基于Google Earth Engine和NDVI时序差异指数的作物种植区提取

为提高农作物种植信息遥感监测的效率,扩展数据适用范围,本文提出了一种基于时间序列NDVI差异指数的作物种植区提取方法。随着海量遥感与云计算的发展,Google Earth Engine作为一个全球尺度地理空间分析云平台,弥补了单机计算耗时长的不足,为快速遥感分类带来了新机遇。基于Google Earth Engine平台,以河南省开封市杞县为研究区,以2019—2020年杞县地区多时相Sentinel-2影像为数据源,结合物候信息,根据不同作物在时间序列NDVI曲线上的差异构建NDVI时序差异指数,从而提取作物种植区,区分不同作物类型,并与其他方法进行了精度验证和对比。结果表明：① NDVI时序差异指数法以作物物候信息为基础,与GEE高性能的计算能力相结合,形成了作物种植信息快速提取框架,可以方便快捷地进行作物种植区提取,较本地处理具有明显优势;② 杞县冬小麦和大蒜种植区有明显的空间分异性,冬小麦种植区主要集中在研究区西北部以及南部的农村居民点周围,而杞县大蒜则由于产品流通需要,主要集中在研究区中部以及东北部,居民点较为密集,交通便利的城市周边;③ 与时间序列支持向量机法和最大似然法相比较, NDVI时序差异指数进行作物种植区提取的总体精度达到83.72%, Kappa系数为0.67,分别比最大似然法提高了10.02%和0.21,比支持向量机法提高了4.18%和0.09,表明该方法能更高效率,更高精度地提取作物种植信息,实现区域作物种植信息的高效准确监测。总体来看,该方法在一定程度上可拓展遥感数据在农业领域的应用范围,具有推广价值。     

Conflation and aggregation of spatial data improve predictive models for species with limited habitats: A case of the threatened yellow-billed cuckoo in Arizona,USA

Riparian vegetation provides important wildlife habitat in the southwestern United States, but limited distributions and spatial complexity often leads to inaccurate representation in maps used to guide conservation. We test the use of data conflation and aggregation on multiple vegetation/land-cover maps to improve the accuracy of habitat models for the threatened western yellow-billed cuckoo (Coccyzus americanus occidentalis). We used species observations (n = 479) from a state-wide survey to develop habitat models from 1) three vegetation/land-cover maps produced at different geographic scales ranging from state to national, and 2) new aggregate maps defined by the spatial agreement of cover types, which were defined as high (agreement = all data sets), moderate (agreement ≥ 2), and low (no agreement required). Model accuracies, predicted habitat locations, and total area of predicted habitat varied considerably, illustrating the effects of input data quality on habitat predictions and resulting potential impacts on conservation planning. Habitat models based on aggregated and conflated data were more accurate and had higher model sensitivity than original vegetation/land-cover, but this accuracy came at the cost of reduced geographic extent of predicted habitat. Using the highest performing models, we assessed cuckoo habitat preference and distribution in Arizona and found that major watersheds containing high-probably habitat are fragmented by a wide swath of low-probability habitat. Focus on riparian restoration in these areas could provide more breeding habitat for the threatened cuckoo, offset potential future habitat losses in adjacent watershed, and increase regional connectivity for other threatened vertebrates that also use riparian corridors.     

A new MODIS daffy cloud free snow cover mapping algorithm on the Tibetan Plateau

Because of similar reflective characteristics of snow and cloud, the weather status seriously affects snow monitoring using optical remote sensing data. Cloud amount analysis during 2010 to 2011 snow seasons shows that cloud cover is the major limitation for snow cover monitoring using MOD10A1 and MYD10A1. By use of MODIS daily snow cover products and AMSR-E snow wa- ter equivalent products （SWE）, several cloud elimination methods were integrated to produce a new daily cloud flee snow cover product, and information of snow depth from 85 climate stations in Tibetan Plateau area （TP） were used to validate the accuracy of the new composite snow cover product. The results indicate that snow classification accuracy of the new daily snow cover product reaches 91.7% when snow depth is over 3 cm. This suggests that the new daily snow cover mapping algorithm is suitable for monitoring snow cover dynamic changes in TP.     

Results of a CLM4 Land Surface Simulation over China Using a Multisource Integrated Land Cover Dataset

In this study, the high-accuracy multisource integrated Chinese land cover （MICLCover） dataset was used in version 4 of the Community Land Model （CLM4） to assess how the new land cover information affected land surface simulation over China. Compared to the default land cover dataset in CLM4, the MICL data indicated lower values for bare soil （14.6% reduction）, nee- dleleaf tree （3.6%）, and broadleaf tree （1.9%）; higher values for shrub cover （1.8% increase）, grassland （9.9%）, cropland （5.0%）, glaciers （0.5%）, lakes （1.6%）, and wetland （1.1%）; and unchanged for urban areas. Two comparative CLM4 simulations were conducted for the 33-yr period from 1972 to 2004, one using the MICL dataset and the other using the default dataset. The results revealed that the MICL data produced a 0.3% lower mean annual surface albedo over China than the original data. The largest contributor to the reduced value was semiarid regions （2.1% reduction）. The MICL-data albedo value agreed more closely with observations （MODIS broad- band black-sky albedo products） over arid and semiarid regions than for the original data to some extent. The simulated average sensible heat flux over China increased by only 0.1 W m 2 owing to the reduced values in arid and semiarid regions, as opposed to increases in humid and semihumid regions, while an increased latent heat flux of I W m-2 was reflected in almost identical changes over the whole region. In addition, the mean annual runoff simulated by CLM4 using MICL data decreased by 6.8 mm yr-1, primarily due to large simulated decreases in humid regions.     

基于MODIS的春、夏季中国近海气溶胶对于云参数的影响差异研究

通过分析中国近海MODIS数据中气溶胶参数与云参数的相互关系,讨论了该区域气溶胶的间接效应及其对于云的可能影响.结果表明,在中国近海,气溶胶具有明显的间接效应,而且由于气溶胶种类和水汽的季节变化,使得气溶胶的间接作用具有很强的时间变化特征.在夏季,由于人为气溶胶占主导,它作为有效的云凝结核,使得气溶胶光学厚度（AOT）分别与云凝结核数（CCN）有正相关、与云滴有效半径（CER）有负相关性、与云光学厚度（COT）也存在着正相关,气溶胶的间接效应明显;在春季,由于沙尘气溶胶盛行,同时沙尘并不是很好的云凝结核,使得气溶胶光学厚度（AOT）分别与云凝结核数（CCN）的正相关减弱、与云滴有效半径（CER）则由夏季的负相关变为正相关、与云光学厚度（COT）存在弱的负相关,气溶胶的间接效应不明显.     

Soil respiration mapped by exclusively use of MODIS data for forest landscapes of Saskatchewan,Canada

Soil respiration (R_s) is of great importance to the global carbon balance. Remote sensing of R_s is challenging because of (1) the lack of long-term R_s data for model development and (2) limited knowledge of using satellite-based products to estimate R_s. Using 8-years (2002–2009) of continuous R_s measurements with nonsteady-state automated chamber systems at a Canadian boreal black spruce stand (SK-OBS), we found that R_s was strongly correlated with the product of the normalized difference vegetation index (NDVI) and the nighttime land surface temperature (LSTn) derived from Moderate Resolution Imaging Spectroradiometer (MODIS) imagery. The coefficients of the linear regression equation of this correlation between R_s and NDVI × LSTn could be further calibrated using the MODIS leaf area index (LAI) product, resulting in an algorithm that is driven solely by remote sensing observations. Modeled R_s closely tracked the seasonal patterns of measured R_s and explained 74–92% of the variance in R_s with a root mean square error (RMSE) less than 1.0 g C/m²/d. Further validation of the model from SK-OBS site at another two independent sites (SK-OA and SK-OJP, old aspen and old jack pine, respectively) showed that the algorithm can produce good estimates of R_s with an overall R² of 0.78 (p < 0.001) for data of these two sites. Consequently, we mapped R_s of forest landscapes of Saskatchewan using entirely MODIS observations for 2003 and spatial and temporal patterns of R_s were well modeled. These results point to a strong relationship between the soil respiratory process and canopy photosynthesis as indicated from the greenness index (i.e., NDVI), thereby implying the potential of remote sensing data for detecting variations in R_s. A combination of both biological and environmental variables estimated from remote sensing in this analysis may be valuable in future investigations of spatial and temporal characteristics of R_s.     

An improved dark object method to retrieve 500 m-resolution AOT (Aerosol Optical Thickness) image from MODIS data: A case study in the Pearl River Delta area,China

This paper presents an improved Dark Dense Vegetation (DDV) method for retrieving 500 m-resolution aerosol optical depth (AOT) based on MOD04-C005 arithmetic with the Moderate Resolution Imaging Spectroradiometer (MODIS) from the National Aeronautics and Space Administration (NASA). The improvements include change of the movement pattern of retrieval window, selection of a more suitable aerosol type, and storage of the look-up table. The method is then applied to obtain the AOT over the Pearl River Delta region (PRD). By comparing the results with the co-temporal ground sunphotometer observations in 2010, the correlation coefficient is found to be 0.794 with RMSE 0.139 and their variations remain consistent. Contrasts between model values in 2008 and MODIS AOT products in the same date also reveal a high accuracy of the improved DDV method. We also performed sensitivity tests to analyze the impacts of several parameters on apparent reflectance at different bands, and the results show that apparent reflectance is much more sensitive to surface reflectance and AOT than to elevation.     

改善MODIS BRDF产品热点效应的方法研究

核驱动的Ross Thick-LiSparse Reciprocal(RTLSR)双向反射函数(BRDF)模型已广泛地应用于MODIS等星载传感器的业务化产品处理中。但是,对于多年MODIS二向反射产品历史数据,如何基于RTMLSR模型发展一种简单有效快速的方法,进行热点效应的校正是一个迫切需要解决的问题。本文提出了一种简单有效的方法,不需要对观测数据重新反演,直接在现有MODIS二向反射产品的基础上进行热点校正,方便用户对历史MODIS二向反射产品的使用。该方法应用POLDER-3/BRDF数据库和部分经MODIS业务化算法筛选的反射率数据进行验证,并与RTLSR模型和RTMLSR模型的结果进行比较,结果表明:(1)该方法比现有的MODIS业务化RTLSR算法,对热点反射率有明显改善,拟合相对误差平均降低了10.12%;(2)该方法相对于RTMLSR模型在热点反射率拟合效果上差别不大,相对误差相差2.10%;(3)该方法对热点和冷点归一化的植被指数(NDHD)的估算效果有一定程度的改善,相对于RTLSR模型降低了约4.99%,与RTMLSR模型的相对误差相差1.32%,该方法对直接应用现有MODIS BRDF产品,基于热点方向反射率反演植被结构参数(如植被聚集指数)的精度提高有现实应用价值。     

基于MODIS温度的青藏高原多年冻土活动层厚度变化研究

基于MODIS温度数据,采用TTOP模型和Stefan公式模拟了青藏高原地区的冻土分布并计算了活动层厚度,并与地面观测结果进行了对比。结果表明：2003—2019年青藏高原多年冻土面积为1.01×10⁶ km²;多年冻土活动层厚度区域平均值为1.79 m, 活动层厚度区域平均的变化率为3.67 cm/10a,且草甸地区的变化率明显大于草原地区,5100~5300 m高程带的活动层厚度变化速率最大。     

黄河宁夏段凌汛期气温变化特征

利用宁夏1991~2010年20 a凌汛期水文和气温序列资料,研究了宁夏凌汛期特征、凌汛灾害发生主要时段,黄河封河、开河集中期与日平均气温之间的关系。结果表明:宁夏凌汛特征为封开河时序相反,小流量、高水位行洪,受强冷空气影响较大;凌汛灾害主要发生在封河和开河期间;日平均气温正负转化日期与封、开河期有对应关系,宁夏封河集中期在12月20日至1月19日;开河集中日期在2月13~20日。在凌汛关键期结合MODIS遥感数据对凌汛灾害监测防范有一定指导意义。