云是影响当前气候变化模拟的最大不确定性因素之一。云量和云高的观测仪器和方法 非常多。现有研究着眼于不同的时空范围和尺度，使用了不同的观测方法，其结果差异很 大。弄清观测方法对观测结果的影响是准确估计云量和云高变化特征的基础。近年来，我 国开始大规模布设自动化云观测仪器，极有必要对这些观测仪器的观测特点进行深入的讨 论。美国等国家已经在自动化云量和云底高度观测方面积累了长期的观测资料。因此本文 拟对比研究美国云观测方法及其对云长期变化特征的影响，为我国大规模开展自动化云量、云高观测和结果分析储备经验和知识。 首先，本文基于美国自动地面观测系统（ASOS，Automated Surface Observing System） 的云高仪观测，研究了美国大陆地区地面到 3600 米范围内，云频率、云底高度的区域差 异和不同尺度的时间变化特征。云频率可以理解为一种时间平均云量，其长期气候特征与 云量基本一致。结果发现，对于不同类型的云，云频率的空间分布、日变化和季节变化特 征是不同的，而云底高度的空间分布、日变化和季节变化则相对一致。美国大陆地区云底 高度呈西高东低的空间分布，但在西海岸沿线地区受海洋低云的影响，云底高度极低。云 底高度的日变化最低值出现在早上 9 点左右，最高值出现在傍晚左右。大体上云底高度在 夏季较高，冬季较低。2000-2015 年，美国大陆地区地面至 3600 米的云频率有着每年增加 0.28%（p<0.1）的显著趋势，云底高度有每年降低 4 米（p<0.1）的显著趋势，并伴随着降 水日数每年增加 0.14 天（p<0.1）的显著趋势。这种变化趋势在美国大陆东部的冬季更为 明显。地面至 3600 米的云频率和云底高度与近地面空气相对湿度及气温日较差的相关性 在整个美国大陆大陆地区不同季节均有较为一致的相关性。然而云频率和云底高度与近地 面空气温度的年际对应关系则呈现出明显的季节差异和区域差异，在夏季整个美国大陆地 区云频率与近地面空气温度呈负相关，但在冬季云频率与近地面空气温度在美国大陆东部 呈正相关。 其次，本文基于云高仪和云雷达的观测，对比检验了北美区域再分析资料（NARR， North American Regional Reanalysis）中云底高度和云频率的模拟结果。发现北美区域再分 析资料可以整体模拟出云频率和云底高度的气候平均特征，NARR 对积云云底高度模拟较 好。在美国大陆地区地面至 3600 米范围内，NARR 模拟云底高度比 ASOS 观测平均低 568 米，而 NARR 模拟的云频率比 ASOS 观测高 26%。但是由于模式分辨率的影响以及北美区 域再分析资料云底高度算法的固有偏差，使得模拟出的云底高度在一定程度上小于地基主动遥感仪器的观测。并且在加利福尼亚州海岸线地区，受海洋低云和内陆干旱地区高云造 成的云底高度突变的影响，容易出现与观测相差较大情况。 最后，本论文从仪器观测视域角度出发，探究了卫星被动观测与地基被动观测云量的 差别，以及地基主动观测的时间平均云量与地基被动观测的空间平均云量的差别。结果发 现 MODIS（MODerate Resolution Imaging Spectroradiometer）遥感反演云量与地基可见光 全天空成像仪观测云量的差异主要由于仪器灵敏度的不同以及卫星与地基不同的观测视 域造成的。并且由于观测视域造成的影响是与 MODIS 仪器天顶角的大小以及视域中是否 有小而厚的积云密切相关的。结果发现在地面至 7620 米内，地基主动遥感仪器的垂直范 围越大，观测到的云量与地基可见光全天空成像仪越为接近。除此之外研究结果表明美国 第一代云高仪 3600 米的仪器量程太小，全年平均低估云量高达 38%，不能充分研究云和 天气、气候的相互作用。

Cloud is one of the largest uncertainties in predictions of climate change. At present, there are many different instruments and methods for observing cloud amount and cloud height. The existing researches focused on different time-space ranges and scales, and used different observation methods, thus have shown different results. Understanding the effect of the different methods on the observed results is the basis for accurately estimating the characteristics of clouds. In recent years, China began large-scale deployment of automated cloud observation devices. It is necessary to study the features of these devices. The United States and some other countries have recorded tremendous amounts of cloud amount and cloud base height data from automated devices until now. Therefore, the study of comparisons between different methods and devices and their influences on the analyzing the long-term variation of cloud in the United States will give inspirations to the automation of cloud observing in China. First, the geographic and temporal variability characteristics of the surface-3600 m cloud frequency and cloud base height over the contiguous United States are described using information observed by ceilometers from the Automated Surface Observing System (ASOS). Cloud frequency can be interpreted as a time-averaging cloud amount that tends to have the same characteristics in long-term cloud climatology. The results show that the geographic distributions and seasonal and diurnal cycles of cloud frequencies for different cloud types have different patterns. However, the geographic distributions and seasonal and diurnal cycles of cloud base heights for different cloud types are consistent. Cloud base height is lower in the eastern and higher in the western contiguous United States except in the west coastal area, where cloud base heights are very low. The diurnal cycles of the cloud base height present a minimum in the morning around 9 a.m. and peak in the late afternoon or early evening. In general, cloud base heights are higher in summer than in winter. From 2000 to 2015, the cloud frequency in the contiguous United States showed a positive trend of 0.28% yr-1 while cloud base height showed a negative trend of -4m yr-1 (p<0.1) for the surface-3600 m level, accompanied with a positive trend of precipitation days (0.14 days yr-1, p<0.1). Moreover, the increase of cloud frequency and the decrease of cloud base height were most obvious in winter in the eastern part of the contiguous United States. The surface-3600 m cloud frequency and cloud base height are closely related to surface air relative humidity as well as diurnal temperature range, and the correlations are consistent in most of the stations over the contiguous United States in different seasons. However, the correlations between the surface-3600 m cloud frequency and cloud base height with surface air temperature are not consistent in different seasons in the contiguous United States. In summer, these correlations are negative throughout the contiguous United States. However, in winter, cloud frequency is positively related with the surface air temperature in the eastern of the contiguous United States. Second, cloud frequencies and cloud base heights in North American Regional Reanalysis (NARR) are evaluated using ceilometers and cloud radar observations. It is found that NARR can generally simulate the climatology features of cloud frequency and cloud base height. Besides, NARR simulate well of cloud base height of cumulus. On average, NARR simulated cloud base height is lower than ASOS for 568 m in the contiguous United States while NARR simulated cloud frequency is higher than ASOS for 26% in the surface-3600 m level. However, due to the influence of the model resolution and the inherent effect from the cloud base height algorithm in NARR, the simulated cloud base height is generally lower than observations from surface active remote sensing instruments. Moreover, in the California coastline area, there exist large discrepancies between NARR simulated and surface observed cloud base height because of the discontinuity of cloud base height in these area, i.e., low cloud bases from marine layer and high cloud bases from the inner arid area. Last, the discrepancies between the observed cloud-amounts by passive satellite romote sensor and passive surface instrument, and by passive surface instrument and active surface instrument are discussed, focused on the different Field of Views (FOVs) of the devices. It is found that the discrepancies between the Moderate Resolution Imaging Spectroradiometer (MODIS) and surface total sky imager (TSI) observed cloud amount are generally caused by different instrument sensitivities and FOVs. Moreover, the influence of different FOVs is determined by the MODIS sensor zenith angles and if there are small and vertical-developed cumuliform clouds in the FOVs. Within surface-7620 m level, the higher the vertical detect range of the surface active remote sensing devices, the higher the correlations between cloud amount detected by those surface active remote sensing devices and observed by a TSI. Besides, the results indicate that the vertical detect range (0-3600 m) of the first generation ceilometer (Model: CT12k) from ASOS is too small, underestimating cloud frequency for 38% in a year, therefore it is not adequate to study cloud-weather and cloud-climate interactions.