Comparative performance of convolutional neural network,weighted and conventional support vector machine and random forest for classifying tree species using hyperspectral and photogrammetric data

ABSTRACT

The classification of tree species can significantly benefit from high spatial and spectral information acquired by unmanned aerial vehicles (UAVs) associated with advanced classification methods. This study investigated the following topics concerning the classification of 16 tree species in two subtropical forest fragments of Southern Brazil: i) the potential integration of UAV-borne hyperspectral images with 3D information derived from their photogrammetric point cloud (PPC); ii) the performance of two machine learning methods (support vector machine – SVM and random forest – RF) when employing different datasets at a pixel and individual tree crown (ITC) levels; iii) the potential of two methods for dealing with the imbalanced sample set problem: a new weighted SVM (wSVM) approach, which attributes different weights to each sample and class, and a deep learning classifier (convolutional neural network – CNN), associated with a previous step to balance the sample set; and finally, iv) the potential of this last classifier for tree species classification as compared to the above mentioned machine learning methods. Results showed that the inclusion of the PPC features to the hyperspectral data provided a great accuracy increase in tree species classification results when conventional machine learning methods were applied, between 13 and 17% depending on the classifier and the study area characteristics. When using the PPC features and the canopy height model (CHM), associated with the majority vote (MV) rule, the SVM, wSVM and RF classifiers reached accuracies similar to the CNN, which outperformed these classifiers for both areas when considering the pixel-based classifications (overall accuracy of 84.4% in Area 1, and 74.95% in Area 2). The CNN was between 22% and 26% more accurate than the SVM and RF when only the hyperspectral bands were employed. The wSVM provided a slight increase in accuracy not only for some lesser represented classes, but also some major classes in Area 2. While conventional machine learning methods are faster, they demonstrated to be less stable to changes in datasets, depending on prior segmentation and hand-engineered features to reach similar accuracies to those attained by the CNN. To date, CNNs have been barely explored for the classification of tree species, and CNN-based classifications in the literature have not dealt with hyperspectral data specifically focusing on tropical environments. This paper thus presents innovative strategies for classifying tree species in subtropical forest areas at a refined legend level, integrating UAV-borne 2D hyperspectral and 3D photogrammetric data and relying on both deep and conventional machine learning approaches.     

Satellite measurements of solar X-ray flux and their use for interpretation of sudden ionospheric disturbances

Summary The apparatus for the reception and evaluation of the real-time telemetry (RTT) data on the X-ray flux, measured by the satellite Solrad-9, is briefly described. A comparison of the RTT data obtained with the data published by the authors of the experiment shows a good correlation of the data compared. The results of the comparison of our RTT data with SID data of the Panská Ves Observatory confirm many known facts concerning the response of the lower ionosphere to solar flares and, moreover, seem to support the idea of a less important role of X-rays in the lower ionosphere under slightly disturbed conditions.     

When less is more: limits to international transfers under Article 6 of the Paris Agreement

International carbon markets can be an important tool in achieving countries’ mitigation targets under the Paris Agreement, but they are subject to a number of environmental integrity risks. An important risk is that some countries have mitigation targets that correspond to higher levels of emissions than independent projections of their likely emissions. If such ‘hot air’ can be transferred to other countries, it could increase aggregated emissions and create a perverse incentive for countries not to enhance the ambition of future mitigation targets. Limits to international transfers of mitigation outcomes have been proposed to address this risk. This article proposes a typology for such limits, explores key design options, and tests different types of limits in the context of 15 countries. Our analysis indicates that limits to international transfers could, if designed appropriately, prevent most of the hot air contained in current mitigation targets from being transferred, but also involve trade-offs between different policy objectives. Given the risks from international transfer of hot air and the uncertainty over whether other approaches will be effective in ensuring environmental integrity, we recommend that countries take a cautious approach and pursue a portfolio of approaches to ensure environmental integrity, in which case limits could provide for additional safeguards.

Key policy insights

• Limits to international transfers involve trade-offs between different policy objectives, in particular reducing the risk that countries transfer hot air and enabling participation in carbon markets.

• Under ‘relative’ limits a country may transfer mitigation outcomes to the extent that its actual emissions are below the limit. Relative limits derived from historical emissions data have significant limitations, and none of the tested approaches was found to be effective for all countries. Relative limits based on emission projections could be a more valid approach, although they are also technically and politically challenging.

• Under ‘absolute’ limits a country could only issue, transfer or acquire a certain amount of mitigation outcomes. Absolute limits set at sufficiently low levels could prevent countries from transferring large amounts of hot air, but are bluntly applicable to all countries, whether or not they have hot air.

     

The selection of filters for reduction of optical contamination in astronomical CCD X-ray images

The European Photon Imaging Camera(EPIC), is the X-ray imaging and medium spectroscopy instrument for theESA X-ray Multi Mirror telescope(XMM) mission. TheCCD detectors to be used in the three focal plane cameras will provide images in the energy band from 0.1 to 10 keV. However, spectral studies may be compromised by low energy, optical photon contamination. In order to reduce this effect, a number of filters will be incorporated onto a rotating mechanism in the camera head. The filters will be chosen to provide a significant reduction in the optical contamination from a source whilst minimising the attenuation of the X-ray flux. Four commercial filters are described here and their effects on calculated typical source fluxes evaluated. In addition, two alternative filter designs are described and their effects on a simulated source spectra are debated. In both cases, particular attention is given to the problem of maintaining high sensitivity at soft X-ray energies (less than 2 keV).     

Time variant analysis of geomagnetic signals describes the volcanic activity of Mt. Etna between early 2000 and late 2002

Volcanomagnetic anomalies have been mostly observed during strong eruptions. Our aim is to improve the geomagnetic data analysis to evidence the anomalies occurring in a larger time span, especially in the phases preceding the eruptive events. We developed a time variant statistical approach and applied it to the 2000–2002 Etna geomagnetic temporal series. It is based on an algorithm that statistically predicts the geomagnetic field at the station on the volcanic edifice by that recorded at the remote one. In such a way a number of significant changes in the time series (called statistical innovations), marking the local magnetic field change, were detected. The distribution of such statistical innovations accurately describes the Etna volcanic evolution: we note a progressive increase of the innovation occurrence as the eruptive cycles were approaching and only few and weak innovations at times between the various eruptive cycles. The significance of this analysis is further confirmed by the close agreement among the mean square prediction error, strain release and the volcanic activity behavior. On the contrary, the geomagnetic field at a single station or its difference at two stations do not have any clear correlation with other measured physical quantities. The complex pattern of the prediction error was also investigated by a multifractal analysis. We found that the Holder regularity increases with the intensification of the volcanic activity, implying that innovations tend to be less sporadic and correlated during the major volcanic phases.     

Simultaneous infrasonic,seismic, magnetic and ionospheric observations in an earthquake epicentre

Various pre-seismic and co-seismic effects have been reported in the literature in the solid Earth, hydrosphere, atmosphere, electric/magnetic field and in the ionosphere. Some of the effects observed above the surface, particularly some of the pre-seismic effects, are still a matter of debate. Here we analyze the co-seismic effects of a relatively weak earthquake of 28 October 2008, which was a part of an earthquake swarm in the westernmost region of the Czech Republic. Special attention is paid to unique measurements of infrasonic phenomena. As far as we know, these have been the first infrasonic measurements during earthquake in the epicentre zone. Infrasonic oscillations (～1–12 Hz) in the epicentre region appear to be excited essentially by the vertical seismic oscillations. The observed oscillations are real epicentral infrasound not caused by seismic shaking of the instruments or by meteorological phenomena. Seismo-infrasonic oscillations observed 155 km apart from the epicentre were excited in situ by seismic waves. No earthquake-related infrasonic effects have been observed in the ionosphere. Necessity to make vibration tests of instruments is pointed out in order to be sure that observed effects are not effects of mechanical shaking of the instrument.     

Greenhouse gas emission associated with sugar production in southern Brazil

Background  

Since sugarcane areas have increased rapidly in Brazil, the contribution of the sugarcane production, and, especially, of the sugarcane harvest system to the greenhouse gas emissions of the country is an issue of national concern. Here we analyze some data characterizing various activities of two sugarcane mills during the harvest period of 2006-2007 and quantify the carbon footprint of sugar production.     

Sedimentary evidence of prehistoric distant-source tsunamis in the Hawaiian Islands

Over the past 200 years of written records, the Hawaiian Islands have experienced tens of tsunamis generated by earthquakes in the subduction zones of the Pacific ‘Ring of Fire’ (for example, Alaska–Aleutian, Kuril–Kamchatka, Chile and Japan). Mapping and dating anomalous beds of sand and silt deposited by tsunamis in low-lying areas along Pacific coasts, even those distant from subduction zones, is critical for assessing tsunami hazard throughout the Pacific basin. This study searched for evidence of tsunami inundation using stratigraphic and sedimentological analyses of potential tsunami deposits beneath present and former Hawaiian wetlands, coastal lagoons, and river floodplains. Coastal wetland sites on the islands of Hawai΄i, Maui, O΄ahu and Kaua΄i were selected based on historical tsunami runup, numerical inundation modelling, proximity to sandy source sediments, degree of historical wetland disturbance, and breadth of prior geological and archaeological investigations. Sand beds containing marine calcareous sediment within peaty and/or muddy wetland deposits on the north and north-eastern shores of Kaua΄i, O΄ahu and Hawai΄i were interpreted as tsunami deposits. At some sites, deposits of the 1946 and 1957 Aleutian tsunamis are analogues for deeper, older probable tsunami deposits. Radiocarbon-based age models date sand beds from three sites to ca 700 to 500 cal yr bp , which overlaps ages for tsunami deposits in the eastern Aleutian Islands that record a local subduction zone earthquake. The overlapping modelled ages for tsunami deposits at the study sites support a plausible correlation with an eastern Aleutian earthquake source for a large prehistoric tsunami in the Hawaiian Islands.     

������GPSվ���΢��̬�仯�����о�

???????GPS????????????????????????????λ????????????????????????????о????????????λ??仯?????????????????????????????2006????????2008?????????????????仯?????????????????????????2006??????????λ???????2004??2010??????????????????????2006??3??4????????????????????????????????????????????????????2006??7??4????????????????????????????Ms5.1??????2008????????????????λ?????????????????????Ms8.0?????????????α?     

Accurate Measurement of Lithium Isotopes in Eleven Carbonate Reference Materials by MC‐ICP‐MS with Soft Extraction Mode and 1012 Ω Resistor High‐Gain Faraday Amplifiers

Lithium isotopes in carbonate rocks and minerals can serve as important tools for assessing palaeoclimates and palaeoenvironments. However, carbonate bulk rock samples are commonly mixtures of carbonate and silicate minerals, which require the complete digestion of the carbonate without digesting the silicate. Additionally, the low Li content (ng g^?1 level) in carbonates provides an additional challenge. Hence, despite their wide applications, few carbonates have had their δ⁷Li values characterised, particularly carbonate reference materials, which hinders comparisons of Li isotope measurement results obtained in different laboratories and the further application of Li isotopes in geological studies. This study aimed to provide precise and accurate δ⁷Li values for carbonate reference materials based on an evaluation of sample leaching and the Li purification method for carbonates, as well as the adoption of soft extraction and 10¹² Ω amplifiers to increase the intensity/blank ratio and matrix effect on Li isotope measurement. The precision and accuracy of the proposed procedure were verified by analysing synthetic carbonate samples and mono‐elemental Li solutions. With the developed method we provide δ⁷Li values for eleven carbonate reference materials with a precision of ~ 0.4‰. The accuracy of the δ⁷Li values was validated using the standard addition method.