CSTR:11738.11.ncdc.isnow.db2527.2022
10.12072/ncdc.I-SNOW.db0024.2021
<pre><code> <pre><code>
Northeast China is one of the Three Snow areas in China. The region is located in high latitude, with low solar altitude angle in winter, poor lighting conditions, and high forest coverage. At present, the normalized difference snow index (NDSI) provided by MODIS V6 snow product has an obvious problem of excessive cloud mask in Northeast China, especially in northeast forest areas, which seriously affects the application of this product in the study of temporal and spatial variation of snow. At the same time, MODIS V6 snow products adopt low NDSI shielding, resulting in snow leakage in forest areas. Taking various MODIS V6 data sets as the source data, using the methods of morning and afternoon star synthesis, decision tree classification and spatio-temporal adaptive cloud removal respectively, the NDSI daily cloudless products with high precision and spatio-temporal continuity in Northeast China during the snow season from 2000 to 2020 were produced. This data set can be used for snow area mapping and snow phenology in Northeast China.
<pre><code> <pre><code>
In MODIS V6 snow products, raw in mod10a1 / myd10a1 products is used_ NDSI、NDSI_ snow_ Cover dataset and sur in mod09ga products_ refl_ B04 band dataset. The above product data comes from Google Earth engine（ https://code.eartengine.google.com )
<pre><code> <pre><code>
The first step is image clipping. Firstly, the snow product data set in the study area is cut on Google Earth engine.
The second step is star synthesis in the morning and afternoon to recover the NDSI values of some pixels. The algorithm rule is that if the original product is defined with cloud in the morning and no cloud in the afternoon, the NDSI value of the afternoon star will be assigned to the current day; If the stars are cloudy in the morning and afternoon, it is not restored, and the pixel is still retained as a cloud pixel; In other cases, the NDSI value of morning star is used.
The third step is to use the decision tree discrimination method. First, raw_ NDSI performs mean filtering. The algorithm rule is that the values of 9 pixels around the pixel are directly averaged and assigned to the central pixel to obtain a new data layer mean_ NDSI。 If the cloud pixel in mod10a1 corresponds to mean_ If the NDSI mean is greater than 0.4, it is considered that the pixel is not a cloud pixel, and the raw of the pixel is restored_ NDSI value. When the NDSI of a pixel is between 0 and 0.4, if the surface reflectance in the green band is less than or equal to 0.4, the pixel is a non cloud pixel, and its raw NDSI value is restored; If the surface reflectance in the green band is greater than 0.4, the pixel is still marked as a cloud. When the NDSI is less than 0, it is still recognized as a cloud pixel (this step value is only used to judge the pixel still marked as a cloud after star synthesis in the morning and afternoon).
The fourth step is to synthesize the adjacent day. The algorithm rule is that if the target pixel is a non cloud pixel before and after the day, the NDSI values of the before and after the day are averaged and assigned to the day; If one of the two days is a non cloud pixel, the NDSI value of the non cloud pixel is directly assigned to the current day; If the front and back days are cloud pixels, it is necessary to find the pixels in the front and back days to continue discrimination, and the algorithm rules are the same as above; If there are also cloud pixels in the first two days and the second two days, the target pixel has not been recovered. The calculation ends and waits for the next cloud operation.
In the fifth step, the spatiotemporal adaptive cloud removal algorithm proposed by Chen is used to completely remove the cloud in the whole image layer. The main idea is to judge similar pixels, and the NDSI based on similar pixels gives corresponding weight to cloud removal in adaptive space-time.
<pre><code> <pre><code>
Quality control is carried out from each link of data production. MODIS V6 version is greatly improved compared with previous V5, and atmospheric correction is carried out for water, cloud and heavy aerosol. Therefore, the data source of this study is reliable.
From the original data to the calculation of the final product, each processing step ensures the accuracy of the data. In the morning and afternoon, star synthesis cloud removal is a common method for snow products, mainly considering the stability of the surface on the same day; The threshold in the decision tree is also selected based on literature or statistical features, and a large number of results after executing the decision tree are selected and compared with false color synthetic images. This step can effectively restore NDSI_ snow_ There are some problems of mistakenly dividing forest snow into clouds in cover data set. Finally, Chen's method is used for complete cloud removal, and its accuracy is proved by cloud hypothesis test. Finally, visual inspection was conducted on the daily data products in 2018. The results show that the NDSI distribution of the data set is reasonable and can be used as the basic data source for snow cover research in Northeast China.
| # | number | name | type |
| 1 | 41771373 | National Natural Science Foundation of China | |
| 2 | 2017FY100500 | Investigation on the characteristics and distribution of snow cover in China | Special project for the investigation of basic scientific and technological resources |
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
