基于温度植被干旱指数的江苏淮北地区农业旱情监测

为实现江苏省淮北地区农业旱情监测,利用Savitzky—Golay(S-G)滤波方法,对2011—2012年江苏省淮北地区1-5月MODIS的归一化植被指数(normalized difference vegetation index,NDVI)和地表温度(land Surface temperature,LST)8 d产品进行重构,去除原8 d数据的噪声,填补受云影响而缺失的数据。基于重建后的NDVI和LST数据,计算温度植被干旱指数(temperature vegetation dryness index,TVDI);分析TVDI和土壤湿度之间的关系,构建土壤湿度反演模型。最后,利用另外1组数据验证所建土壤湿度模型的精度。研究结果表明:1)S-G滤波方法能够提高MODIS LST和NDVI数据质量,并能对缺失数据进行填补;2)TVDI方法能够实现试验区土壤湿度反演,所建模型在试验区具有一定的普适性,反演精度较高(R2=0.575,RMSE=2.59%);3)TVDI方法在江苏省淮北地区干旱监测中得到了较好的应用,能够成功地监测出江苏淮北地区2011年和2012年春旱。该研究可为农业旱情的快速监测提供借鉴。

Agricultural drought monitoring in north Jiangsu by using temperature vegetation dryness index

Abstract: This paper focuses on developing an agricultural droughty monitoring method in north Jiangsu province based on the measurements from the Moderate Resolution Imaging Spectroradiometer (MODIS). In order to build soil moisture estimation model, we collected gravimetric water content of soil at experimental sites in 2011, measured the soil moisture of the sites in 2012, and downloaded the 8-day MODIS reflectance and land surface temperature data from January to May in 2011 and 2012 in this study region. The gravimetric water content of soil included soil moisture at 10 cm depth and at 20 cm depth. The used MODIS data have some noise from atmospheric effects, and some data can not be acquired because of cloud. Therefore, a Savitzky-Golay (S-G) filter method was selected to remove NDVI and LST noise, and generate lost NDVI and LST. Then, the Temperature-Vegetation Dryness Index (TVDI) was calculated from the re-created NDVI and LST data. A correlation analysis between TVDI and soil moisture at 10 cm and 20 cm depth were conducted. The results showed that TVDI was more correlative with soil moisture at 10 cm depth compared to at 20 cm depth, and that soil moisture at 10 cm depth was highly correlative with soil moisture at 20 cm depth. Based on the TVDI and soil moisture data at 10 cm depth, an empirical model for soil moisture estimation was built and validated. In addition, an empirical model was also built to describe the relationship between soil moisture at 10 cm and 20 cm depth. Finally, the two models was utilized to estimate soil moisture at 20 cm depth in the area from MODIS data, and the estimated soil moisture was used to monitor field droughty status with a criterion about wheat field draughty evaluation. The results show that S-G filter method removes the MODIS data noise, and can be used to generate the lost data. The correlation analysis between soil moisture and TVDI shows that TVDI has higher correlation with soil moisture at 10 cm depth, and a linear model can be used to best-fit the relationship between TVDI and the soil moisture at 10 cm depth. The correlation analysis between soil moisture at 10 cm depth and at 20 cm depth shows that soil moisture at 20 cm depth has higher correlation with soil moisture at 10 cm depth, and a linear model can be used to best-fit the relationship between soil moisture at 10 cm depth and at 20 cm depth. The validation experiments show that the model obtains a high accuracy of soil moisture estimation with an r2 of 0.575 and a RMSE of 2.59 %. Using this model, soil moisture maps at 10 cm depth were obtained. The linear model describing the relationship between soil moisture at 10 cm and 20 cm depth was used to obtain soil moisture maps at 20 cm depth. Wheat field draught maps in north Jiangsu Province were obtained by the criterion about wheat field draughty evaluation. Validation experiments showed that the experiments showed the droughty monitoring method was promising in monitoring the droughty, which appeared in north Jiangsu province.