Geometric consistency for regression model estimation and testing in climatology and meteorology

Abstract

The effects of outliers on linear regression are examined. The sensitivity of classical least‐squares (LS) procedures to outliers is shown to be associated with the geometric inconsistency between the data space and the analysis space. This is illustrated for both estimation and inference. A geometrically consistent procedure based on the Euclidean distance is proposed. This procedure involves the least absolute deviation (LAD) regression and a new permutation test for matched pairs (PTMP). Comparisons made with LS techniques demonstrate that the proposed procedure is more resistant to the existence of outliers in the data set and leads to more intuitive results. Applications and illustrations using meteorological and climatological data are also discussed.     

A simple model of hillslope response for overland flow generation

This paper deals with the derivation of the hydrological response of a hillslope on the assumption of quick runoff by surface runoff generation. By using the simple non‐linear storage based model, first proposed by Horton, an analytical solution of the overland flow equations over a plane hillslope was derived. This solution establishes a generalization for different flow regimes of Horton's original solution, which is valid for the transitional flow regime only. The solution proposed was compared successfully with that of Horton and, for the turbulent flow regime, to the one derived from kinematic wave theory. This solution can be applied easily to both stationary and non‐stationary rainfall excess events. An analytical solution for the instantaneous response function (IRF) was also derived. Finally, simple expressions to compute peak and time to peak of IRF are proposed. Copyright © 2001 John Wiley & Sons, Ltd.     

Jupiter's ammonia clouds—localized or ubiquitous?

From an analysis of the Galileo Near Infrared Imaging Spectrometer (NIMS) data, Baines et al. (Icarus 159 (2002) 74) have reported that spectrally identifiable ammonia clouds (SIACs) cover less than 1% of Jupiter. Localized ammonia clouds have been identified also in the Cassini Composite Infrared Spectrometer (CIRS) observations (Planet. Space Sci. 52 (2004a) 385). Yet, ground-based, satellite and spacecraft observations show that clouds exist everywhere on Jupiter. Thermochemical models also predict that Jupiter must be covered with clouds, with the top layer made up of ammonia ice. For a solar composition atmosphere, models predict the base of the ammonia clouds to be at 720 mb, at 1000 mb if N/H were 4×solar, and at 0.5 bar for depleted ammonia of 10⁻²×solar (Planet. Space Sci. 47 (1999) 1243). Thus, the above NIMS and CIRS findings are seemingly at odds with other observations and cloud physics models. We suggest that the clouds of ammonia ice are ubiquitous on Jupiter, but that spectral identification of all but the freshest of the ammonia clouds and high altitude ammonia haze is inhibited by a combination of (i) dusting, starting with hydrocarbon haze particles falling from Jupiter's stratosphere and combining with an even much larger source—the hydrazine haze; (ii) cloud properties, including ammonia aerosol particle size effects. In this paper, we investigate the role of photochemical haze and find that a substantial amount of haze material can deposit on the upper cloud layer of Jupiter, possibly enough to mask its spectral signature. The stratospheric haze particles result from condensation of polycyclic aromatic hydrocarbons (PAHs), whereas hydrazine ice is formed from ammonia photochemistry. We anticipate similar conditions to prevail on Saturn.     

“Dust” streaks on Mars

Global mapping and photometry of selected areas on Mars are used to investigate the nature of bright and dark wind streaks that extend from topographic obstacles. Occurrence of both bright and dark streaks is strongly latitude dependent and is only weakly correlated with surface properties such as albedo and thermal inertia. Data on the colors, albedos, and phase behavior of streaks are consistent with models of bright streaks as mosaics of plains material and brighter, redder dust. Less than 20% of the ground need be covered by the optically thick dust in the brightest parts of the streaks; the amount of dust in optically thick layers could be as little as 10^?3 g/cm². Dark streaks can be interpreted as erosional windows in a patchy dust cover. Our model of dust deposition in optically thick patches is sedimentologically different from scenarios involving the deposition of ubiquitous, optically thin layers. It has the advantage that large amounts of dust can be deposited without affecting regional albedos.     

A study on the flux,speciation, and budget of mercury in controlled experimental ecosystems

The lowest addition of mercury (0.1 ug Hg 1⁻¹) was used in CEEs for research on mercury flux, speciation and budget. The removal behavior of mercury by phytoplankton in water columns of CEEs can be described by first order kinetic equations for total and particulate mercury in the CEE spiked by mercury. The removal rate of mercury in water columns depends on the size and productivity of phytoplanton in a water column to which mercuric ions were added. A 4.4 day half-life time and a 2.8 day half-life time for total and particulate mercury respectively was obtained in diatom bloom. During microflagellate bloom a 30 day total mercury half-life time was estimated with increase of particulate mercury in the water column. The 0.010 ug Hg cm⁻²y⁻¹ mercury flux rate that was attained in the control bag agreed with the values from field measurements in Saanich Inlet where the bags were launched. The proportion of total mercury to dissolved and particulate mercury depended also on the size, productivity, and concentration of mercury in a water column. A more or less constant distribution of mercury species in the control bag was observed as follows: dissolved Hg 0.73, particulate Hg 0.27, inorganic Hg 0.42, dissolved organic Hg 0.31. After spiking with mercury, the particulate mercury rose rapidly and reached to over 70% of the total mercury. The concentration factor of mercury by phytoplankton in the CEEs in the order of 10⁵ was consistent with the results from field measurements in Saanich Inlet. The mercury recovery from the water column, sediment, water with sediment, and the CEE bag walls was only 52.3% of total mercury spiked in the CEE. The losses of mercury by vaporization into the ambient air and diffusion through the wall of the enclosure should be considered. This paper was published in Chinese inOcean. Limn. Sinica,17(4):307–317, 1986.     

Isolation of PAH-degrading bacteria from mangrove sediments and their biodegradation potential

Surface sediment samples were collected from seven mangrove swamps in Hong Kong SAR with different degrees of contamination. The total concentrations of 16 PAHs in these sediments ranged from 169.41 to 1058.37 ng g⁻¹ with the highest concentration found in Ma Wan and the lowest in Kei Ling Ha Lo Wai mangrove swamp. In each swamp, three bacterial consortia were enriched from sediments using phenanthrene (Phe) as the sole carbon and energy source, and individual bacterial colony showing Phe degradation was isolated and identified by 16S rDNA gene sequence. The consortia enriched from Sai Keng and Ho Chung sediments had highest ability to degrade mixed PAHs in liquid medium, with 90% Phe and Fla (fluoranthene) degraded in 7 days. On the other hand, Kei Ling Ha Lo Wai-enriched consortia degraded less than 40% Phe and Fla. Pyrene (Pyr) was hardly degraded by the consortia enriched from sediments. Bacterial isolates, namely Rhodococcus (HCCS), Sphingomonas (MWFG) and Paracoccus (SPNT) were capable to degrade mixed PAHs (Phe + Fla + Pyr). Their degradation percentages could be lower, comparable or even higher than their respective enriched consortia, depending on the consortium and the type of PAH compounds. These results suggest that PAH-degrading bacteria enriched from mangrove sediments, either as a mixed culture or as a single isolate could be used for PAHs bioremediation.     

Natural attenuation, biostimulation and bioaugmentation on biodegradation of polycyclic aromatic hydrocarbons (PAHs) in mangrove sediments

The biodegradability of a mixture of PAHs, namely fluorene (Fl), phenanthrene (Phe) and pyrene (Pyr), in mangrove sediment slurry was investigated. At the end of week 4, natural attenuation based on the presence of autochthonous microorganisms degraded more than 99% Fl and Phe but only around 30% of Pyr were degraded. Biostimulation with addition of mineral salt medium degraded over 97% of all three PAHs, showing that nutrient amendment could enhance Pyr degradation. Bioaugmentation with inoculation of a PAH-degrading bacterial consortium enriched from mangrove sediments did not show any promotion effect and the degradation percentages of three PAHs were similar to that by natural attenuation. Some inhibitory effect was observed in bioaugmentation treatment in week 1 with only 50% Fl and 70% Phe degraded. These results indicate that autochthonous microbes may interact and even compete with the enriched consortium during PAH biodegradation. Natural attenuation appeared to be the most appropriate way to remedy Fl- and Phe-contaminated mangrove sediments while biostimulation was more capable to degrade Pyr-contaminated sediments. The study also shows that although a large portion of the added PAHs (more than 95%) was adsorbed onto the sediments at the beginning of the experiment, most PAHs were degraded in 4 weeks, suggesting that the degraders could utilize the adsorbed PAHs efficiently.     

Controlling mechanisms for the trace metals, (Cu,Zn, Cd,Fe, Co and Ni) in anoxic sea water in Saanich Inlet,British Columbia,Canada

A study was conducted on Cu, Zn, Cd, Fe, Co and Ni levels in Saanich Inlet anoxic seawater. Data on the concentration of these trace metals and H₂S, and other oceanographic parameters were obtained in four cruises. Equilibrium models are presented together with in situ data. The results strongly support the assumption that the solubilities of Cu, Zn, Cd, Fe, Co and Ni are controlled by bisulfide and/or polysulfide complexes. The species of Cu, Zn, Cd and Ni are shown to be Cu(S₄) ₂ ³⁻ and Cu(Sn₄S₅)³⁻, Zn(HS) ₂ ⁰ and ZnHS ₂ ⁻ , Cd(HS) ₂ ⁰ and Ni(HS) ₂ ⁰ , respectively. The solid species controlling Fe²⁺ and Co²⁺ are respectively the pyrohotites FeS and CoS. The data illustrates that thermodynamic equilibrium has been established in the H₂S-controlled seawater of Saanich Inlet, and that equilibrium has not been established in the H₂S−O₂-coexisted seawater of Saanich Inlet. The lack of equilibrium in the H₂S−O₂-coexisted seawater is a result of the flushing or mixing of oxygenated seawater with anoxic water. The new species of trace metals are still in the processes of reduction and precipitation. Contribution No. 1428 from Institute of Oceanology, Academia Sinica