Efficient collection of training data for sub-pixel land cover classification using neural networks

Artificial neural networks (ANNs) are a popular class of techniques for performing soft classifications of satellite images. They have successfully been applied for estimating crop areas through sub-pixel classification of medium to low resolution images. Before a network can be used for classification and estimation, however, it has to be trained. The collection of the reference area fractions needed to train an ANN is often both time-consuming and expensive. This study focuses on strategies for decreasing the efforts needed to collect the necessary reference data, without compromising the accuracy of the resulting area estimates. Two aspects were studied: the spatial sampling scheme (i) and the possibility for reusing trained networks in multiple consecutive seasons (ii). Belgium was chosen as the study area because of the vast amount of reference data available. Time series of monthly NDVI composites for both SPOT-VGT and MODIS were used as the network inputs. The results showed that accurate regional crop area estimation (R² > 80%) is possible using only 1% of the entire area for network training, provided that the training samples used are representative for the land use variability present in the study area. Limiting the training samples to a specific subset of the population, either geographically or thematically, significantly decreased the accuracy of the estimates. The results also indicate that the use of ANNs trained with data from one season to estimate area fractions in another season is not to be recommended. The interannual variability observed in the endmembers’ spectral signatures underlines the importance of using up-to-date training samples. It can thus be concluded that the representativeness of the training samples, both regarding the spatial and the temporal aspects, is an important issue in crop area estimation using ANNs that should not easily be ignored.     

The Proterozoic,albitite-hosted,Valhalla uranium deposit,Queensland, Australia: a description of the alteration assemblage associated with uranium mineralisation in diamond drill hole V39

The Valhalla uranium deposit, located 40 km north of Mount Isa, Queensland, Australia, is an albitite-hosted, Mesoproterozoic U deposit similar to albitite-hosted uranium deposits in the Ukraine, Sweden, Brazil and Guyana. Uranium mineralisation is hosted by a thick package of interbedded fine-grained sandstones, arkoses and gritty siltstones that are bound by metabasalts belonging to the ca. 1,780 Ma Eastern Creek Volcanics in the Western Succession of the Mount Isa basin. Alteration associated with U mineralisation can be divided into an early, main and late stage. The early stage is dominated by laminated and intensely altered rock comprising albite, reibeckite, calcite, (titano)magnetite ± brannerite. The main stage of mineralisation is dominated by brecciated and intensely altered rocks that comprise laminated and intensely altered rock cemented by brannerite, apatite, (uranoan)-zircon, uraninite, anatase, albite, reibeckite, calcite and hematite. The late stage of mineralisation comprises uraninite, red hematite, dolomite, calcite, chlorite, quartz and Pb-, Fe-, Cu-sulfides. Brannerite has U–Pb and Pb–Pb ages that indicate formation between 1,555 and 1,510 Ma, with significant Pb loss evident at ca. 1,200 Ma, coincident with the assemblage of Rodinia. The oldest ages of the brannerite overlap with ⁴⁰Ar/³⁹Ar ages of 1,533 ± 9 Ma and 1,551 ± 7 Ma from early and main-stage reibeckite and are interpreted to represent the timing of formation of the deposit. These ages coincide with the timing of peak metamorphism in the Mount Isa area during the Isan Orogeny. Lithogeochemical assessment of whole rock data that includes mineralised and unmineralised samples from the greater Mount Isa district reveals that mineralisation involved the removal of K, Ba and Si and the addition of Na, Ca, U, V, Zr, P, Sr, F and Y. U/Th ratios indicate that the ore-forming fluid was oxidised, whereas the crystal chemistry of apatite and reibeckite within the ore zone suggests that F⁻ and were important ore-transporting complexes. δ¹⁸O values of co-existing calcite and reibeckite indicate that mineralisation occurred between 340 and 380°C and involved a fluid having δ¹⁸O_fluid values between 6.5 and 8.6‰. Reibeckite δD values reveal that the ore fluid had a δD_fluid value between −98 and −54‰. The mineral assemblages associated with early and main stages of alteration, plus δ¹⁸O_fluid and δD_fluid values, and timing of the U mineralisation are all very similar to those associated with Na–Ca alteration in the Eastern Succession of the Mount Isa basin, where a magmatic fluid is favoured for this style of alteration. However, isotopic data from Valhalla is also consistent with that from the nearby Mount Isa Cu deposit where a basinal brine is proposed for the transport of metals to the deposit. Based on the evidence to hand, the source fluids could have been derived from either or both the metasediments that underlie the Eastern Creek Volcanics or magmatism that is manifest in the Mount Isa area as small pegmatite dykes that intruded during the Isan Orogeny.     

Trace Metals in Selected Benthic Invertebrates from Belgian Coastal Waters (1981–1996)

Mercury, cadmium, lead, copper, zinc and chromium were assessed in brown shrimp (Crangon crangon), swimming crab (Liocarcinus holsatus), hermit crab (Pagurus bernhardus), starfish (Asterias rubens) and cut trough shell (Spisula subtruncata) from ten sampling stations off the Belgian coast, including three dredge spoils disposal sites. Calculations of time trends indicated that on the whole concentrations of trace metals had significantly decreased since 1981, also on dredged material dumping sites. Starfish and hermit crab were found to be the most promising benthic indicator species for metal contamination. Globally, the location of the sites, including the dredge spoils disposal zones, showed to be of little influence on the concentrations of trace metals.     

Estimation of commercial diamond grades based on microdiamonds: a case study of the Koidu diamond mine,Sierra Leone

This paper documents the application of a microdiamond-based approach to the estimation of diamond grade in the Pipe 1 kimberlite at the Koidu mine in Sierra Leone. A geological model of Pipe 1 was constructed to represent the distribution and volume of the dominant kimberlite units within the pipe. Bulk samples, along with representative microdiamond samples, were collected from these units at surface and were used to define the ratio between microdiamond stone frequency (+212 μm stones per kilogram) and recoverable macrodiamond grade (+1.2 mm carats per tonne; 1 carat = 0.2 g). These ratios were applied to a comprehensive, spatially representative microdiamond sample dataset and were combined with a spatial model of country-rock xenolith dilution within the pipe to estimate +1.2 mm recoverable grades. The resource estimate was reconciled with subsequent production results in the elevation range 160 to 100 m above sea level. Production results for each of the six 10 m benches covering this elevation range were compared to the estimated average grades for these zones in the pipe. For the five cases where most of the kimberlite mass on a given bench is represented in the production data, the results show a maximum discrepancy of 6% between predicted and reported production grade with no indication of any consistent bias. This indicates that, when supported by a sound geological model and suitable microdiamond and macrodiamond data, the microdiamond-based estimation approach can provide reliable constraints on macrodiamond grade, even in the case of geologically complex bodies such as Koidu Pipe 1.

     

X-ray imagery and physical core logging as a proxy of the content of sediment cores in cold-water coral mound provinces: a case study from Porcupine Seabight,SW of Ireland

Three provinces, characterized by the presence of carbonate mounds interpreted as cold-water coral banks have been reported in Porcupine Seabight, SW of Ireland and were recently subjected to many detailed studies. This contribution discusses the use of X-ray imagery and physical properties in deciphering palaeoceanographic, sedimentological and biological processes. Physical property core logging and X-ray imagery are used to identify and describe sedimentation regimes and so their respective palaeoceanographic and palaeoclimatological settings in two mound provinces, respectively the Belgica mound province and the Magellan mound province. Both provinces show at present time clear differences in the hydrodynamic environment. This study confirms that also during the past the oceanographic and sedimentological environment of both provinces differ clearly. Impacts of glacial–interglacial variations and locally derived ice rafting events (IRE), comparable with the North Atlantic Heinrich events (HE) have been recognized in both provinces. Moreover, the combination of X-ray imagery, magnetic susceptibility, gamma density and P-wave velocity makes it possible to estimate the coral content and coral distribution in unopened cores localized on top of carbonate mounds. A comparison between on-mound and off-mound cores in both provinces allowed revealing some mechanisms of mound evolution and coral growth versus time.     

Sequence stratigraphy of the Mississippi fan (Plio-Pleistocene), Gulf of Mexico

The Mississippi Fan is a large, mud-dominated submarine fan over 4 km thick, deposited in the deep Gulf of Mexico during the late Pliocene and Pleistocene. Analysis of 19,000 km of multifold seismic data defined 17 seismic sequences, each characterized by channel, levee, and associated overbank deposits, as well as mass transport deposits. At the base of nine sequences are a series of seismic facies consisting of mounded, hummocky, chaotic, and subparallel reflections, which constitute 10–20% of the sediments in each the sequences. These facies are externally mounded and occur in two general regions of the fan: (1) in the upper and middle fan they are elongate in shape and mimic the channel's distribution; (2) in the middle fan to lower fan they are characterized by a fan-shaped distribution, increasing in width downfan. These facies are interpreted to have formed as disorganized slides, debris flows, and turbidites (informally called “mass transport complexes”). Overlying this basal interval, characteristic of all sequences, are well-developed channel-levee systems that constitute 80–90% of the fan's sediments. Channels consist of high amplitude, subparallel reflections, whereas the flanking levee sediments appear as subparallel reflections that have high amplitudes at the base changing upward to low amplitude. The vertical change in amplitude may reflect a decrease in grain size and bed thicknesses. Overbank sediments are characterized by interbedded subparallel to hummocky and mounded reflections, suggesting both turbidites from the channel, as well as slides and debris flows derived both locally and from the slope updip.     

Temporal patterns and controls on runoff magnitude and solution chemistry of urban catchments in the semiarid southwestern United States

Urban expansion and the scarcity of water supplies in arid and semiarid regions have increased the importance of urban runoff to localized water resources. However, urban catchment responses to precipitation are poorly understood in semiarid regions where intense rainfall often results in large runoff events during the short summer monsoon season. To evaluate how urban runoff quantity and quality respond to rainfall magnitude and timing, we collected stream stage data and runoff samples throughout the 2007 and 2008 summer monsoons from four ephemeral drainages in Tucson, Arizona. Antecedent rainfall explained 20% to 30% of discharge (mm) and runoff ratio in the least impervious (22%) catchment but was not statistically related to hydrologic responses at more impervious sites. Regression models indicated that rainfall depth, imperviousness and their combined effect control discharge and runoff ratios (p < 0.01, r² = 0.91 and 0.75, respectively). In contrast, runoff quality did not vary with imperviousness or catchment size. Rainfall depth and duration, time since antecedent rainfall and event and cumulative discharge controlled runoff hydrochemistry and resulted in five specific solute response patterns: (i) strong event and seasonal solute mobilization (solute flush), (ii) event chemostasis and strong seasonal flush, (iii) event chemostasis and weak seasonal flush, (iv) event and seasonal chemostasis and (v) late seasonal flush. Our results indicate that hydrologic responses of semiarid catchments are controlled by rainfall partitioning at the event scale, whereas wetting magnitude, frequency and timing alter solute stores readily available for transport and control temporal runoff quality. Copyright © 2012 John Wiley & Sons, Ltd.     

Seismic loss estimation tool as rapid survey for prioritizing buildings for disaster preparedness: case study to hospital buildings

Natural Hazards - Hospital buildings must be fully operational after the earthquake to protect the lives of patients as well as to provide emergency care and medical treatment to the victims....