滑坡空间易发性统计模型的构建需要正样本(滑坡点)和负样本(非滑坡点)两类数据,但历史观测数据仅记录了正样本,而负样本的选取容易受到正样本污染,因为没有滑坡记录的地方也可能在过去或未来发生滑坡,从而导致模型的预测精度与稳定性受...滑坡空间易发性统计模型的构建需要正样本(滑坡点)和负样本(非滑坡点)两类数据,但历史观测数据仅记录了正样本,而负样本的选取容易受到正样本污染,因为没有滑坡记录的地方也可能在过去或未来发生滑坡,从而导致模型的预测精度与稳定性受到影响。针对此问题,将前期提出的半监督学习算法PBLC(positive and background learning with constraints)应用于滑坡空间易发性分析,探讨其解决负样本污染问题的有效性。本文以粤东地区为研究区,选择高程、坡度、坡向、剖面曲率、距离道路最短距离、距离断层线最短距离、距水系最短距离、年平均降雨量、归一化植被指数和地理坐标共11个影响因子作为环境变量。结果表明,与传统的人工神经网络模型相比,基于PBLC算法的预测概率取值范围更为合理,预测结果更加稳定,且预测精度随背景样本数量增加而提高;粤东地区的滑坡灾害高易发区集中于北部和西南区域,坡度和高程是影响该地区滑坡易发性的主要因子。结果表明,半监督学习算法PBLC可以有效解决滑坡统计建模过程负样本污染的问题,提高模型预测精度。展开更多
The Queershan composite granitic pluton is located in the north of the late Paleozoic Yidun arc collision-orogenic belt, eastern Tibetan Plateau. The main rock types are coarse-grained porphyritic alkalic-monzonite gr...The Queershan composite granitic pluton is located in the north of the late Paleozoic Yidun arc collision-orogenic belt, eastern Tibetan Plateau. The main rock types are coarse-grained porphyritic alkalic-monzonite granite with minor fine-grained porphyritic monzogranite and granodiorite distributed in the eastern and southwestern regions. Here we report their zircon U-Pb ages and geo chemical data. The intrusive contact relations indicate that granodiorite was formed earlier than the alkalic-monzonite granite （105.9±1.3 Ma） and monzogranite （102.6±1.1 Ma）. These suggest that the Queershan composite granitic pluton was formed through three-stage magmatic events. The alkalic-monzonite granite （105.9±1.3 Ma） and monzogranite （102.6±1.1 Ma） are characterized by high SiO2 （73.5%-77.7%）, KzO＋Na20 （6.9%-8.5%）, Ga/AI ratios （2.6-3.4） and low A1203 （11.8%-14.5%）, CaO （0.25%-1.5%）, MgO （0.18%-0.69%）, negative Ba, Sr and Eu anomalies, showing A-type granite affinities. The granodi- orite exhibits lower SiO2, P205 and K20＋Na20 contents, but higher A1203, CaO and MgO contents than alkalic-monzonite granite and monzogranite, showing I-type granite affinity. 176Hff177Hf ratios of the alkalic-monzonite granite and the monzogranite are 0.282692-0.282749 and 0.282685-0.282765, respectively, and with similar cHf（t） values （-0.56 to 1.43 and -0.87 to 1.90 respectively）. They also present similar T1-M2 model ages （1.04-1.22 and 1.07-1.2 Ga respectively）, indicating they may be sourced from a similar rock source, mostly like Kangding Complex. The homogeneity of the Hf isotopic compositions and the absence of the MMEs demonstrate that little depleted mantle materials have contributed to the source. We propose that the Mesoproterozoic crust materials of the Yangtze Craton exist beneath the Yidun arc terrane and support it was a dismembered part of the Yangtze Craton. The A-type granites of Queershan composite granitic pluton are most probably related to the clo- sure of the Bangong-Nujiang Tethys ocean.展开更多
Combining the latest Planck, Wilkinson Microwave Anisotropy Probe （WMAP）, and baryon acoustic oscillation （BAO） data, we exploit the recent cosmic microwave background （CMB） B-mode power spectra data released by...Combining the latest Planck, Wilkinson Microwave Anisotropy Probe （WMAP）, and baryon acoustic oscillation （BAO） data, we exploit the recent cosmic microwave background （CMB） B-mode power spectra data released by the BICEP2 collaboration to constrain the cosmological parameters of the ACDM model, especially the primordial power spectra parameters of the scalar and the tensor modes, ns, as, r, nt. We obtain constraints on the parameters for a lensed ACDM model using the Markov Chain -- ＋0,0307 ＋0,0061 ＋0,0105 Monte Carlo （MCMC） technique, the marginalized 68% bounds are r -0.1043 -0.0914, ns -0.9617-0.0061, as =-0.0175-0.0097, nt = 0.5198＋-0.4579. We find that a blue tilt for nt is favored slightly, but it is still well consistent with flat or even red tilt. Our r value is slightly smaller than the one obtained by the BICEP group, in that we permit nt as a free parameter without imposing the single-field slow roll inflation consistency relation. When we impose this relation, then r= 0.2130-0.06096 ＋0.0446. For most other parameters, the best fit values and measurement errors are not altered significantly by the introduction of the BICEP2 data.展开更多
文摘滑坡空间易发性统计模型的构建需要正样本(滑坡点)和负样本(非滑坡点)两类数据,但历史观测数据仅记录了正样本,而负样本的选取容易受到正样本污染,因为没有滑坡记录的地方也可能在过去或未来发生滑坡,从而导致模型的预测精度与稳定性受到影响。针对此问题,将前期提出的半监督学习算法PBLC(positive and background learning with constraints)应用于滑坡空间易发性分析,探讨其解决负样本污染问题的有效性。本文以粤东地区为研究区,选择高程、坡度、坡向、剖面曲率、距离道路最短距离、距离断层线最短距离、距水系最短距离、年平均降雨量、归一化植被指数和地理坐标共11个影响因子作为环境变量。结果表明,与传统的人工神经网络模型相比,基于PBLC算法的预测概率取值范围更为合理,预测结果更加稳定,且预测精度随背景样本数量增加而提高;粤东地区的滑坡灾害高易发区集中于北部和西南区域,坡度和高程是影响该地区滑坡易发性的主要因子。结果表明,半监督学习算法PBLC可以有效解决滑坡统计建模过程负样本污染的问题,提高模型预测精度。
基金supported by China Geological Survey(Grant No.1212011121270)
文摘The Queershan composite granitic pluton is located in the north of the late Paleozoic Yidun arc collision-orogenic belt, eastern Tibetan Plateau. The main rock types are coarse-grained porphyritic alkalic-monzonite granite with minor fine-grained porphyritic monzogranite and granodiorite distributed in the eastern and southwestern regions. Here we report their zircon U-Pb ages and geo chemical data. The intrusive contact relations indicate that granodiorite was formed earlier than the alkalic-monzonite granite （105.9±1.3 Ma） and monzogranite （102.6±1.1 Ma）. These suggest that the Queershan composite granitic pluton was formed through three-stage magmatic events. The alkalic-monzonite granite （105.9±1.3 Ma） and monzogranite （102.6±1.1 Ma） are characterized by high SiO2 （73.5%-77.7%）, KzO＋Na20 （6.9%-8.5%）, Ga/AI ratios （2.6-3.4） and low A1203 （11.8%-14.5%）, CaO （0.25%-1.5%）, MgO （0.18%-0.69%）, negative Ba, Sr and Eu anomalies, showing A-type granite affinities. The granodi- orite exhibits lower SiO2, P205 and K20＋Na20 contents, but higher A1203, CaO and MgO contents than alkalic-monzonite granite and monzogranite, showing I-type granite affinity. 176Hff177Hf ratios of the alkalic-monzonite granite and the monzogranite are 0.282692-0.282749 and 0.282685-0.282765, respectively, and with similar cHf（t） values （-0.56 to 1.43 and -0.87 to 1.90 respectively）. They also present similar T1-M2 model ages （1.04-1.22 and 1.07-1.2 Ga respectively）, indicating they may be sourced from a similar rock source, mostly like Kangding Complex. The homogeneity of the Hf isotopic compositions and the absence of the MMEs demonstrate that little depleted mantle materials have contributed to the source. We propose that the Mesoproterozoic crust materials of the Yangtze Craton exist beneath the Yidun arc terrane and support it was a dismembered part of the Yangtze Craton. The A-type granites of Queershan composite granitic pluton are most probably related to the clo- sure of the Bangong-Nujiang Tethys ocean.
基金supported by the Chinese Academy of Science Strategic Priority Research Program "The Emergence of Cosmological Structures" (Grant No. XDB09000000)the National Natural Science Foundation of China (Grant Nos. 11103027, 11373030 and 11373031)the National High Technology Research and Development Program of China (Grant No. 2012AA121701)
文摘Combining the latest Planck, Wilkinson Microwave Anisotropy Probe （WMAP）, and baryon acoustic oscillation （BAO） data, we exploit the recent cosmic microwave background （CMB） B-mode power spectra data released by the BICEP2 collaboration to constrain the cosmological parameters of the ACDM model, especially the primordial power spectra parameters of the scalar and the tensor modes, ns, as, r, nt. We obtain constraints on the parameters for a lensed ACDM model using the Markov Chain -- ＋0,0307 ＋0,0061 ＋0,0105 Monte Carlo （MCMC） technique, the marginalized 68% bounds are r -0.1043 -0.0914, ns -0.9617-0.0061, as =-0.0175-0.0097, nt = 0.5198＋-0.4579. We find that a blue tilt for nt is favored slightly, but it is still well consistent with flat or even red tilt. Our r value is slightly smaller than the one obtained by the BICEP group, in that we permit nt as a free parameter without imposing the single-field slow roll inflation consistency relation. When we impose this relation, then r= 0.2130-0.06096 ＋0.0446. For most other parameters, the best fit values and measurement errors are not altered significantly by the introduction of the BICEP2 data.
