Comparing Human and Neural Network Learning of Climate Categories

This study compares how humans and neural networks classify climate types. Human subjects were asked to classify climates from monthly temperature and precipitation patterns. To model their learning process, the same data were used to produce input vectors that trained a pattern associator neural network. Both human subjects and the neural network classified climates accurately after 10 rounds of supervised learning. The neural network successfully modeled the rate of human learning and the ability to learn specific climate categories. Moreover, the neural network weights used to classify climates correspond to distinct visual characteristics in temperature and precipitation. These results suggest that neural networks can model the formation of visual categories.     

A Boolean delay equation model of an interdecadal Arctic climate cycle

A Boolean delay equation (BDE) model is presented for the interdecadal Arctic and Greenland Sea climate cycle recently proposed by Mysak, Manak and Marsden. It is shown that 15- to 20-year oscillations can occur in the model for a variety of time delays in the BDEs. However, both the period and structure of the oscillations are sensitive to the initial conditions. In an extended model, in which the convection in the Greenland Sea is dependent upon the ice conditions during each of several previous years as well as the current year, the solution structure is more realistic, with two jumps per period of oscillation.     

Fluctuation dissipation in a general circulation model

This paper considers the climate response to step function changes in the solar constant in two versions of a general circulation model with simplified geography. The NCAR CCM0 model is applied to an all-land planet with no topography (Terra Blanda). In one version there is moisture in the air (as well as self-generated clouds) as evaporated from an ideal surface at a fixed 80% of saturation. In the other version there is no moisture in the atmosphere. We examine the decay of natural anomalies in the large-scale temperature field in each model and compare the time dependence of the ensemble average with the average temporal behavior of the response to step function changes in the solar constant. The fluctuation-dissipation theorem of statistical mechanics makes specific predictions about the relationship between the two curves. We conduct the experiments for both versions of the model since the sensitivity is quite different for each. The theorem is found to hold reasonably well in each case.     

Global vegetation change predicted by the modified Budyko model

A modified Budyko global vegetation model is used to predict changes in global vegetation patterns resulting from climate change (CO₂ doubling). Vegetation patterns are predicted using a model based on a dryness index and potential evaporation determined by solving radiation balance equations. Climate change scenarios are derived from predictions from four General Circulation Models (GCM's) of the atmosphere (GFDL, GISS, OSU, and UKMO). Global vegetation maps after climate change are compared to the current climate vegetation map using the kappa statistic for judging agreement, as well as by calculating area statistics. All four GCM scenarios show similar trends in vegetation shifts and in areas that remain stable, although the UKMO scenario predicts greater warming than the others. Climate change maps produced by all four GCM scenarios show good agreement with the current climate vegetation map for the globe as a whole, although over half of the vegetation classes show only poor to fair agreement. The most stable areas are Desert and Ice/Polar Desert. Because most of the predicted warming is concentrated in the Boreal and Temperate zones, vegetation there is predicted to undergo the greatest change. Specifically, all Boreal vegetation classes are predicted to shrink. The interrelated classes of Tundra, Taiga, and Temperate Forest are predicted to replace much of their poleward (mostly northern) neighbors. Most vegetation classes in the Subtropics and Tropics are predicted to expand. Any shift in the Tropics favoring either Forest over Savanna, or vice versa, will be determined by the magnitude of the increased precipitation accompanying global warming. Although the model predicts equilibrium conditions to which many plant species cannot adjust (through migration or microevolution) in the 50–100 y needed for CO₂ doubling, it is nevertheless not clear if projected global warming will result in drastic or benign vegetation change.     

World Data Center-A for Paleoclimatology at the NOAA Paleoclimatology Program,Boulder, CO

Conclusion The World Data Center-A for Paleoclimatology, located in the NOAA/NGDC Paleoclimatology Program, is committed to providing the scientific community with easy access to all paleoenvironmental data. Efforts to make archived data readily available include international coordination of data acquisition, management, and distribution, sponsoring workshops and data cooperatives to facilitate the compilation of important data sets, development of a browse and visualization software package (PaleoVu), and dispersal of archived data on magnetic media or over ANONYMOUS FTP/INTERNET. The program publishes a semi-annual newsletter that highlights latest developments and accomplishments in the area of paleoenvironmental data for global change research. Contributions to the newsletter are welcome from researchers describing their efforts to coordinate the free flow of paleoclimate data throughout the international scientific community.For information on the program or to be added to the mailing list contact Mrs Mildred England (phone: 303-497-6227; Fax: 303 497-6513; e-mail: MKE@mail.ngdc.noaa.gov), NOAA National Geophysical Data Center, Paleoclimatology Program/World Data Center-A for Paleoclimatology, 325 Broadway, E/GC Boulder, CO 80303 USA     

Early use by fish of a mitigation salt marsh,Humboldt Bay,California

Fish numbers and biomass in a mitigation salt marsh, Humboldt Bay, California, were examined from July 1981 to October 1982 and were compared with a nearby established marsh to determine whether the restored marsh provided adequate mitigation for habitat lost due to construction of a nearby marina. The use by fish of channels adjacent to the two marshes and the channel at the Woodley Island Marina construction site, for which mitigation was required, were also compared. The mitigation marsh, adjacent to Freshwater Slough channel, was 5.2 km from the marina site. Fishes were sampled by otter trawl, ichthyoplankton net, fixed channel nets, drop traps, and beach seines. Thirty-one fish species and two crab species were collected. Wide ranges in seasonal salinities and water temperatures, and differences in marsh elevation influenced fish use of the mitigation marsh area. The intertidal area of the mitigation marsh, dominated by euryhaline sticklebacks and topsmelts, did not replace intertidal and subtidal habitat lost by marina construction, which had more stable salinities and water temperatures and was used extensively by juvenile English sole.     

Distribution and abundance of the cownose ray,Rhinoptera bonasus,in lower Chesapeake Bay

Aerial surveys were conducted in the lower Chesapeake Bay during 1986–1989 to estimate abundance and examine the distribution of the cownose ray,Rhinoptera bonasus, during its seasonal residence, May–October. Most of the survey effort was concentrated in the lower and mid-bay regions. Cownose rays appeared uniformly distributed across the bay during mid-summer, but were more abundant in the eastern portion of the bay during migration. North-south distribution varied and reflected the general seasonal migration pattern. Mean abundance increased stepwise monthly from June through September and declined dramatically in October with their emigration from the bay. Abundance estimates from individual surveys varied. The greatest range of individual survey abundance estimates occurred in September (0–3.7×10⁷ cownose rays0 due to high variation in school size and abundance between surveys. Monthly mean cownose ray abundance ranged from 0 in May and November to an estimated maximum of 9.3×10⁶ individuals in September. The magnitude of the population suggests that the cownose ray plays an important role in the trophic dynamics of the Chesapeake Bay ecosystem. The historical data were insufficient to determine whether the population has increased, but these surveys provided the baseline data which would allow future investigation of cownose ray population dynamics in lower Chesapeake Bay.     

Determination of part-per-trillion levels of atmospheric dimethyl sulfide by isotope dilution gas chromatography/mass spectrometry

Stable isotopic dilution was applied to the determination of dimethyl sulfide (DMS) in ambient air at the low part-per-trillion by volume (pptrv) levels. Perdeuterated DMS was used as an internal standard in the gas chromatography/mass spectrometry determination. The isotopically labelled internal standard provided insensitivity to possible losses of DMS in sampling and analysis. The lower limit of detection (LLD) was 1 pptrv with a sample acquisition time of 2 min.     

Subsurface Transport of Inorganic and Organic Solutes from Experimental Road Spreading of Oil-Field Brine

A study designed to evaluate ground water quality changes resulting from spreading oil-field brine on roads for ice and dust control was conducted using a gravel roadbed that received weekly applications of brine eight times during the winter phase and 11 times during the summer phase of the study. A network of 11 monitoring wells and five pressure-vacuum lysimeters was installed to obtain ground water and soil water samples. Thirteen sets of water- quality samples were collected and analyzed for major ions, trace metals, and volatile organic compounds. Two sets of samples were taken prior to brine spreading, four sets during winter-phase spreading, five sets during summer- phase spreading, and two sets during the interim between the winter and summer phases. A brine plume delineated by elevated specific-conductance values and elevated chloride concentrations developed downgradient of the roadbed during both the winter and summer phases. The brine plume caused chloride concentrations in ground water samples to exceed U.S. EPA public drinking-water standards by two-fold during the winter phase and five-fold during the summer phase. No other major ions, trace metals, or volatile organic compounds exceeded the standards during the winter or summer phases. More than 99 percent dilution of the solutes in the brine occurred between the roadbed surface and the local ground water flow system. Further attenuation of calcium, sodium, potassium, and strontium resulted from adsorption, whereas further attenuation of benzene resulted from volatilization and adsorption.