Earth Science Applications Research Articles

Deep learning in environmental remote sensing: Achievements and challenges

Various forms of machine learning (ML) methods have historically played a valuable role in environmental remote sensing research. With an increasing amount of “big data” from earth observation and rapid advances in ML, increasing opportunities for novel methods have emerged to aid in earth environmental monitoring. Over the last decade, a typical and state-of-the-art ML framework named deep learning (DL), which is developed from the traditional neural network (NN), has outperformed traditional models with considerable improvement in performance. Substantial progress in developing a DL methodology for a variety of earth science applications has been observed. Therefore, this review will concentrate on the use of the traditional NN and DL methods to advance the environmental remote sensing process. First, the potential of DL in environmental remote sensing, including land cover mapping, environmental parameter retrieval, data fusion and downscaling, and information reconstruction and prediction, will be analyzed. A typical network structure will then be introduced. Afterward, the applications of DL environmental monitoring in the atmosphere, vegetation, hydrology, air and land surface temperature, evapotranspiration, solar radiation, and ocean color are specifically reviewed. Finally, challenges and future perspectives will be comprehensively analyzed and discussed.

Forest height estimation and change monitoring based on artificial neural network using Geoscience Laser Altimeter System and Landsat data

The continuous forest inventory is critical to forestry monitoring, Earth science applications, and global carbon cycle researching. Traditional field measurements of forest inventory are time-consuming. Targeting the problem, our study proposes an approach for monitoring the forest tree heights by combining geoscience laser altimeter system data with land surface reflectance data from Landsat TM based on artificial neural network (ANN) model in the Nanning research region. The land surface reflectance data are used as inputs of ANN model. The GLAS data are used to derive tree heights for training ANN model. Then the forest height spatial distribution and corresponding height change during 2003 to 2006 are investigated. The validation shows the high precision of modeling tree heights by, respectively, calculating determination coefficient ( R 2) and root-mean-square error (RMSE) of evaluations from 2003 to 2006 ( R 2 = 0.667, mean RMSE = 3.219 m). Finally, the change trends of forest tree heights in Nanning region are obtained by calculating the difference of tree heights from 2003 to 2006.

Next Generation Mapping: Combining Deep Learning, Cloud Computing, and Big Remote Sensing Data

The rapid growth of satellites orbiting the planet is generating massive amounts of data for Earth science applications. Concurrently, state-of-the-art deep-learning-based algorithms and cloud computing infrastructure have become available with a great potential to revolutionize the image processing of satellite remote sensing. Within this context, this study evaluated, based on thousands of PlanetScope images obtained over a 12-month period, the performance of three machine learning approaches (random forest, long short-term memory-LSTM, and U-Net). We applied these approaches to mapped pasturelands in a Central Brazil region. The deep learning algorithms were implemented using TensorFlow, while the random forest utilized the Google Earth Engine platform. The accuracy assessment presented F1 scores for U-Net, LSTM, and random forest of, respectively, 96.94%, 98.83%, and 95.53% in the validation data, and 94.06%, 87.97%, and 82.57% in the test data, indicating a better classification efficiency using the deep learning approaches. Although the use of deep learning algorithms depends on a high investment in calibration samples and the generalization of these methods requires further investigations, our results suggest that the neural network architectures developed in this study can be used to map large geographic regions that consider a wide variety of satellite data (e.g., PlanetScope, Sentinel-2, Landsat-8).

Public response to news reports on the Mount Kusatsu-Shirane volcanic eruption: A content analysis of online reader comments

This paper reports on the real-time public responses to the eruption disaster on Mt. Kusatsu-Shirane in January 2018. This volcanic disaster attracted significant media coverage, with many people again acknowledging the risk of volcanic hazards from the many news reports. These days, as it has become increasingly common for people to express their opinions using online sites, online reader comments can be important sources of public opinion. This paper used a content analysis approach to analyze the comments posted in online news reports on the volcanic disaster to determine the public's perspective of the risks associated with volcanic disasters and volcanology. Applying an inductive category-extraction strategy, the comments were sorted into six categories of comments (1) that made reference to volcanic risk perception, (2) on volcanology or volcanologists, (3) that had fatalistic perceptions of natural disasters as blameless misfortunes, (4) on eruption predictability, (5) on disaster risk reduction measures implemented by central/local governments and disaster management agencies, and (6) explaining the phreatic explosion mechanism. It was found that many comments tended to have a negative tone, and a prominent criticism was that modern science could not detect volcanic activities reliably. However, some commenters suggested that experts with a sophisticated knowledge of volcanology should be available to support local communities in a similar way to the “home doctor”. Volcanologists need to disseminate information, such as the current state of geological and earth science education systems, volcanological knowledge, and application of the lessons learned from volcanic disasters.

Peakaboo: Advanced software for the interpretation of X-ray fluorescence spectra from synchrotrons and other intense X-ray sources

Abstract Peakaboo is a platform for the analysis of full spectrum synchrotron X-ray fluorescence (XRF) map data. It offers robust, automated fitting of XRF spectral peaks, increasing the sensitivity to trace chemical elements. Large arrays of spectral data can also be processed into element maps to define spatial relationships. The modifications to Peakaboo introduced since 2017 are particularly designed to help new users to become rapidly competent in the handling of complex XRF spectra. Some of the most important innovations of the software are described below. Finally, we share examples of its application in Earth and Environmental Sciences.

A Survey on Spatial Prediction Methods

With the advancement of GPS and remote sensing technologies, large amounts of geospatial data are being collected from various domains, driving the need for effective and efficient prediction methods. Given spatial data samples with explanatory features and targeted responses (categorical or continuous) at a set of locations, the spatial prediction problem aims to learn a model that can predict the response variable based on explanatory features. The problem is important with broad applications in earth science, urban informatics, geosocial media analytics, and public health, but is challenging due to the unique characteristics of spatial data, including spatial autocorrelation, heterogeneity, limited ground truth, and multiple scales and resolutions. This paper provides a systematic review on principles and methods in spatial prediction. We provide a taxonomy of methods categorized by the key challenge they address. For each method, we introduce its underlying assumption, theoretical foundation, and discuss its advantages and disadvantages. We also discuss spatiotemporal extensions of methods. Our goal is to help interdisciplinary domain scientists choose techniques to solve their problems, and more importantly, to help data mining researchers to understand the main principles and methods in spatial prediction and identify future research opportunities.

The actinide beamline at VERA

Interest in the long-lived radioisotope 236U (T1/2 = 23.4 million years) has significantly increased recently due to the emergence of environmental and earth science applications. Compared to the previous setup at VERA, which was based on oxygen stripping and a time-of-flight detector, we have improved the sensitivity of VERA by more than an order of magnitude by switching to helium stripping and by installing a second 90° magnet in our analyzer beamline. The new setup has been successfully employed for several research projects.Here, we present the characterization of the upgraded spectrometer. We discuss the design of the new beamline and present benchmark measurements, suggesting an instrumental sensitivity limit well below 236U/238U = 10−14. The yield for the 3+ charge state is 19%, of which 90% are recorded in the detector. In environmental samples, one in 4500 sputtered actinide atoms is detected. While this allows tackling natural 236U, also measurements of anthropogenic 236U or other actinides profit from the higher sensitivity. This allows analysis of smaller samples, but also has made the rare anthropogenic isotope 233U accessible.

A model of local thermal non-equilibrium during infiltration

A realistic temperature estimation is crucial for many earth-science applications, ranging from hydro-thermal systems to plant physiology. The most common approach to calculate the temperature in multi-phase systems assumes immediate local thermal equilibrium (LTE) between the phases. However, local thermal equilibrium between the phases is not applicable in various scenarios like during the infiltration of rain or melt water in frozen soil, limiting the applicability of the approach and inhibiting the implementation of separate initial and boundary conditions for non-equilibrium situations. In local thermal non-equilibrium (LTNE) models, phase temperatures are described separately to the cost of additional differential equations and an explicitly formulated heat transfer between the phases. Especially a cumbersome parameterization of the explicit heat transfer restricts the use of the LTNE models in multi-phase conditions so far. In this work, we derive a general local thermal non-equilibrium model for dynamic, partly saturated porous media. Heat transfer between the phases is described explicitly using well-known semi-empirical parameterization accounting for velocity changes of the mobile phases. The change in volume fraction introduces an additional term in the heat equation, causing a coupling with the hydraulic model. We validate our model with a numerical simulation of historic experimental data from soil infiltration experiments of warm and cold water into drained soil, posing a perfect example of local thermal non-equilibrium conditions between the phases. Experimentally obtained mixture temperatures are reproduced within experimental accuracy. We further show the benefits of our model by applying it to rainwater infiltration into cold soil. Besides a consistent formulation of initial and boundary conditions, the derived model allows physically based conclusions about the thermal state of the separate phases.

Comparison of methods for the detection of 10Be with AMS and a new approach based on a silicon nitride foil stack

Natural 10Be (t1/2 = 1.387 ± 0.012 Ma) is produced by cosmic rays and is present on Earth's surface only at ultratrace concentrations (typically 104 to 1010 atoms/g). Its cosmogenic origin makes it an important tracer for many applications in Earth and environmental sciences. An improved accelerator mass spectrometry (AMS) method has been developed at the Vienna Environmental Research Accelerator (VERA) at the University of Vienna to detect the long-lived radionuclide 10Be and separate it from its isobar 10B. Recently installed and projected AMS facilities mainly apply a degrader foil followed by an electrostatic or magnetic separator to remove 10B from the ion beam. This provides the highest suppression of 10B, but suffers from significant transmission losses of 10Be ions. The new technique described here achieves comparable 10B suppression with a passive absorber, consisting of a stack of silicon nitride foils. Compared to a gas absorber, the smaller energy straggling in foils allows separation at lower energies. For a tandem accelerator operated at 3 MV, the charge state 2 + instead of 3 + can be used, with a stripping yield as high as 55%. This way, a high overall efficiency is gained. The setup is simple to operate, and provides good precision and accuracy. We compare this new approach with other methods used at VERA and at other AMS facilities. The foil stack setup was fully characterized with artificial samples from chemically and isotopically well-defined reagents, and is now routinely applied to real samples in various research projects at VERA. The new method is straightforward to be implemented, and was already adopted at another AMS facility with higher terminal voltage, the potential use at tandem accelerators with lower terminal voltage is under exploration.

A joint velocity-intermittency analysis reveals similarity in the vertical structure of atmospheric and hydrospheric canopy turbulence

Turbulent flow through and over vegetation continues to draw significant research attention given its relevance to a plethora of applications in earth and environmental science. Canopy flows are characterized by three-dimensional coherent vortical motions not directly accessible from single-point measurements, which pose a challenge to formalizing links between vegetation structure and turbulent motion. A joint velocity-intermittency technique is applied to velocity data collected within and above aquatic vegetation in a hydraulic flume and above a forested canopy. The approach reveals behavior that provides greater insight into canopy flow dynamics than may be inferred from the vertical profiles of mean velocity, turbulence intensity and Reynolds stresses, which are the quantities usually studied. There is a remarkable similarity in the structure of such flows between the forest canopy and the flume study despite large differences in morphology and stem rigidity. In particular, these results determine an outer flow type arising above 1.5 canopy heights, while turbulent in-rushing events are most significant at the zero-plane displacement. The approach also implies ways in which improved models for canopy turbulence may be developed.

Ground, Proximal, and Satellite Remote Sensing of Soil Moisture

AbstractSoil moisture (SM) is a key hydrologic state variable that is of significant importance for numerous Earth and environmental science applications that directly impact the global environment and human society. Potential applications include, but are not limited to, forecasting of weather and climate variability; prediction and monitoring of drought conditions; management and allocation of water resources; agricultural plant production and alleviation of famine; prevention of natural disasters such as wild fires, landslides, floods, and dust storms; or monitoring of ecosystem response to climate change. Because of the importance and wide‐ranging applicability of highly variable spatial and temporal SM information that links the water, energy, and carbon cycles, significant efforts and resources have been devoted in recent years to advance SM measurement and monitoring capabilities from the point to the global scales. This review encompasses recent advances and the state‐of‐the‐art of ground, proximal, and novel SM remote sensing techniques at various spatial and temporal scales and identifies critical future research needs and directions to further advance and optimize technology, analysis and retrieval methods, and the application of SM information to improve the understanding of critical zone moisture dynamics. Despite the impressive progress over the last decade, there are still many opportunities and needs to, for example, improve SM retrieval from remotely sensed optical, thermal, and microwave data and opportunities for novel applications of SM information for water resources management, sustainable environmental development, and food security.

Accurate Determination of Boron Content in Halite by ICP-OES and ICP-MS.

Boron element is widely distributed in different geologic bodies, and there are important geo-chemical applications in earth science. Halite is a common mineral found in sediment basin. However there is no good method to accurately measure the boron content in halite, which is mainly because Inductively Coupled Plasma Optical Emission Spectrometer (ICP-OES) and Inductively Coupled Plasma Mass Spectrometer (ICP-MS) are limited by the high salt matrix interference and the instrument detection limit. Thus enriching the boron element and removing the matrix interference are necessary before the measuring. In this paper, Amberlite IRA 743 boron-specific resin was applied to enrich the boron element and remove most of the high-salt matrix. The strong acid cation resin (Dowex 50 W×8, 200-400 mesh, USA) and weak-base anion resin (Ion Exchanger II, Germany) were mixed with equal volume, which could remove the foreign ions completely: meanwhile, the relative content of boron in the solution reached above 98%, and the recoveries ranged from 97.8% to 104%. 208.900 nm was chosen as the detection wavelength for ICP-OES, and the detection identification and quantification limits were 0.006 mg·L−1 and 0.02 mg·L−1, respectively. 11B was chosen as the measuring element for ICP-MS, and the detection identification and quantification limits were severally 0.036 mg·L−1 and 0.12 mg·L−1. The relative standard deviations ranged from 1.4% to 3.4% through six replicates under different salinities. Therefore, the process could be regarded as a feasible method to measure boron content in halite by ICP-OES and ICP-MS.

Machine Learning Algorithms for Automatic Lithological Mapping Using Remote Sensing Data: A Case Study from Souk Arbaa Sahel, Sidi Ifni Inlier, Western Anti-Atlas, Morocco

Remote sensing data proved to be a valuable resource in a variety of earth science applications. Using high-dimensional data with advanced methods such as machine learning algorithms (MLAs), a sub-domain of artificial intelligence, enhances lithological mapping by spectral classification. Support vector machines (SVM) are one of the most popular MLAs with the ability to define non-linear decision boundaries in high-dimensional feature space by solving a quadratic optimization problem. This paper describes a supervised classification method considering SVM for lithological mapping in the region of Souk Arbaa Sahel belonging to the Sidi Ifni inlier, located in southern Morocco (Western Anti-Atlas). The aims of this study were (1) to refine the existing lithological map of this region, and (2) to evaluate and study the performance of the SVM approach by using combined spectral features of Landsat 8 OLI with digital elevation model (DEM) geomorphometric attributes of ALOS/PALSAR data. We performed an SVM classification method to allow the joint use of geomorphometric features and multispectral data of Landsat 8 OLI. The results indicated an overall classification accuracy of 85%. From the results obtained, we can conclude that the classification approach produced an image containing lithological units which easily identified formations such as silt, alluvium, limestone, dolomite, conglomerate, sandstone, rhyolite, andesite, granodiorite, quartzite, lutite, and ignimbrite, coinciding with those already existing on the published geological map. This result confirms the ability of SVM as a supervised learning algorithm for lithological mapping purposes.

An Overlap Store Optimization for Large-Scale Parallel Earth Science Application

An Overlap Store Optimization for Large-Scale Parallel Earth Science Application

Determine In-Situ Stress and Characterize Complex Fractures in Naturally Fractured Reservoirs from Diagnostic Fracture Injection Tests

Estimation of in-situ stresses has significant applications in earth sciences and subsurface engineering, such as fault zone studies, underground CO2 sequestration, nuclear waste repositories, oil and gas reservoir development, and geothermal energy exploitation. Over the past few decades, Diagnostic Fracture Injection Tests (DFIT), which have also been referred to as Injection-Falloff Tests, Fracture Calibration Tests, and Mini-Frac Tests, have evolved into a commonly used and reliable technique to obtain in-situ stress. Simplifying assumptions used in traditional methods often lead to inaccurate estimation of the in-situ stress, even for a planar fracture geometry. When a DFIT is conducted in naturally fractured reservoirs, the stimulated natural fractures can either alter the effective reservoir permeability within the distance of investigation or interact with the hydraulic fracture to form a complex fracture geometry, this further complicates stress estimation. In this study, we present a new pressure transient model for DFIT analysis in naturally fractured reservoirs. By analyzing synthetic, laboratory and field cases, we found that fracture complexity and permeability evolution can be detected from DFIT data. Most importantly, it is shown that using established methods to pick minimum in-situ stress often lead to over or underestimates, regardless of whether the reservoir is heavily fractured or sparsely fractured. Our proposed “variable compliance method” gives a much more accurate and reliable estimation of in-situ stress in both homogenous and naturally fractured reservoirs. By combining the unique pressure signatures associated with the closure of natural fractures, a lower bound on the horizontal stress anisotropy can be estimated.

High-Precision Measurements of the Isotopic Composition of Common Lead Using MC-ICPMS: Comparison of Calibration Strategies Based on Full Gravimetric Isotope Mixture and Regression Models.

The determination of isotope ratios of lead finds many important applications in earth and planetary sciences. In this study, we report the first independent and fully calibrated absolute isotope ratio measurements of a common lead since the seminal work by the NIST in the 1960s, which has provided the widely used standards SRM 981 and SRM 982. This work employs MC-ICPMS for the absolute isotope ratio measurements, which were calibrated using two independent methods: the full gravimetric isotope mixture method based on the use of all four near-pure lead isotopes (the first application of this primary method to lead) and the regression method with NIST SRM 997 thallium isotopic standard as the primary calibrator. Both calibration methods provide results consistent to a few parts in 104, which were used to characterize isotope ratios of lead in an NRC candidate reference material, high-purity common lead HIPB-1.

MACHINE LEARNING APPLICATIONS IN OCEANOGRAPHY

Machine learning (ML) is a subset of artificial intelligence that enables to take decision based on data. Artificial intelligence makes possible to integrate ML capabilities into data driven modelling systems in order to bridge the gaps and lessen demands on human experts in oceanographic research .ML algorithms have proven to be a powerful tool for analysing oceanographic and climate data with high accuracy in efficient way. ML has a wide spectrum of real time applications in oceanography and Earth sciences. This study has explained in simple way the realistic uses and applications of major ML algorithms. The main application of machine learning in oceanography is prediction of ocean weather and climate, habitat modelling and distribution, species identification, coastal water monitoring, marine resources management, detection of oil spill and pollution and wave modelling.

Optimization of analytical conditions for precise and accurate isotope analyses of Li, Mg, Fe, Cu, and Zn by MC-ICPMS

This work presents optimized analytical protocols for the accurate and precise measurements of a selection of stable metal isotopes on Nu Plasma II MC-ICPMS to facilitate their applications in earth and planetary sciences.

Design, Construction and In Situ Testing of a Muon Camera for Earth Science and Civil Engineering Applications

The MUST2 (MUon Survey Tomography based on Micromegas detectors for Unreachable Sites Technology) camera is based on a thin Time Projection Chamber read by a resistive Micromegas. This innovative combination presents interesting distinctive features compared to existing muon detection technologies. It allows a wide angular acceptance of the detector with a low weight and compact volume, well adapted for confined spaces or underground operation. The current work presents the results obtained during the calibration measurements at the reference site, the Low Background Noise Laboratory (LBNL). Preliminary results from field measurement campaign carried out at the dam overlooking the village of Saint-Saturnin-les-Apt (South-East of France) are presented and discussed.

Hierarchical Optimization: Fast and Robust Multiscale Stochastic Reconstructions with Rescaled Correlation Functions.

Stochastic reconstructions based on universal correlation functions allow obtaining spatial structures based on limited input data or to fuse multiscale images from different sources. Current application of such techniques is severely hampered by the computational cost of the annealing optimization procedure. In this study we propose a novel hierarchical annealing method based on rescaled correlation functions, which improves both accuracy and computational efficiency of reconstructions while not suffering from disadvantages of existing speeding-up techniques. A significant order of magnitude gains in computational efficiency now allows us to add more correlation functions into consideration and, thus, to further improve the accuracy of the method. In addition, the method provides a robust multiscale framework to solve the universal upscaling or downscaling problem. The novel algorithm is extensively tested on binary (two-phase) microstructures of different genesis. In spite of significant improvements already in place, the current implementation of the hierarchical annealing method leaves significant room for additional accuracy and computational performance tweaks. As described here, (multiscale) stochastic reconstructions will find numerous applications in material and Earth sciences. Moreover, the proposed hierarchical approach can be readily applied to a wide spectrum of constrained optimization problems.