EFFECTS OF GF-6 SATELLITE RED-EDGE BANDS ON PEANUT DROUGHT MONITORING

: Red-edge band is an indicator band to describe the health status of crops. In order to explore the impact of the red-edge bands on the accuracy of agricultural drought monitoring, this study used GF-6 WFV and Landsat8 to calculate the Red-edge Normalized Difference Vegetation Index (NDVI705), Vogelmann red-edge index 1 (VOG1), Normalized Difference Vegetation Index (NDVI) and Land Surface Temperature (LST) to construct temperature vegetation drought index (TVDI), this index integrated vegetation index and land surface temperature information, the soil relative humidity data and the measured drought grade data were respectively used for correlation analysis and classification accuracy verification of the TVDI results. The results showed that dry edge equation fit with red edge bands were higher than that of non-red edge bands. The dry edge equation based on VOG1 index has the highest fit, with a maximum coefficient determination of 0.92; TVDI constructed by the above vegetation index have significant negative correlation with soil relative humidity. The TVDI based on the VOG1 index has a better correlation with the relative soil humidity, with a maximum correlation coefficient of 0.85; and realized the dynamic monitoring of peanut drought grade, the overall accuracy rate of peanut drought grade monitoring reached 92.59%, and the result of TVDI classification was in good agreement with the actual drought grade measurement result. The red edge bands of GF-6 satellite can effectively improve the accuracy of peanut drought monitoring and better characterize peanut drought information. This research provides a data reference for the application of GF-6 WFV data in agricultural drought monitoring.


INTRODUCTION
In the 21th century, the rapid development of satellite remote sensing technology and near-Earth remote sensing technology provided a new way for drought monitoring [1] . Modern remote sensing can affect the detection of plants through the spectral reflection information of red, green, blue, near-infrared, red edge and other bands received by sensors [2] . growth status was characterized. Studies have shown that when crops are subjected to drought stress [3] , the leaf water content [4] , leaf area index [5] , and chlorophyll content [6] all show a downward trend with the increase of the degree of stress, which can be used as important indicators for evaluating the degree of drought in farmland [7] , and the red-edge band (wavelength range 0.69～0.73 m) is sensitive to the changes of these indicators, and has broad application prospects in crop drought monitoring. Su Wei [8] constructed a regression analysis of multiple vegetation indices and canopy chlorophyll content based on UAV images, and found that the correlation between the vegetation index and chlorophyll content with the participation of the red edge band was high, and the determination coefficient was up to 0.702, the red edge band is sensitive to the change of the chlorophyll content of the maize canopy, and the addition of the red edge band can effectively improve the accuracy of chlorophyll content estimation. Vaz [9] used a portable plant reflectance spectrometer to measure the spectral reflectance of grape leaves and calculated the red edge parameters such as the normalized red edge vegetation index (Red-edge Normalized Difference Vegetation Index, NDVI705), red edge position, etc. The correlation between the relative water content of grape leaves and the red edge vegetation index under soil moisture treatment was analysed. arid variety. Lin Yi [10] used hyperspectral data to study the variation characteristics of the spectral reflectance of maize canopy under different drought stress conditions, and found that the red edge parameter responded quickly to drought stress and could be used as a reference for judging the degree of drought in farmland. The above research has laid a good foundation for the research of red-edge band in agricultural drought remote sensing monitoring. However, most of these studies use hyperspectral near-Earth remote sensing data or UAV remote sensing data, and the scope of application can only be carried out in small areas, and it is difficult to achieve continuous monitoring in large-scale areas.
With the successful launch of satellites with red-edge band such as WorldView-2, WorldView-3, Super Dove, Sentinel-2 and Gaofen-6, it is possible to use the red-edge band to monitor agricultural drought in a large area. Harry West constructed NDVI to invert soil moisture in the southwestern United States by replacing the red band of the Sentinel-2 data with the red band. The results showed that the vegetation index involved in the red band had a high inversion accuracy for soil moisture [11] , and higher than the red band. As China's first satellite with a red-edge band sensor, the domestically produced Gaofen-6 satellite has realized the replacement of similar foreign data by domestically produced high-resolution satellite data, breaking the long-term dependence on foreign satellites for medium-resolution and highresolution data in agricultural remote sensing monitoring. Gaofen-6 has the characteristics of wide coverage, high revisit, high resolution, and multi-spectral characteristics [12] , and it satisfies the application requirements of regional-scale agricultural remote sensing monitoring to a higher degree. There have been some applications in agricultural monitoring such as crop growth diagnosis [13] , orchard [14] , and the newly added rededge band has improved the monitoring accuracy [15,16] . However, there are few applied studies using the red-edge band of GF-6 satellite data for agricultural drought. Therefore, in this study, Sheqi County, Henan Province was talked as an example, based on GaoFen-6 Wide Field of View (GF-6 WFV) and Landsat8 data, the vegetation index was constructed using the red edge band to determine the relationship between temperature and vegetation drought. Improve the Temperature Vegetation Drought Index (TVDI), carried out remote sensing monitoring research on peanut drought, explore the advantages of GF-6 rededge band in peanut drought monitoring, and provide reference for the application of GF-6 satellite in agricultural drought monitoring. Sheqi County is affiliated to Henan Province, China, with a geographical location ranging from 32°47′N to 33°07′N and 112°4 5′E to 113°11′E, as shown in Figure 1. The climate of Sheqi County belongs to the transition zone from the north subtropical zone to the warm and humid zone, and has obvious continental monsoon climate characteristics. Peanut was the oil crop with the largest sown area in Sheqi County, second only to wheat. The peanut planting method in this study area was mainly based on direct seeding after wheat is harvested. Generally, it was sown in June and harvested in September. During the period from June 1 to September 9, 2019, Henan Province continued to experience high temperatures and little rainfall, and most areas experienced droughts of varying degrees. The spatial distribution of peanut planting areas in Sheqi County in 2019 was shown in Figure 1.

Remote Sensing Data
Based on the availability of data, the remote sensing data of GF-6 WFV and Landsat8 satellites on July 7, August 24, and September 9, 2019 were selected, and the cloud cover was less than 5%.  [17] , red Table 1 shows the calculation formulas of Rededge Normalized Difference Vegetation Index (NDVI705) [18] and Vogelmann Red Edge Index 1 (VOG1) [19] . The Landsat8 data comes from the official website of the United States Geological Survey (http: ∥ glovis.usgs.gov/), the image row number is 124/37, and the spatial resolution is 15 m. After resampling to 16 m, the single-window algorithm for the 10th band of Landsat8 TIRS was used to invert the Land Surface Temperature (LST) [20] .

Vegetation index Formula
Bands of GF-6 WFV NDVI Table 1 Vegetation index and its Calculation formula

Soil Relative Moisture Data
The soil relative humidity data comes from the China Meteorological Science Data Sharing Network (http: ∥ data.cma.cn) "China Meteorological Administration Land Surface Data Assimilation System (CLDAS-V2.0) Real-time Product Dataset", with a spatial resolution of 0.062 5 °×0.062 5°, which is in good agreement with the actual observation value on the ground [21] . This data is used to verify the accuracy of the inversion TVDI. The data selects the 0-20 cm soil relative humidity data in the time period corresponding to the satellite data acquisition period.

TVDI
The combination of land surface temperature and vegetation index can provide information on surface vegetation and water conditions [22] . When the vegetation coverage in the study area is from bare soil to full coverage, and the soil moisture is from extremely dry to extremely humid, the land surface temperature (LST) is the vertical coordinate, and the scatter plot with the Vegetation Index (VI) as the horizontal coordinate is a triangle [23] . TVDI comprehensively utilizes the information of surface temperature and vegetation index, and its calculation formula is as follows: Where, LST = surface temperature of any pixel; =the minimum surface temperature corresponding to the same VI value; =the maximum surface temperature corresponding to the same VI value.
The TVDI value range is [0, 1]. The larger the TVDI value, the lower the soil moisture and the more severe the drought; on the contrary, the smaller the TVDI value, the higher the soil moisture and the lighter the drought [24] . Using the TVDI index value as the drought grading index, the drought is divided into four grades, namely, 0～0.6 is no drought, 0.6 ～0.7 is mild drought, 0.7～0.8 is moderate drought, and 0.8～1.0 is severe drought [25] .

Drought Classification Accuracy
In order to analyse the ability of TVDI to reflect the drought degree of peanuts, referring to the accuracy rate detection method proposed by Liu Dan [26] , combined with the measured data of drought level, the accuracy rate of TVDI drought level monitoring results were evaluated. The calculation formula of the accuracy rate is as follows: (2) Where, ACC = the accuracy rate, H = the number of survey points where the TVDI monitoring result is consistent with the ground-measured drought level, M = the number of survey points where the TVDI monitoring result is inconsistent with the ground-level drought level.

LST-VI Feature Space
Using NDVI, NDVI705, VOG1 on July 7, 2019, August 24, 2019, and September 9, 2019, respectively, and the LST of the corresponding date to construct the Land Surface Temperature-Vegetation Index (LST-VI) feature space, and the results are shown in Figure 2. It can be seen from the figure that the scatter distribution shapes of NDVI, NDVI705, VOG1 and LST are all approximately triangular, which is in line with the characteristic space distribution proposed by Sandholt research area.  Table2. Equations of dry edge and wet edge of different vegetation index

Correlation Analysis Between TVDI And Soil Relative Humidity
Through the correlation analysis between 0-20 cm soil relative humidity and TVDI in the corresponding period, the effectiveness of TVDI in monitoring peanut drought was verified. The results are shown in Table 3. It can be seen from the table that when comparing the same period, the TVDI (TVDI_NDVI, TVDI_NDVI705, TVDI_VOG1) obtained by using the three vegetation indices NDVI, NDVI705 and VOG1 were all significantly negatively correlated with soil relative humidity (P<0.05), the soil relative humidity showed a downward trend, that is, the higher the TVDI, the lower the soil relative humidity, and the more severe the drought. The order of the correlation coefficients is TVDI_VOG1>TVDI_NDVI705>TVDI_NDVI. The TVDI constructed by adding the vegetation index of the red edge band has a higher correlation with the soil relative humidity and can better represent the drought information. When comparing different indices, the correlations between TVDI_VOG1 and soil relative moisture were the highest in the three periods, and the maximum correlation coefficient was 0.85, which were better than TVDI_NDVI and TVDI_NDVI705 in the same period, indicating that TVDI_VOG1 could better monitor drought information. When comparing different periods, the correlation between TVDI and soil relative humidity was higher on August 24, 2019 than on July 7, 2019 and September 9, 2019, because July 7 was the peanut seedling stage, and the ground cover The soil information affects the expression of the vegetation index; September 9 is the mature period of peanuts, and the peanut vegetation index has a saturation effect; and August 24 is the podding period of peanuts, with vigorous growth and high vegetation coverage, and soil in the vegetation information. The influence of background noise is small, and the correlation between TVDI and soil relative humidity is high.
Table3. Correlation coefficient between soil relative humidity and TVDI based on different vegetation index

Accuracy Verification
In order to reduce the influence of cloud and soil background, the three-phase TVDI_VOG1 was synthesized and classified into drought grades by the maximum value synthesis  Table 4. Accuracy of drought grade based on TVDI

5.CONCLUSIONS
The scatter distribution shapes of NDVI705, VOG1 and LST calculated by the red-edge band of the GF-6 WFV satellite are in line with the "triangular" characteristic spatial distribution. The dry edge equation of the VOG1 index has the highest fitting degree, and the maximum coefficient of determination is 0.92; the TVDI constructed by the three vegetation indices has a significant negative correlation with the soil relative humidity, and the TVDI constructed based on the VOG1 index has a better correlation with the soil relative humidity, the maximum correlation coefficient was 0.85; the dynamic monitoring of peanut drought grade was realized. The overall accuracy of peanut drought grade monitoring in the study area was 92.59%, and the TVDI classification results were in good agreement with the measured results of drought grades. The red-edge band of GF-6 satellite can effectively improve the accuracy of peanut drought monitoring and better characterize peanut drought information.