基于RBF神经网络的遥感影像分类算法研究

提出了运用径向基函数神经网络和K均值法进行遥感影像分类的算法，以实际的遥感影像分类为例，通过与传统的最小距离法进行比较，对RBF神经网络分类器的优点进行了归纳，并就算法实施中的一些问题进行了探讨。分析结果表明，RBF神经网络是一种有效的图像分类器。     

GPS-measured land subsidence in Ojiya City, Niigata Prefecture, Japan

Land subsidence caused by compression of clay layers in Ojiya City, Japan was measured by global positioning system (GPS) between 1 April 1996 and 31 December 1998.

Three baselines were selected in and around the city, and height difference on a WGS-84 ellipsoid was measured by GPS on each baseline. The ground at the GPS station in the city subsides and rebounds 7 cm every winter and spring, respectively. Measurement accuracy was 9.5 mm standard deviation. Ground water level was observed at a well near the GPS station. Regression analysis between total strain, calculated as ratio of the height difference displacement to the total thickness of the clay layers, and the layers' effective stress change with ground water level change gave good correlation. The slope of regression line 7.0×10⁻¹¹ m²/N was obtained as an average apparent coefficient of volume compressibility of the layers.     

岩体工程质量分级应用研究

岩体质量评价采用的测试技术包括地质勘察、弹性波检测和室内岩石力学试验。结合水布垭水利枢纽大坝建基岩体工程质量评价实例,对岩体工程质量分级的应用进行了研究。     

黄河流域夏季分区面雨量预报研究

介绍黄河流域分区夏季面雨量预报的研究成果，精心挑选51个具有较好代表性的测站对黄河流域夏季降水的时空演变特征进行分析，使用K均值动态聚类对黄河流域的夏季降水进行了客观分区，并计算出各流域夏季面雨量。通过对黄河流域夏季雨量与500hPa环流，海温、OLR、中纬阻高，高原积雪，欧亚积雪等重要影响因子的关系分析，结合黄河流域夏季面雨量年降和年代际演变特征的分析，研究出黄河流域分区夏季面雨量预测的基本方法和模型，并通过客观化的数学方法建立黄河流域夏季面雨量预测系统，预测系统十年回报的结果显示出具有较的预测技巧。     

日照市短时冰雹定时-定点-定量预报

利用1960～2000年青岛、射阳、徐州3站08时探空资料及日照站08时地面气象资料，分析了日照地区出现的16次降雹天气的时空分布特征，在对降雹天气分型的基础上利用积云数值模式计算降雹因子，用降雹因子和单站要素因子建立历史降雹因子个例库，用历史实况资料建立实况个例库，用距离相似法实现日照地区短时冰雹定时、定点、定量的客观预报。     

黑龙江省湿地资源遥感信息解译分析

按黑龙江省湿地分类标准，应用资源卫星遥感技术，建立湿地资源解译标准，判别分析全省湿地资源现状与区域分布。     

陆面过程模式的改进及其检验

文中对陆面过程模式 (BATS)进行了改进 ,改进后的模式能较好地模拟地表物理量的年、季和日变化 ,它有两方面的特点 :采用热扩散方程模拟 7层土壤温度 ,模拟的温度可与实测值进行比较 ;在BATS的地表径流方案中 ,考虑了空间不均匀性的一般地表径流 (GVIC)过程 ,研究结果表明 :⑴模式能很好地模拟各层土壤温度的年、季和日变化。冬季土壤温度下层高于上层 ,而在夏季上层高于下层 ,这种上下层温度的转换时间大约在 4和 10月份 ,这与实测土壤温度的年变化非常一致。较为准确地模拟了各层土壤温度日变化的时滞效应。⑵用南京和武汉站的资料 ,将BATS地表径流方案模拟的地表水分分量与GVIC方案进行比较 ,BATS地表径流方案模拟的地表水分分量 ,与总水量的平衡相差较大 ,而GVIC模拟的效果相对较好 ,地表总水量基本上与降水总量达到了平衡     

Modelling of the OASIS Energy Flux Measurements Using Two Canopy Concepts

Two land surface schemes, SCAM and CSIRO9, were used to model the measured energy fluxes during the OASIS (Observations At Several Interacting Scales) field program. The measurements were taken at six sites along a 100 km rainfall gradient. Two types of simulations were conducted: (1) offline simulations forced with measured atmospheric input data at each of the six sites, and (2) regional simulations with the two land surface schemes coupled to the regional climate model DARLAM.The two land surface schemes employ two different canopy modelling concepts: in SCAM the vegetation is conceptually above the ground surface, while CSIRO9 employs the more commonly used `horizontally tiled' approach in which the vegetation cover is modelled by conceptually placing it beside bare ground. Both schemes utilize the same below-ground components (soil hydrological and thermal models) to reduce the comparison to canopy processes only. However, the ground heat flux, soil evaporation and evapotranspiration are parameterised by the two canopy treatments somewhat differently.Both canopy concepts reproduce the measured energy fluxes. SCAM has a slightly higher root-mean standard error in the model-measurement comparison for the ground heat flux. The mean surface radiative temperature simulated by SCAM is approximately 1K lower than in the CSIRO9 simulations. However, the soil and vegetation temperatures (which contribute to the radiative temperature) varied more in the CSIRO9 simulations. These larger variations are due to the absence of a representation of the aerodynamic interactions between vegetation and ground.