Sentinel-2 time series of Switzerland

We processed Sentinel-2 image time series from 2017 to 2023 for Switzerland with the Software FORCE (Frantz 2019) on the basis of [Sentinel-2 images](https://envidat.ch//metadata/sentinel-2-imagery-of-switzerland). The respective parameter files can be found here: [Github](https://github.com/TLKoch/Sentinel-2_CH). All the image time series consist of several TB and therefore access will be granted upon request.
The available bands (in spatial reference system EPSG 3035) are the following:
Red, Green, Blue, NIR, Red-Edge-1, Red-Edge-2, Red-Edge-3, SWIR-1, SWIR-2
The available indices (in spatial reference system EPSG 3035) are the following:
CCI, CIRE, NDWI/NDMI, NDVI, EVI
Processing
On the basis of processed Sentinel-2 images for the 14 Sentinel-2 tiles covering Switzerland (T31TGN, T32TLT, T32TMT, T32UMU, T32TNT, T32TPT, T31TGM, T32TLS, T32TMS, T32TNS, T32TPS, T32TLR, T32TMR, T32TNR), we processed the image time series further with FORCE v. 3.7.8-12.
We generated interpolated Sentinel-2 time series with a 5-day interval, corresponding to the theoretical revisit time of the Sentinel-2 satellites. It's important to note that the 5-day time series consist of interpolated and smoothed composites, not the original images. We used the radial basis convolutional filtering (RBF) available in the FORCE time series analysis (TSA) submodule (Schwieder et al. 2016). The RBF is similar to a spatial moving window average approach over time (Schwieder et al. 2016). We applied kernel width values of 10, 20, 30, and 50 days. We spectrally adjusted all the images to match Sentinel-2A, and we removed curve outliers and pixels that failed the quality checks for clouds and their shadows, snow, saturation, and limited illumination. The processed image time series are available in tiles of 30 by 30 km.
Example images
Uploaded is an example of the index EVI for one of the generated 30 by 30 km tiles located around the city of Zürich. The values are multiplied by 10.000. The time series spans the month of July from 2018.

我们基于[Sentinel-2影像](https://envidat.ch//metadata/sentinel-2-imagery-of-switzerland)，使用软件FORCE（Frantz 2019）处理了瑞士2017至2023年的哨兵-2 (Sentinel-2)影像时间序列。相关参数文件可在此处获取：[GitHub](https://github.com/TLKoch/Sentinel-2_CH)。所有影像时间序列数据量达数TB，因此需通过申请获取访问权限。

采用空间参考系EPSG 3035的可用波段如下：红光波段、绿光波段、蓝光波段、近红外波段 (NIR)、红边1波段、红边2波段、红边3波段、短波红外1波段 (SWIR-1)、短波红外2波段 (SWIR-2)。

采用空间参考系EPSG 3035的可用指数如下：叶绿素含量指数 (CCI)、叶绿素反射指数 (CIRE)、归一化水指数/归一化植被水分指数 (NDWI/NDMI)、归一化植被指数 (NDVI)、增强型植被指数 (EVI)。

## 处理
针对覆盖瑞士的14幅哨兵-2 (Sentinel-2)分幅（T31TGN、T32TLT、T32TMT、T32UMU、T32TNT、T32TPT、T31TGM、T32TLS、T32TMS、T32TNS、T32TPS、T32TLR、T32TMR、T32TNR）的已处理哨兵-2影像，我们进一步使用FORCE v.3.7.8-12对影像时间序列进行处理。
我们生成了间隔为5天的插值哨兵-2影像时间序列，该间隔对应哨兵-2卫星的理论重访周期。需注意，此5天时间序列由插值与平滑合成影像构成，而非原始影像。我们采用了FORCE时间序列分析 (TSA)子模块中的径向基卷积滤波 (RBF)方法（Schwieder et al. 2016），该方法类似于基于时间的空间移动窗口平均法（Schwieder et al. 2016）。我们设置的核宽度参数为10、20、30和50天。我们对所有影像进行了光谱校正以匹配哨兵-2A (Sentinel-2A)，并移除了曲线异常值以及未通过云与云阴影、积雪、饱和度、光照不足质量检查的像素。处理后的影像时间序列以30 km×30 km的分幅形式提供。

## 示例影像
本次上传的示例为围绕苏黎世市生成的一幅30 km×30 km分幅的增强型植被指数 (EVI)数据，数值已乘以10.000，时间序列覆盖2018年7月。