Machine learning for radioxenon event classification for the Comprehensive Nuclear-Test-Ban Treaty

A method of weapon detection for the Comprehensive nuclear-Test-Ban-Treaty (CTBT) consists of monitoring the amount of radioxenon in the atmosphere by measuring and sampling the activity concentration of ^131mXe, ¹³³Xe, ^133mXe, and ¹³⁵Xe by radionuclide monitoring. Several explosion samples were simulated based on real data since the measured data of this type is quite rare. These data sets consisted of different circumstances of a nuclear explosion, and are used as training data sets to establish an effective classification model employing state-of-the-art technologies in machine learning. A study was conducted involving classic induction algorithms in machine learning including Naïve Bayes, Neural Networks, Decision Trees, k-Nearest Neighbors, and Support Vector Machines, that revealed that they can successfully be used in this practical application. In particular, our studies show that many induction algorithms in machine learning outperform a simple linear discriminator when a signal is found in a high radioxenon background environment.     

Landscape Correlates of Breeding Bird Richness Across the United States Mid-Atlantic Region

Using a new set of landscape indicator data generated by the U.S.EPA, and a comprehensive breeding bird database from the National Breeding Bird Survey, we evaluated associations between breeding bird richness and landscape characteristics across the entire mid-Atlantic region of the United States. We evaluated how these relationships varied among different groupings (guilds) of birds based on functional, structural, and compositional aspects of individual species demographics. Forest edge was by far the most important landscape attribute affecting the richness of the lumped specialist and generalist guilds; specialist species richness was negatively associated with forest edge and generalist richness was positively associated with forest edge. Landscape variables (indicators) explained a greater proportion of specialist species richness than the generalist guild (46% and 31%, respectively). The lower value in generalists may reflect finer-scale distributions of open habitat that go undetected by the Landsat satellite, open habitats created by roads (the areas from which breeding bird data are obtained), and the lumping of a wide variety of species into the generalist category. A further breakdown of species into 16 guilds showed considerable variation in the response of breeding birds to landscape conditions; forest obligate species had the strongest association with landscape indicators measured in this study (55% of the total variation explained) and forest generalists and open ground nesters the lowest (17% of the total variation explained). The variable response of guild species richness to landscape pattern suggests that one must consider species' demographics when assessing the consequences of landscape change on breeding birds.     

Detecting Temporal Change in Watershed Nutrient Yields

Meta-analyses reveal that nutrient yields tend to be higher for watersheds dominated by anthropogenic uses (e.g., urban, agriculture) and lower for watersheds dominated by natural vegetation. One implication of this pattern is that loss of natural vegetation will produce increases in watershed nutrient yields. Yet, the same meta-analyses also reveal that, absent land-cover change, watershed nutrient yields vary from one year to the next due to many exogenous factors. The interacting effects of land cover and exogenous factors suggest nutrient yields should be treated as distributions, and the effect of land-cover change should be examined by looking for significant changes in the distributions. We compiled nutrient yield distributions from published data. The published data included watersheds with homogeneous land cover that typically reported two or more years of annual nutrient yields for the same watershed. These data were used to construct statistical models, and the models were used to estimate changes in the nutrient yield distributions as a result of land-cover change. Land-cover changes were derived from the National Land Cover Database (NLCD). Total nitrogen (TN) yield distributions increased significantly for 35 of 1550 watersheds and decreased significantly for 51. Total phosphorus (TP) yield distributions increased significantly for 142 watersheds and decreased significantly for 17. The amount of land-cover change required to produce significant shifts in nutrient yield distributions was not constant. Small land-cover changes led to significant shifts in nutrient yield distributions when watersheds were dominated by natural vegetation, whereas much larger land-cover changes were needed to produce significant shifts when watersheds were dominated by urban or agriculture. We discuss our results in the context of the Clean Water Act.     

Forest carbon management in the United States: 1600-2100

This paper reviews the effects of past forest management on carbon stocks in the United States, and the challenges for managing forest carbon resources in the 21st century. Forests in the United States were in approximate carbon balance with the atmosphere from 1600-1800. Utilization and land clearing caused a large pulse of forest carbon emissions during the 19th century, followed by regrowth and net forest carbon sequestration in the 20th century. Recent data and knowledge of the general behavior of forests after disturbance suggest that the rate of forest carbon sequestration is declining. A goal of an additional 100 to 200 Tg C/yr of forest carbon sequestration is achievable, but would require investment in inventory and monitoring, development of technology and practices, and assistance for land managers.