Limitations of Majority Agreement in Crowdsourced Image Interpretation

Crowdsourcing can efficiently complete tasks that are difficult to automate, but the quality of crowdsourced data is tricky to evaluate. Algorithms to grade volunteer work often assume that all tasks are similarly difficult, an assumption that is frequently false. We use a cropland identification game with over 2,600 participants and 165,000 unique tasks to investigate how best to evaluate the difficulty of crowdsourced tasks and to what extent this is possible based on volunteer responses alone. Inter‐volunteer agreement exceeded 90% for about 80% of the images and was negatively correlated with volunteer‐expressed uncertainty about image classification. A total of 343 relatively difficult images were independently classified as cropland, non‐cropland or impossible by two experts. The experts disagreed weakly (one said impossible while the other rated as cropland or non‐cropland) on 27% of the images, but disagreed strongly (cropland vs. non‐cropland) on only 7%. Inter‐volunteer disagreement increased significantly with inter‐expert disagreement. While volunteers agreed with expert classifications for most images, over 20% would have been mis‐categorized if only the volunteers’ majority vote was used. We end with a series of recommendations for managing the challenges posed by heterogeneous tasks in crowdsourcing campaigns.     

Spatial and temporal variability in South San Francisco Bay (USA). I. Horizontal distributions of salinity, suspended sediments, and phytoplankton biomass and productivity

The horizontal pattern of mesoscale (1–4 km) variability in salinity was a poor predictor of mesoscale patterns in chlorophyll a, suspended particulate matter, and daily primary productivity in the South San Francisco Bay estuary during spring 1987. The tidally-averaged salinity distribution varied over weekly time scales, reflecting inputs of freshwater as well as transport processes. Spatial distributions of the other quantities also varied weekly, but not in concert with the salt field. Spatial patterns of phytoplankton biomass (chlorophyll a) deviated from the salinity patterns, largely reflecting in situ production of phytoplankton biomass during the spring bloom. The tidally-averaged distribution of suspended particulate matter (SPM) was highly dynamic and responded to (1) the riverine input of suspended sediment during a freshet, (2) neap-spring variations in tidally-driven resuspension, and (3) resuspension in shallows following a period of wind mixing. Two-dimensional distributions of primary productivity P′, derived from maps of biomass and turbidity (SPM), also varied weekly, but the spatial variability of P′ was only about half that of SPM and chlorophyll. Since the magnitude and patterns of spatial variability differ among nonconservative quantities, at least in part because of local sources and sinks, we conclude that the spatial distributions of nonconservative quantities cannot be predicted from distributions of conservative tracers, such as salinity.     

Growth of estuarine fish is associated with the combined concentration of sediment contaminants and shows no adaptation or acclimation to past conditions

We tested whether the growth rates of small benthic fish (Gillichthys mirabilis) in three southern California estuaries corresponded with the local concentrations of contaminants. Fish originating from each estuary were transplanted to cages in each estuary in two reciprocal transplant experiments. The growth rates of caged fish, and the size-distribution of natural populations, showed the same pattern of difference among estuaries. Twelve metals and organic contaminants occurred in bulk sediments at concentrations close to their individual ERL values, and a simple index of their combined concentration (the mean ERL quotient) was inversely correlated to the growth of caged fish. Metals in the water column occurred at lower concentrations, relative to toxicity thresholds, than those in sediments and were unrelated to fish growth. Fish used in the field caging experiments, and other fish held in the laboratory under constant conditions, showed no difference in growth according to their estuary of origin. Fish originating from different estuaries also showed no consistent differences in their tissue burden of organic contaminants. Our results thus suggested no genetic adaptation or physiological acclimation to the past contaminant regime, but revealed a possible association between fish growth rates and the combined concentration of multiple sediment contaminants.     

Effects of Direct Ocean CO<Subscript>2</Subscript> Injection on Deep-Sea Meiofauna

Purposeful deep-sea carbon dioxide sequestration by direct injection of liquid CO₂ into the deep waters of the ocean has the potential to mitigate the rapid rise in atmospheric levels of greenhouse gases. One issue of concern for this carbon sequestration option is the impact of changes in seawater chemistry caused by CO₂ injection on deep-sea ecosystems. The effects of deep-sea carbon dioxide injection on infaunal deep-sea organisms were evaluated during a field experiment in 3600 m depth off California, in which liquid CO₂ was released on the seafloor. Exposure to the dissolution plume emanating from the liquid CO₂ resulted in high rates of mortality for flagellates, amoebae, and nematodes inhabiting sediments in close proximity to sites of CO₂ release. Results from this study indicate that large changes in seawater chemistry (i.e. pH reductions of ∼0.5–1.0 pH units) near CO₂ release sites will cause high mortality rates for nearby infaunal deep-sea communities. This revised version was published online in July 2006 with corrections to the Cover Date.     

Chemical characteristics of atmospheric bulk deposition in a semi-rural area of the <Emphasis Type="Bold">Po Valley</Emphasis> (Italy)

This study provides an analysis of a five-year time series chemical composition of the bulk deposition (2009–2013), collected within a farm surrounded by industrial and urban settlements in a semi-rural area of the Po Valley, with the aim of characterizing potential emission sources affecting precipitation composition at the site. Most monitoring efforts in this region, recognized as one of the most polluted in the world both due to the intense industrialisation and urbanisation as well as to frequent air stagnation conditions, are presently devoted more to gaseous and particulate pollutants than to precipitation chemistry. The bulk deposition samples were very concentrated in chemical species, both acidic and alkaline, high compared to other polluted sites in the world and to locations in the same district. The mean ions concentrations (in μeq l^?1) are: NO₃ ^? (243) > SO₄ ^2? (220) > PO₄ ^3? (176) > Cl^? (153) > NO₂ ^? (29) > F^? (2.6); NH₄ ⁺ (504) > Ca²⁺ (489) > K⁺ (151) > Na⁺ (127) > Mg²⁺ (127). pH data shows a trend toward slightly alkaline conditions attributed to the large presence of ammonium and crustal elements, in spite of high concentrations of nitrates and sulphates. The relevant concentrations of Ca²⁺ and Mg²⁺ further suggests that these alkaline conditions might be due to the correspondingly significant concentrations of carbonates/bicarbonates in our dataset. While back-trajectories analysis suggests the stronger importance of local resuspension over long-range transport, statistical analyses on ion composition highlight the key role exerted by agricultural activity, especially in the case of NH₄ ⁺, K⁺, Ca²⁺ and PO₄ ^3? (especially linked to fertilisation practices and soil resuspension due to mechanical operations). Apart from Na⁺ and Cl^? ions which correlate well as expected, indicating their likely common origin from marine salt, the identification of the origin of the other ions is very complex due to the contribution of diverse local sources, such as industrial and residential settlements.     

Assessing the suitability of GlobeLand30 for mapping land cover in Germany

Global land cover (LC) maps have been widely employed as the base layer for a number of applications including climate change, food security, water quality, biodiversity, change detection, and environmental planning. Due to the importance of LC, there is a pressing need to increase the temporal and spatial resolution of global LC maps. A recent advance in this direction has been the GlobeLand30 dataset derived from Landsat imagery, which has been developed by the National Geomatics Center of China (NGCC). Although overall accuracy is greater than 80%, the NGCC would like help in assessing the accuracy of the product in different regions of the world. To assist in this process, this study compares the GlobeLand30 product with existing public and online datasets, that is, CORINE, Urban Atlas (UA), OpenStreetMap, and ATKIS for Germany in order to assess overall and per class agreement. The results of the analysis reveal high agreement of up to 92% between these datasets and GlobeLand30 but that large disagreements for certain classes are evident, in particular wetlands. However, overall, GlobeLand30 is shown to be a useful product for characterizing LC in Germany, and paves the way for further regional and national validation efforts.