通过训练样本采样处理改善小宗作物遥感识别精度

训练样本质量是决定农作物遥感识别精度的关键因素,虽然高空间分辨率卫星的发展有效地解决了农作物遥感识别过程中的混合像元问题,但是当区域内不同作物种植面积差异较大时,训练集中不同类别样本数量往往相差较大,这样的不均衡数据集影响分类器的训练,导致少数类别的识别精度不理想。为研究作物遥感识别过程中的不均衡样本问题,本文基于GF-2号卫星数据,首先挖掘了地物的光谱信息、纹理信息,用特征递归消除RFE (Recursive Feature Elimination)方法进行特征优选,然后从数据处理的角度采用了5种采样算法对不均衡训练集进行处理,最后使用采样后的均衡数据集训练分类器,对比数据采样前后决策树与Adaboost(Adaptive Boosting)两种分类器的识别结果,发现:(1)经过采样处理后两种分类算法明显提升了小宗作物的分类精度;(2)经过ADASYS (Adaptive synthetic sampling)采样处理后,分类器性能提升最多,决策树的Kappa系数提高了14.32%,Adaboost的Kappa系数提高了10.23%,达到最高值0.9336;(3)过采样的处理效果优于欠采样,过采样对分类器的性能提升更多。综上所述,选择合适的采样方法和分类方法是提高不均衡数据集遥感分类精度的有效途径。     

完全残差连接与多尺度特征融合遥感图像分割

遥感图像数据规模大,光照、遮挡等情况复杂,目标密集、尺度不一以及缺乏大量带标注图像用于训练深度网络等特点对遥感图像分割的完整性和正确性造成了更大的挑战。针对深度卷积网络中因多次卷积造成分辨率显著下降,像素类别预测精度降低的问题,本文在深度卷积编码—解码网络的基础上设计了一个采用完全残差连接和多尺度特征融合的端到端遥感图像分割模型。该模型具有两方面优点：首先,长距离和短距离的完全残差连接既简化了深层网络的训练,又为本层末端融入了原始输入信息,增强了特征融合。其次,不同尺度和方式的特征融合使网络能够提取丰富的上下文信息,应对目标尺度变化,提升分割性能。本文通过对ISPRS Vaihingen和Road Detection数据集做数据扩充并进行实验,分别从平均IOU、平均F1值两方面对模型进行评价。通过与目前先进的模型以及文献中的研究成果进行比较,结果表明本文所提模型优于对比模型,在两个数据集上的平均IOU分别达到了85%和84%,平均F1值分别达到了92%和93%,能够有效提高遥感图像目标分割的完整性和正确性。     

可降水份的误差分析

分析了由无线电探空数据计算可降水份中逼近误差和观测误差的影响 ,利用香港的无线电探空资料计算出逼近误差和观测误差的影响分别为 0 .5mm和 1 .2mm ,两者的综合影响为 1 .3mm。     

Sensitivity and uncertainty analysis of mesoscale model downscaled hydro‐meteorological variables for discharge prediction

Precipitation and Reference Evapotranspiration (ETo) are the most important variables for rainfall–runoff modelling. However, it is not always possible to get access to them from ground‐based measurements, particularly in ungauged catchments. This study explores the performance of rainfall and ETo data from the global European Centre for Medium Range Weather Forecasts (ECMWF) ERA interim reanalysis data for the discharge prediction. The Weather Research and Forecasting (WRF) mesoscale model coupled with the NOAH Land Surface Model is used for the retrieval of hydro‐meteorological variables by downscaling ECMWF datasets. The conceptual Probability Distribution Model (PDM) is chosen for this study for the discharge prediction. The input data and model parameter sensitivity analysis and uncertainty estimations are taken into account for the PDM calibration and prediction in the case study catchment in England following the Generalized Likelihood Uncertainty Estimation approach. The goodness of calibration and prediction uncertainty is judged on the basis of the p‐factor (observations bracketed by the prediction uncertainty) and the r‐factor (achievement of small uncertainty band). The overall analysis suggests that the uncertainty estimates using WRF downscaled ETo have slightly smaller p and r values (p= 0.65; r= 0.58) as compared to ground‐based observation datasets (p= 0.71; r= 0.65) during the validation and hence promising for discharge prediction. On the contrary, WRF precipitation has the worst performance, and further research is needed for its improvement (p= 0.04; r= 0.10). Copyright © 2013 John Wiley & Sons, Ltd.     

阳江国际探空试验的GPS、探空、微波辐射计水汽资料对比分析

GPS/MET水汽资料因其建设维护成本低、时间分辨率高、可全天候、无人、自动观测等优点,已经成为获取大气水汽资料新的重要手段。以阳江第八届国际探空比对试验资料为基础,对比分析了GPS/MET与4种探空仪和微波辐射计获取的水汽资料。探空与GPS/MET测量的水汽值有较强的相关性,相关系数在0.57~0.74之间,但探空水汽值比GPS/MET偏高0.71~1.47mm,方差在3.48~3.96之间。其中RS92探空仪测量的水汽变幅与GPS/MET基本相当,但国内探空仪水汽测量值的变幅明显高于GPS/MET,偏高近1.6~1.8倍。微波辐射计在天气较好时,测量精度较高,基本上以3个国产探空仪数据相当;但微波辐射计在雨天受罩子上的雨水影响,误差较大,与GPS/MET的测量值不相关。     

Validation of the “HY-2” altimeter wet tropospheric path delay correction based on radiosonde data

Wet tropospheric path delay （PD） is a highly variable term for the altimeter measurement of a sea surface height, caused by the refraction effect of atmospheric water vapor and cloud liquid water. In order to esti- mate PD values, the ＂HY-2＂ system includes a calibration microwave radiometer （CMR） operating at 18.7, 23.8 and 37 GHz. The PD data of the CMR were compared and validated by coincident radiosonde profiles from ten globally distributed radiosonde stations during October 2011 to August 2012. The temporal interval was 1 h. In order to avoid land contamination, different spatial intervals between these two data sets were tested. The empirical fit function of PD uncertainty and spatial interval was found and extrapolated to the ideal situation that the data of CMR and radiosonde were totally coincident. The stability of the brightness temperature of the CMR and its impact on the PD correction was also studied. Consequently, the uncertainty of the PD algorithm of the CMR was estimated to be 2.1 cm.     

An Analysis of the Discontinuity in Chinese Radiosonde Temperature Data Using Satellite Observation as a Reference

Reconciling upper-air temperature trends derived from radiosonde and satellite observations is a necessary step to confidently determine the global warming rate. This study examines the raw and homogenized radiosonde observations over China and compares them with layer-mean atmospheric temperatures derived from satellite microwave observations for the lower-troposphere(TLT), mid-troposphere(TMT), upper-troposphere(TUT), and lower-stratosphere(TLS) by three research groups. Comparisons are for averages over China, excluding the Tibetan Plateau, and at individual stations where metadata contain information on radiosonde instrument changes. It is found that major differences between the satellite and radiosonde observations are related to artificial systematic changes. The radiosonde system updates in the early 2000 s over China caused significant discontinuities and led the radiosonde temperature trends to exhibit less warming in the middle and upper troposphere and more cooling in the lower stratosphere than satellite temperatures. Homogenized radiosonde data have been further adjusted by using the shift-point adjustment approaches to match with satellite products for China averages. The obtained trends during 1979–2015 from the re-adjusted radiosonde observation are respectively 0.203 ± 0.066, 0.128 ± 0.044, 0.034 ± 0.039, and –0.329 ± 0.135 K decade^–1 for TLT, TMT, TUT, and TLS equivalents. Compared to satellite trends, the re-adjusted radiosonde trends are within 0.01 K decade^–1 for TMT and TUT, 0.054 K decade^–1 warmer for TLT, and 0.051 K decade^–1 cooler for TLS. The results suggest that the use of satellite data as a reference is helpful in identifying and removing inhomogeneities of radiosonde temperatures over China and reconciling their trends to satellite microwave observations. Future efforts are to homogenize radiosonde temperatures at individual stations over China by using similar approaches.     

Current Status of Global Ocean Reanalysis Datasets

With the development of the ocean satellite remote sensing technology, the reanalysis of past oceanic observations using modern data assimilation technique and the restructuring of the long-term and consistent gridded data products have made great progress. Such datasets provide us with the most primary research tools to identify the state and evolution of ocean, and understand the role of ocean in climate change and variability at different spatial-temporal scales. In this paper, the current research status in the global reanalysis datasets including some of international global ocean reanalysis projects and the corresponding reanalyzed products were systematically reviewed. In addition, the present status of the domestic research of ocean reanalysis datasets was briefly introduced. The validation of the reanalysis datasets and some quality problems represented by the reanalyzed products, and the Ocean Reanalyses Intercomparison Project were systematically reviewed. Moreover, the prospects of the studies of oceanic reanalysis in the future were also discussed in this paper.     

Exploring nationally and regionally defined models for large area population mapping

Interactions between humans, diseases, and the environment take place across a range of temporal and spatial scales, making accurate, contemporary data on human population distributions critical for a variety of disciplines. Methods for disaggregating census data to finer-scale, gridded population density estimates continue to be refined as computational power increases and more detailed census, input, and validation datasets become available. However, the availability of spatially detailed census data still varies widely by country. In this study, we develop quantitative guidelines for choosing regionally-parameterized census count disaggregation models over country-specific models. We examine underlying methodological considerations for improving gridded population datasets for countries with coarser scale census data by investigating regional versus country-specific models used to estimate density surfaces for redistributing census counts. Consideration is given to the spatial resolution of input census data using examples from East Africa and Southeast Asia. Results suggest that for many countries more accurate population maps can be produced by using regionally-parameterized models where more spatially refined data exists than that which is available for the focal country. This study highlights the advancement of statistical toolsets and considerations for underlying data used in generating widely used gridded population data.     

探空温度资料质量控制技术研究

探空资料质量控制是探空资料在数值模式同化应用取得较好效果的重要保障,随着模式分辨率不断提高,探空资料飘移信息越来越需要应用到同化算子。本文通过全球谱模式T639L60(简称T639)分析场和掩星探测系统COSMIC资料对2011年7月我国119个探空站温度资料进行分析,同时结合探空气球随高度飘移信息以及温度残余误差等因素,开展探空温度质量控制技术研究。研究结果表明：探空气球飘移距离随高度而增大,在300~100 hPa之间飘移速率较大;基于探空温度资料残余误差和飘移因素建立的双权重标准质量控制方法能有效保障资料质量,是否考虑温度残余误差和飘移信息对质量控制结果有明显影响。当考虑温度残余误差时,300 hPa以上离群资料量显著减少;考虑飘移信息时,高层飘移速率较大区间离群资料量明显减少;模式地形与测站地形的高度差异对质量控制结果有明显影响,当测站地形高度低于模式地形时,易出现离群点,且两种地形差异大小与离群点呈正相关,即地形差异越大,低层离群资料越多。