Continuous MODIS land surface temperature dataset over the Eastern Mediterranean

A continuous dataset of Land Surface Temperature (LST) is vital for climatological and environmental studies. LST can be regarded as a combination of seasonal mean temperature (climatology) and daily anomaly, which is attributed mainly to the synoptic-scale atmospheric circulation (weather). To reproduce LST in cloudy pixels, time series (2002-2019) of cloud-free 1km MODIS Aqua LST images were generated and the pixel-based seasonality (climatology) was calculated using temporal Fourier analysis. To add the anomaly, we used the NCEP Climate Forecast System Version 2 (CFSv2) model, which provides air surface temperature under both cloudy and clear sky conditions. The combination of the two sources of data enables the estimation of LST in cloudy pixels. The dataset consists of geo-located continuous LST (Day, Night and Daily) which calculates LST values of cloudy pixels. The spatial domain of the data is the Eastern Mediterranean, at the resolution of the MYD11A1 product (~1 Km). Data are stored in GeoTIFF format as signed 16-bit integers using a scale factor of 0.02, with one file per day, each defined by 4 dimensions (Night LST Cont., Day LST Cont., Daily Average LST Cont., QA). The QA band stores information about the presence of cloud in the original pixel. If in both original files, Day LST and Night LST there was NoData due to clouds, then the QA value is 0. QA value of 1 indicates NoData at original Day LST, 2 indicates NoData at Night LST and 3 indicates valid data at both, day and night. File names follow this naming convention: LST_ &lt;YYYY_MM_DD&gt; .tif, where &lt;YYYY &gt; represents the year, &lt;MM&gt; represents the month and &lt;DD&gt; represents the day. Files of each year (2002-2019) are compressed in a ZIP file. The file LSTcont_validation.tif contains the validation dataset in which the MAE, RMSE, and Pearson (<em>r</em>) of the validation with true LST are provided. Data are stored in GeoTIFF format as signed 32-bit floats, with the same spatial extent and resolution as the LSTcont dataset. These data are stored with one file containing three bands (MAE, RMSE, and Perarson_r).地表温度（Land Surface Temperature, LST）连续数据集是气候学与环境研究的关键基础数据。LST可拆解为季节平均温度（即气候态）与日距平两部分，其中日距平主要由天气尺度大气环流（即天气过程）驱动。为实现云像元内LST的重构，研究人员构建了2002-2019年时段的无云1km分辨率MODIS Aqua LST影像时间序列，并通过时间傅里叶分析计算了基于单像元的季节变化特征（即气候态）。为引入日距平分量，本研究采用了NCEP气候预报系统第2版（CFSv2）模型，该模型可输出阴天与晴空条件下的地表气温数据。两类数据源的融合实现了云像元LST的精准估算。本数据集包含经地理配准的连续LST产品（分为日间、夜间与日平均三类），可直接估算云像元的LST数值。该数据集的空间覆盖范围为东地中海区域，空间分辨率与MYD11A1产品一致（约1km）。数据以GeoTIFF格式存储，采用缩放因子0.02，以有符号16位整数格式存储；每日对应一个独立文件，每个文件包含4个波段：夜间连续LST、日间连续LST、日平均连续LST以及质量评估（Quality Assessment, QA）波段。QA波段存储了原始像元的云覆盖状态信息：若原始日间LST与夜间LST文件均因云覆盖出现无数据（NoData），则QA值为0；QA值为1代表原始日间LST存在无数据情况，值为2代表原始夜间LST存在无数据情况，值为3则表示日间与夜间的原始LST均为有效数据。文件命名遵循如下规则：LST_<YYYY_MM_DD>.tif，其中<YYYY>代表年份，<MM>代表月份，<DD>代表日期。2002-2019年的各年度数据均打包压缩为单个ZIP文件。LSTcont_validation.tif文件包含验证数据集，其中提供了与真实LST进行比对验证得到的平均绝对误差（Mean Absolute Error, MAE）、均方根误差（Root Mean Square Error, RMSE）以及皮尔逊相关系数（Pearson <em>r</em>）。验证数据以GeoTIFF格式存储，采用有符号32位浮点数格式，其空间范围与分辨率与LSTcont数据集完全一致。该验证数据集仅包含一个文件，内含3个波段：MAE、RMSE以及皮尔逊相关系数（Pearson <em>r</em>）。