical framework is lacking. We present a way of quantifying the relative weight of natural selection and coadaptation grounded in information theory, to assess the relative role of bottom-up and top-down control in the evolution of ecological systems, and analyse the information transfer in an individual based stochastic complex systems model, the Tangled Nature Model of evolutionary ecology. We show that ecological communities evolve from mainly bottom-up controlled early-successional systems to more strongly top-down controlled late-successional systems, as coadaptation progresses. Species which have a high influence on selection are also generally more abundant. Hence our findings imply that ecological communities are shaped by a dialogue of bottom-up and top-down control, where the role of the systemic selection and integrity becomes more pronounced the further the ecosystem is developed.

Nowadays, Earth Observation systems provide a multitude of heterogeneous remote sensing data. How to manage such richness leveraging its complementarity is a crucial chal- lenge in modern remote sensing analysis. Data Fusion techniques deal with this point proposing method to combine and exploit complementarity among the different data sensors. Considering optical Very High Spatial Resolution (VHSR) images, satellites obtain both Multi Spectral (MS) and panchro- matic (PAN) images at different spatial resolution. VHSR images are extensively exploited to produce land cover maps to deal with agricultural, ecological, and socioeconomic issues as well as assessing ecosystem status, monitoring biodiversity and provid- ing inputs to conceive food risk monitoring systems. Common techniques to produce land cover maps from such VHSR images typically opt for a prior pansharpening of the multi-resolution source for a full resolution processing. Here, we propose a new deep learning architecture to jointly use PAN and MS imagery for a direct classification without any prior image fusion or resampling process. By managing the spectral information at its native spatial resolution, our method, named MRFusion, aims at avoiding the possible infor- mation loss induced by pansharpening or any other hand-crafted preprocessing. Moreover, the proposed architecture is suitably designed to learn non-linear transformations of the sources with the explicit aim of taking as much as possible advantage of the complementarity of PAN and MS imagery. Experiments are carried out on two-real world scenarios depicting large areas with different land cover characteristics. The characteristics of the proposed scenarios underline the applicability and the generality of our method in operational settings.

There is an ever growing need to ensure the quality of food and assess specific quality parameters in all the links of the food chain, ranging from processing, distribution and retail to preparing food. Various imaging and sensing technologies, including X-ray imaging, ultrasound, and near infrared reflectance spectroscopy have been applied to the problem. Cost and other constraints restrict the application of some of these technologies. In this study we test a novel Multiplexing Electric Field Sensor (MEFS), an approach that allows for a completely non-invasive and non-destructive testing approach. Our experiments demonstrate the reliable detection of certain foreign objects and provide evidence that this sensor technology has the capability of measuring fat content in minced meat. Given the fact that this technology can already be deployed at very low cost, low maintenance and in various different form factors, we conclude that this type of MEFS is an extremely promising technology for addressing specific food quality issues.

Scene parsing is an indispensable component in understanding the semantics within a scene. Traditional methods rely on handcrafted local features and probabilistic graphical models to incorporate local and global cues. Recently, methods based on fully convolutional neural networks have achieved new records on scene parsing. An important strategy common to these methods is the aggregation of hierarchical features yielded by a deep convolutional neural network. However, typical algorithms usually aggregate hierarchical convolutional features via concatenation or linear combination, which cannot sufficiently exploit the diversities of contextual information in multi-scale features and the spatial inhomogeneity of a scene. In this paper, we propose a mixture-of-experts scene parsing network (MoE-SPNet) that incorporates a convolutional mixture-of-experts layer to assess the importance of features from different levels and at different spatial locations. In addition, we propose a variant of mixture-of-experts called the adaptive hierarchical feature aggregation (AHFA) mechanism which can be incorporated into existing scene parsing networks that use skip-connections to fuse features layer-wisely. In the proposed networks, different levels of features at each spatial location are adaptively re-weighted according to the local structure and surrounding contextual information before aggregation. We demonstrate the effectiveness of the proposed methods on two scene parsing datasets including PASCAL VOC 2012 and SceneParse150 based on two kinds of baseline models FCN-8s and DeepLab-ASPP.

In many real-world problems, there is the possibility to configure, to a limited extent, some environmental parameters to improve the performance of a learning agent. In this paper, we propose a novel framework, Configurable Markov Decision Processes (Conf-MDPs), to model this new type of interaction with the environment. Furthermore, we provide a new learning algorithm, Safe Policy-Model Iteration (SPMI), to jointly and adaptively optimize the policy and the environment configuration. After having introduced our approach and derived some theoretical results, we present the experimental evaluation in two explicative problems to show the benefits of the environment configurability on the performance of the learned policy.

With the emergence of edge computing, there is an increasing need for running convolutional neural network based object detection on small form factor edge computing devices with limited compute and thermal budget for applications such as video surveillance. To address this problem, efficient object detection frameworks such as YOLO and SSD were proposed. However, SSD based object detection that uses VGG16 as backend network is insufficient to achieve real time speed on edge devices. To further improve the detection speed, the backend network is replaced by more efficient networks such as SqueezeNet and MobileNet. Although the speed is greatly improved, it comes with a price of lower accuracy. In this paper, we propose an efficient SSD named Fire SSD. Fire SSD achieves 70.7mAP on Pascal VOC 2007 test set. Fire SSD achieves the speed of 30.6FPS on low power mainstream CPU and is about 6 times faster than SSD300 and has about 4 times smaller model size. Fire SSD also achieves 22.2FPS on integrated GPU.

Automated brain tissue segmentation into white matter (WM), gray matter (GM), and cerebro-spinal fluid (CSF) from magnetic resonance images (MRI) is helpful in the diagnosis of neuro-disorders such as epilepsy, Alzheimer's, multiple sclerosis, etc. However, thin GM structures at the periphery of cortex and smooth transitions on tissue boundaries such as between GM and WM, or WM and CSF pose difficulty in building a reliable segmentation tool. This paper proposes a Fully Convolutional Neural Network (FCN) tool, that is a hybrid of two widely used deep learning segmentation architectures SegNet and U-Net, for improved brain tissue segmentation. We propose a skip connection inspired from U-Net, in the SegNet architetcure, to incorporate fine multiscale information for better tissue boundary identification. We show that the proposed U-SegNet architecture, improves segmentation performance, as measured by average dice ratio, to 89.74% on the widely used IBSR dataset consisting of T-1 weighted MRI volumes of 18 subjects.

Many significant applications need land cover information of remote sensing images that are acquired from different areas and times, such as change detection and disaster monitoring. However, it is difficult to find a generic land cover classification scheme for different remote sensing images due to the spectral shift caused by diverse acquisition condition. In this paper, we develop a novel land cover classification method that can deal with large-scale data captured from widely distributed areas and different times. Additionally, we establish a large-scale land cover classification dataset consisting of 150 Gaofen-2 imageries as data support for model training and performance evaluation. Our experiments achieve outstanding classification accuracy compared with traditional methods.

The forecasting and reconstruction of ocean and atmosphere dynamics from satellite observation time series are key challenges. While model-driven representations remain the classic approaches, data-driven representations become more and more appealing to benefit from available large-scale observation and simulation datasets. In this work we investigate the relevance of recently introduced bilinear residual neural network representations, which mimic numerical integration schemes such as Runge-Kutta, for the forecasting and assimilation of geophysical fields from satellite-derived remote sensing data. As a case-study, we consider satellite-derived Sea Surface Temperature time series off South Africa, which involves intense and complex upper ocean dynamics. Our numerical experiments demonstrate that the proposed patch-level neural-network-based representations outperform other data-driven models, including analog schemes, both in terms of forecasting and missing data interpolation performance with a relative gain up to 50\% for highly dynamic areas.

Establishing up-to-date large scale building maps is essential to understand urban dynamics, such as estimating population, urban planning and many other applications. Although many computer vision tasks has been successfully carried out with deep convolutional neural networks, there is a growing need to understand their large scale impact on building mapping with remote sensing imagery. Taking advantage of the scalability of CNNs and using only few areas with the abundance of building footprints, for the first time we conduct a comparative analysis of four state-of-the-art CNNs for extracting building footprints across the entire continental United States. The four CNN architectures namely: branch-out CNN, fully convolutional neural network (FCN), conditional random field as recurrent neural network (CRFasRNN), and SegNet, support semantic pixel-wise labeling and focus on capturing textural information at multi-scale. We use 1-meter resolution aerial images from National Agriculture Imagery Program (NAIP) as the test-bed, and compare the extraction results across the four methods. In addition, we propose to combine signed-distance labels with SegNet, the preferred CNN architecture identified by our extensive evaluations, to advance building extraction results to instance level. We further demonstrate the usefulness of fusing additional near IR information into the building extraction framework. Large scale experimental evaluations are conducted and reported using metrics that include: precision, recall rate, intersection over union, and the number of buildings extracted. With the improved CNN model and no requirement of further post-processing, we have generated building maps for the United States. The quality of extracted buildings and processing time demonstrated the proposed CNN-based framework fits the need of building extraction at scale.

Horizon or skyline detection plays a vital role towards mountainous visual geo-localization, however most of the recently proposed visual geo-localization approaches rely on \textbf{user-in-the-loop} skyline detection methods. Detecting such a segmenting boundary fully autonomously would definitely be a step forward for these localization approaches. This paper provides a quantitative comparison of four such methods for autonomous horizon/sky line detection on an extensive data set. Specifically, we provide the comparison between four recently proposed segmentation methods; one explicitly targeting the problem of horizon detection\cite{Ahmad15}, second focused on visual geo-localization but relying on accurate detection of skyline \cite{Saurer16} and other two proposed for general semantic segmentation -- Fully Convolutional Networks (FCN) \cite{Long15} and SegNet\cite{Badrinarayanan15}. Each of the first two methods is trained on a common training set \cite{Baatz12} comprised of about 200 images while models for the third and fourth method are fine tuned for sky segmentation problem through transfer learning using the same data set. Each of the method is tested on an extensive test set (about 3K images) covering various challenging geographical, weather, illumination and seasonal conditions. We report average accuracy and average absolute pixel error for each of the presented formulation.

Quantitative interpretation of the tidal response of water levels measured in wells has long been made either with a model for perfectly confined aquifers or with a model for purely unconfined aquifers. However, many aquifers may be neither totally confined nor purely unconfined at the frequencies of tidal loading but behave somewhere between the two end members. Here we present a more general model for the tidal response of groundwater in aquifers with both horizontal and vertical flow. The model has three independent parameters: the transmissivity and storativity of the aquifer and the specific leakage of the leaking aquitard. If transmissivity and storativity are known independently, this model may be used to estimate aquitard leakage from the phase shift and amplitude ratio of water level in wells obtained from tidal analysis. We apply the model to interpret the tidal response of water level in a USGS deep monitoring well installed in the Arbuckle aquifer in Oklahoma, into which massive amount of wastewater co-produced from hydrocarbon exploration has been injected. The analysis shows that the Arbuckle aquifer is leaking significantly at this site. We suggest that the present method may be effectively and economically applied to monitor leakage in groundwater systems, which bears on the safety of water resources, the security of underground waste repositories, and the outflow of wastewater during deep injection and hydrocarbon extraction.

We present the DeepGlobe 2018 Satellite Image Understanding Challenge, which includes three public competitions for segmentation, detection, and classification tasks on satellite images. Similar to other challenges in computer vision domain such as DAVIS and COCO, DeepGlobe proposes three datasets and corresponding evaluation methodologies, coherently bundled in three competitions with a dedicated workshop co-located with CVPR 2018. We observed that satellite imagery is a rich and structured source of information, yet it is less investigated than everyday images by computer vision researchers. However, bridging modern computer vision with remote sensing data analysis could have critical impact to the way we understand our environment and lead to major breakthroughs in global urban planning or climate change research. Keeping such bridging objective in mind, DeepGlobe aims to bring together researchers from different domains to raise awareness of remote sensing in the computer vision community and vice-versa. We aim to improve and evaluate state-of-the-art satellite image understanding approaches, which can hopefully serve as reference benchmarks for future research in the same topic. In this paper, we analyze characteristics of each dataset, define the evaluation criteria of the competitions, and provide baselines for each task.

Interannual variability in the Southern Ocean is investigated via nonlinear Laplacian spectral analysis (NLSA), an objective eigendecomposition technique for nonlinear dynamical systems that can simultaneously recover multiple timescales from data with high skill. Applied to modeled and observed sea surface temperature and sea ice concentration data, NLSA recovers the wavenumber-2 eastward propagating signal corresponding to the Antarctic circumpolar wave (ACW). During certain phases of its lifecycle, the spatial patterns of this mode display a structure that can explain the statistical origin of the Antarctic dipole pattern. Another group of modes have combination frequencies consistent with the modulation of the annual cycle by the ACW. Further examination of these newly identified modes reveals that they can have either eastward or westward propagation, combined with meridional pulsation reminiscent of sea ice reemergence patterns in the Arctic. Moreover, they exhibit smaller-scale spatial structures, and explain more Indian Ocean variance than the primary ACW modes. We attribute these modes to teleconnections between ACW and the tropical Indo-Pacific Ocean; in particular, fundamental ENSO modes and their associated combination modes with the annual cycle recovered by NLSA. Another mode extracted from the Antarctic variables displays an eastward propagating wavenumber-3 structure over the Southern Ocean, but exhibits no strong correlation to interannual Indo-Pacific variability.

Modern geosciences have to deal with large quantities and a wide variety of data, including 2-D, 3-D and 4-D seismic surveys, well logs generated by sensors, detailed lithological records, satellite images and meteorological records. These data serve important industries, such as the exploration of mineral deposits and the production of energy (Oil and Gas, Geothermal, Wind, Hydroelectric), are important in the study of the earth crust to reduce the impact of earthquakes, in land use planning, and have a fundamental role in sustainability. The volume of raw data being stored by different earth science archives today makes it impossible to rely on manual examination by scientists. The data volumes resultant of different sources, from terrestrial or aerial to satellite surveys, will reach a terabyte per day by the time all the planned satellites are flown. In particular, the oil industry has been using large quantities of data for quite a long time. Although there are published works in this area since the 70s, these days, the ubiquity of computing and sensor devices enables the collection of higher resolution data in real time, giving a new life to a mature industrial field. Understanding and finding value in this data has an impact on the efficiency of the operations in the oil and gas production chain. Efficiency gains are particularly important since the steep fall in oil prices in 2014, and represent an important opportunity for data mining and data science.

1. Advances in tracking technology have led to an exponential increase in animal location data, greatly enhancing our ability to address interesting questions in movement ecology, but also presenting new challenges related to data manage- ment and analysis. 2. Step-Selection Functions (SSFs) are commonly used to link environmental covariates to animal location data collected at fine temporal resolution. SSFs are estimated by comparing observed steps connecting successive animal locations to random steps, using a likelihood equivalent of a Cox proportional hazards model. By using common statistical distributions to model step length and turn angle distributions, and including habitat- and movement-related covariates (functions of distances between points, angular deviations), it is possible to make inference regarding habitat selection and movement processes, or to control one process while investigating the other. The fitted model can also be used to estimate utilization distributions and mechanistic home ranges. 3. Here, we present the R-package amt (animal movement tools) that allows users to fit SSFs to data and to simulate space use of animals from fitted models. The amt package also provides tools for managing telemetry data. 4. Using fisher (Pekania pennanti ) data as a case study, we illustrate a four-step approach to the analysis of animal movement data, consisting of data management, exploratory data analysis, fitting of models, and simulating from fitted models.

Proximal soil sensors are taking hold in the understanding of soil hydrogeological processes involved in precision agriculture. In this context, permanently installed gamma ray spectroscopy stations represent one of the best space-time trade off methods at field scale. This study proved the feasibility and reliability of soil water content monitoring through a seven-month continuous acquisition of terrestrial gamma radiation in a tomato test field. By employing a 1 L sodium iodide detector placed at a height of 2.25 m, we investigated the gamma signal coming from an area having a ~25 m radius and from a depth of approximately 30 cm. Experimental values, inferred after a calibration measurement and corrected for the presence of biomass, were corroborated with gravimetric data acquired under different soil moisture conditions, giving an average absolute discrepancy of about 2%. A quantitative comparison was carried out with data simulated by AquaCrop, CRITeRIA, and IRRINET soil-crop system models. The different goodness of fit obtained in bare soil condition and during the vegetated period highlighted that CRITeRIA showed the best agreement with the experimental data over the entire data-taking period while, in presence of the tomato crop, IRRINET provided the best results.

Semantic segmentation in high resolution remote sensing images is a fundamental and challenging task. Convolutional neural networks (CNNs), such as fully convolutional network (FCN) and SegNet, have shown outstanding performance in many segmentation tasks. One key pillar of these successes is mining useful information from features in convolutional layers for producing high resolution segmentation maps. For example, FCN nonlinearly combines high-level features extracted from last convolutional layers; whereas SegNet utilizes a deconvolutional network which takes as input only coarse, high-level feature maps of the last convolutional layer. However, how to better fuse multi-level convolutional feature maps for semantic segmentation of remote sensing images is underexplored. In this work, we propose a novel bidirectional network called recurrent network in fully convolutional network (RiFCN), which is end-to-end trainable. It has a forward stream and a backward stream. The former is a classification CNN architecture for feature extraction, which takes an input image and produces multi-level convolutional feature maps from shallow to deep; while in the later, to achieve accurate boundary inference and semantic segmentation, boundary-aware high resolution feature maps in shallower layers and high-level but low-resolution features are recursively embedded into the learning framework (from deep to shallow) to generate a fused feature representation that draws a holistic picture of not only high-level semantic information but also low-level fine-grained details. Experimental results on two widely-used high resolution remote sensing data sets for semantic segmentation tasks, ISPRS Potsdam and Inria Aerial Image Labeling Data Set, demonstrate competitive performance obtained by the proposed methodology compared to other studied approaches.

New statistical methods were employed to improve the ability to distinguish benign from malignant breast tissue ex vivo in a recent study. The ultimately aim was to improve the intraoperative assessment of positive tumour margins in breast-conserving surgery (BCS), potentially reducing patient re-operation rates. A multivariate Bayesian classifier was applied to the waveform samples produced by a Terahertz Pulsed Imaging (TPI) handheld probe system in order to discriminate tumour from benign breast tissue, obtaining a sensitivity of 96% and specificity of 95%. We compare these results to traditional and to state-of-the-art methods for determining resection margins. Given the general nature of the classifier, it is expected that this method can be applied to other tumour types where resection margins are also critical.

The development of precision livestock farming which adjusts the food needs of each animal requires detailed knowledge of its behavior and particularly physical activity. Individual differences between animals can be observed for group-housed sows. Accelerometer technology offers opportunities for automatic monitoring of animal behavior. The aim of the first step was to develop a methodology to attach the accelerometer on the sow's leg, and an algorithm to automatically detect standing and lying posture. Accelerometers (Hobo Pendant G) were put in a metal case and fastened with two cable ties on the leg of 6 group-housed sows. The data loggers recorded the acceleration on one axis every 20 s. Data were then validated by 9 hours of direct observations. The automatic recording device showed data of high sensitivity (98.8%) and specificity (99.8%). Then, accelerometers were placed on 12 to 13 group-housed sows for 2 to 4 consecutive days in 6 commercial farms equipped with electronic sow feeding. On average each day, sows spent 259 minutes ($\pm$ 114) standing and changed posture 29 ($\pm$ 12) times. The sow's standing time was repeatable day to day. Differences between sows and herds were significant. Based on behavioral data, 5 categories of sows were identified. This study suggests that the consideration of individual behavior of each animal would improve herd management.

The anchor mechanism of Faster R-CNN and SSD framework is considered not effective enough to scene text detection, which can be attributed to its IoU based matching criterion between anchors and ground-truth boxes. In order to better enclose scene text instances of various shapes, it requires to design anchors of various scales, aspect ratios and even orientations manually, which makes anchor-based methods sophisticated and inefficient. In this paper, we propose a novel anchor-free region proposal network (AF-RPN) to replace the original anchor-based RPN in the Faster R-CNN framework to address the above problem. Compared with a vanilla RPN and FPN-RPN, AF-RPN can get rid of complicated anchor design and achieve higher recall rate on large-scale COCO-Text dataset. Owing to the high-quality text proposals, our Faster R-CNN based two-stage text detection approach achieves state-of-the-art results on ICDAR-2017 MLT, ICDAR-2015 and ICDAR-2013 text detection benchmarks when using single-scale and single-model (ResNet50) testing only.

Floating marine debris represent a threat to marine and coastal ecology. Since the Mediterranean basin is one of the highly impacted regions, both by the coastal pollution as well as from sea traffic, the potential harm of a floating pollution on the marine ecology could be overwhelming in this area. Our study area covers the central Mediterranean crossing that connects the western and eastern Mediterranean and is one of the areas impacted by a high intensity of sea traffic. To identify regions in the central Mediterranean that could be more exposed by high concentration of floating marine pollutants we use Leeway model for lower windage small-size particles. We perform numerical simulation of a large ensemble of Lagrangian particles that approximate at-sea debris. The particles are forced by high-resolution sea kinematics from the Copernicus Marine Environment Monitoring Service (CMEMS) and 10 m atmospheric wind from the European Centre for Medium-Range Weather Forecasts (ECMWF) for two reference periods in summer and winter of 2013--2016. We identify the regions with a high accumulation of particles in terms of particle surface densities per unit area. Although seasonal and annual variability of ocean current and atmospheric wind is an important factor that influences accumulation regimes across the central Mediterranean, we found that the border of the Libyan shelf harbors larger percentage of particles after 30 days of simulation.

Semantic segmentation is a fundamental task in computer vision, which can be considered as a per-pixel classification problem. Recently, although fully convolutional neural network (FCN) based approaches have made remarkable progress in such task, aggregating local and contextual information in convolutional feature maps is still a challenging problem. In this paper, we argue that, when predicting the category of a given pixel, the regions close to the target are more important than those far from it. To tackle this problem, we then propose an effective yet efficient approach named Vortex Pooling to effectively utilize contextual information. Empirical studies are also provided to validate the effectiveness of the proposed method. To be specific, our approach outperforms the previous state-of-the-art model named DeepLab v3 by 1.5% on the PASCAL VOC 2012 val set and 0.6% on the test set by replacing the Atrous Spatial Pyramid Pooling (ASPP) module in DeepLab v3 with the proposed Vortex Pooling. Moreover, our model (10.13FPS) shares similar computation cost with DeepLab v3 (10.37 FPS).

Recent CNN based object detectors, no matter one-stage methods like YOLO, SSD, and RetinaNe or two-stage detectors like Faster R-CNN, R-FCN and FPN are usually trying to directly finetune from ImageNet pre-trained models designed for image classification. There has been little work discussing on the backbone feature extractor specifically designed for the object detection. More importantly, there are several differences between the tasks of image classification and object detection. 1. Recent object detectors like FPN and RetinaNet usually involve extra stages against the task of image classification to handle the objects with various scales. 2. Object detection not only needs to recognize the category of the object instances but also spatially locate the position. Large downsampling factor brings large valid receptive field, which is good for image classification but compromises the object location ability. Due to the gap between the image classification and object detection, we propose DetNet in this paper, which is a novel backbone network specifically designed for object detection. Moreover, DetNet includes the extra stages against traditional backbone network for image classification, while maintains high spatial resolution in deeper layers. Without any bells and whistles, state-of-the-art results have been obtained for both object detection and instance segmentation on the MSCOCO benchmark based on our DetNet~(4.8G FLOPs) backbone. The code will be released for the reproduction.