Machine learning for cross-gazetteer matching of natural features

ABSTRACT

Defining and identifying duplicate records in a dataset is a challenging task which grows more complex when the modeled entities themselves are hard to delineate. In the geospatial domain, it may not be clear where a mountain, stream, or valley ends and begins, a problem carried over when such entities are catalogued in gazetteers. In this paper, we take two gazetteers, GeoNames and SwissNames3D, and perform matching – identifying records in each that are about the same entity – across a sample of natural feature records. We first perform rule-based matching, establishing competitive results, then apply machine learning using Random Forests, a method well-suited to the matching task. We report on the performance of a wider array of matching features than has been previously studied, including domain-specific ones such as feature type, land cover class, and elevation. Our results show an increase in performance using machine learning over rules, with a notable performance gain from considering feature types, but negligible gains from other specialized matching features. We argue that future work in this area should strive to be more reproducible and report results on a realistic testing pipeline including candidate selection, feature extraction, and classification.     

Flake Morphology,Fluvial Dynamics,and Debitage Transport Potential

Since archaeological sites are often located in or near fluvial environments, a thorough understanding of corresponding formation processes is a critical component of archaeological research. Within this context, we address the following question: how does flake morphology condition transport potential? Specifically, do flakes behave as the idealized spheres often used to characterize geological materials and if so, how might this process modify a debitage assemblage? To address this important issue, we present the results of a fluvial transport experiment designed to evaluate the relationships between flake size, shape, weight, curvature, and transport potential. We begin with an overview of fluvial mechanics and previous archaeological fluvial transport studies and use these studies to design an experiment intended to fill gaps in our knowledge of the relationships between flake morphology and transport potential. Our experimental results suggest that all things being equal, flake weight is the most important variable in both settling velocity and transport distance, although flake shape conditions transport as well. Finally, our study demonstrates the potential structure of fluvially transported assemblages as lag deposits and suggests that fluvially transported assemblages may retain key behavioral information.     

Community structure and grazing of the nano-microzooplankton on the continental shelf of the Bay of Biscay

In order to investigate the parameters controlling the heterotrophic protists (nano-microzooplankton) on the continental shelf of the southern Bay of Biscay, plankton communities and their physico-chemical environment were studied 4 times in February, April, June and September–October 2004 at three stations in the euphotic zone in the Bay of Biscay. The abundance and carbon biomass of heterotrophic protists (ciliates, heterotrophic dinoflagellates and nanoflagellates) as well as all the others groups of plankton (picoplankton, nanophytoplankton, diatoms, autotrophic dinoflagellates, metazoan microzooplankton and mesozooplankton), the environmental parameters and the primary and bacteria production were evaluated at each sampling period. Microzooplankton grazing experiments were undertaken at the same time. Ciliates and heterotrophic dinoflagellates accounted for the main major component of nano- and microzooplankton communities in term of biomass. The total carbon biomass of heterotrophic protists was highest in spring and lowest at the end of summer. The development of heterotrophic protists started after a winter microphytoplankton bloom (principally large diatoms), the biomass was lower in June and was low in September (through inappropriate prey). The carbon requirement of microzooplankton ranged from 50 to more than 100% of daily primary, bacterial and nanoflagellate production. The heterotrophic protist community was predominantly constrained by bottom-up control in spring and at the end of summer via food availability and quality.     

Robustness and uncertainties of seismic damage estimates at urban scale: a methodological comparison on the example of the city of Oran (Algeria)

The city of Oran is exposed to a significant seismic hazard, as almost all the northern Algeria territory, where numerous casualties and severe damage occurred in the last decades due to several moderate to large earthquakes. A mitigation policy should include the establishment of priorities to reduce the vulnerability of existing buildings based on the knowledge of the actual urban fabrics. The complexity of vulnerability assessment requires a gradual approach from the urban scale to the building scale. The study reported in this paper corresponds to the first step of such an approach, i.e., a preliminary study of the seismic vulnerability and expected damage within an urban district of the city of Oran, based on a non-dedicated data base from a building survey previously performed for other purposes. The main goals of this study are twofold: (1) an assessment of the degree of uncertainty and robustness of such results through a comparison of the results derived from different urban vulnerability methods (GNDT 2; RISK-UE LM1; and VULNERALP 2.0) and (2) an assessment of the actual level of seismic risk in the city of Oran. Cross-method comparisons and correlations highlight a satisfactory agreement between mean damage estimates at the urban scale, despite significant scattering at the single building scale, and uncertainty levels which vary significantly from one method to the other. For a given scenario, the three methods provide damage estimates lying within half an EMS damage degree of one another, with some systematic positive bias for VULNERALP and negative bias for RISK-UE LM1, especially for masonry buildings. The expected mean damage is very important for intensities 9 and 10, with an average damage grade around 3–4 for intensity 9 and 4–5 for intensity 10. The spatial distribution of damage systematically exhibits larger values in the northern, older, commercial area, than in the southern, more recent and more residential area, in relation to the building typology and the existence of several aggravating factors. Some areas of higher vulnerability / damage can be distinguished, which should receive particular attention for retrofitting priorities or urban planning decisions, also taking into account their cultural heritage value.     

Secondary Hydrogeologic Regions of the Conterminous United States

The U.S. Geological Survey (USGS) previously identified and mapped 62 Principal Aquifers (PAs) in the U.S., with 57 located in the conterminous states. Areas outside of PAs, which account for about 40% of the conterminous U.S., were collectively identified as “other rocks.” This paper, for the first time, subdivides this large area into internally-consistent features, defined here as Secondary Hydrogeologic Regions (SHRs). SHRs are areas of other rock within which the rocks or deposits are of comparable age, lithology, geologic or physiographic setting, and relationship to the presence or absence of underling PAs or overlying glacial deposits. A total of 69 SHRs were identified. The number and size of SHRs identified in this paper are comparable to the number and size of PAs previously identified by the USGS. From a two-dimensional perspective, SHRs are complementary to PAs, mapped only where the PAs were not identified on the USGS PA map and not mapped where the PAs were identified. SHRs generally consist of low permeability rocks or deposits, but can include locally productive aquifers. The two maps, taken together, provide a comprehensive, national-scale hydrogeologic framework for assessing and understanding groundwater systems.     

Occurrence and concentration of caffeine in Oregon coastal waters

Caffeine, a biologically active drug, is recognized as a contaminant of freshwater and marine systems. We quantified caffeine concentrations in Oregon's coastal ocean to determine whether levels correlated with proximity to caffeine pollution sources. Caffeine was analyzed at 14 coastal locations, stratified between populated areas with sources of caffeine pollution and sparsely populated areas with no major caffeine pollution sources. Caffeine concentrations were measured in major water bodies discharging near sampling locations. Caffeine in seawater ranged from below the reporting limit (8.5 ng/L) to 44.7 ng/L. Caffeine occurrence and concentrations in seawater did not correspond with pollution threats from population density and point and non-point sources, but did correspond with storm event occurrence. Caffeine concentrations in rivers and estuaries draining to the coast ranged from below the reporting limit to 152.2 ng/L. This study establishes the occurrence of caffeine in Oregon's coastal waters, yet relative importance of sources, seasonal variability, and processes affecting caffeine transport into the coastal ocean require further research.     

STRATEGIC PLANNING FOR GEOGRAPHY DEPARTMENTS?

Strategic planning is a technique from the commercial world which could assist geography departments in adapting to today's more demanding administrative environments and thereby further the development of the discipline. This paper sets out the fundamentals of planning as a prelude to a generic departmental plan. Key inclusions are explained and certain problems of application are discussed, should departments adopt such an approach.     

Modelling of scattering of seismic waves from a corrugated rough sea surface: a comparison of three methods

We compare three numerical methods to model the sea surface interaction in a marine seismic reflection experiment (the frequencies considered are in the band 10–100 Hz): the finite-difference method (FDM), the spectral element method (SEM) and the Kirchhoff method (KM). A plane wave is incident at angles of 0° and 30° with respect to the vertical on a rough Pierson–Moskowitz surface with 2 m significant wave height and the response is synthesized at 6, 10 and 50 m below the average height of the sea surface. All three methods display an excellent agreement for the main reflected arrival. The FDM and SEM also agree very well all through the scattered coda. The KM shows some discrepancies, particularly in terms of amplitudes.     

Assessing the Role of pH in Determining Water Column Nitrification Rates in a Coastal System

Ocean acidification is predicted to impact the nitrogen cycle in a variety of ways. Specifically, manipulations of water column pH have shown that nitrification, the microbial conversion of ammonium to nitrate, is inhibited at low pH. A decrease in nitrification may impact phytoplankton composition and production, denitrification, and the production of nitrous oxide. We compiled an existing unique data set of concurrent water column nitrification rates and water column pH values from a temperate New England estuary (Narragansett Bay, RI, USA). Contrary to the current hypothesis, we found that nitrification rates were highest at low pH and significantly (P = 0.0031) lower at high water column pH. In this study, pH varied up to 0.85 units, 20% more than the maximum predicted ocean pH decrease of 0.7 units. These results highlight that nitrifying organisms in coastal systems tolerate a wide range of pH values. Moreover, the degree of negative correlation with pH may depend on site-specific environmental conditions. Combined, these findings indicate that the current hypothesis of the negative impacts of ocean acidification on nitrification, at least for the coastal ocean, might need reevaluation.     

Soil/landform relationships surrounding the Harappa archaeological site,Pakistan

A soil survey around the archaeological site of Harappa, Pakistan revealed alluvial deposits of five distinct ages based on relative position in the landscape and degree of soil profile development. the youngest deposit (age 1) is in the lowest landscape position and has received flood waters as recently as 1988. Soils there are in an incipient stage of development: only organic carbon and soluble salts have accumulated at the surface of the profile. the age 2 deposit has not undergone significant pedogenic change, but is in a slightly higher landscape position than the youngest deposit. Elevated concentrations of P, and the presence of sand-sized pottery and brick fragments, indicate that this deposit was derived at least partially from archaeological material. the presence of small, soft calcite nodules (Stage II) and some soluble salt translocation are the primary pedogenic changes observed in the age 3 deposit. the age 4 deposit shows evidence of both carbonate and gypsum accumulation. Presence of large gypsum nodules in deep By horizons suggests that a high groundwater table has altered these soils. the oldest deposit, age 5, forms a late Pleistocene stream terrace of the Ravi River. the soil formed in this deposit exhibits considerable carbonate accumulation, with large, dense nodules (Stage II + ) and an argillic horizon. A ¹⁴C date from pedogenic calcite gives an age of 7080 ± 90 years B.P., indicating a minimum age of early Holocene. the soil survey suggests that the ancient city of Harappa was built on an age 5 stream terrace remnant, surrounded by Holocene floodplains and a meandering channel of the Ravi River.