Tropical Geography ›› 2020, Vol. 40 ›› Issue (1): 137-144.

### Moderate Resolution Imaging Spectroradiometer Near-Infrared Water Vapor Linear Regression Model Based on the Global Positioning System Data

Yang Jiao1, Shi Lan1(), Wang Qianwen2, He Qiquan1

1. 1. School of Geographical Science, Nanjing University of Information Science & Technology, Nanjing 210044, China;
2. Nanjing Guotu Information Industry Co. Ltd., Nanjing 210036, China
• Received:2019-03-28 Revised:2019-09-02 Online:2020-01-10 Published:2020-02-24
• Contact: Shi Lan E-mail:sl_nim@163.com

Abstract:

We studied the Global Positioning System (GPS) Precipitable Water Vapor (PWV) and moderate resolution imaging spectroradiometer (MODIS) PWV in Hong Kong (using the data from 2010) and hierarchically stacked the sounding data based on the MATLAB platform because a monthly model of MODIS PWV products corrected by the ground-based GPS data has not yet been established and the accuracy evaluation of the corrected MODIS PWV products has not yet been conducted. We relied on the sounding dry delay calculations performed using the vertical integration method and compared and analyzed three universal dry delay models (Saastamoinen, Hopfield, and Black). The results denote that the Mean Relative Error (MRE), Root Mean Square Error (RMSE), and Percent BIAS (PBIAS) values calculated using the Hopfield model are the smallest, and the R values are slightly higher when compared with those calculated using the remaining two models. Therefore, for our Hong Kong study, we selected the Hopfield model as the most appropriate to invert the zenith dry delay and to solve the zenith hydrostatic delay. Further, we calculated the zenith total delay using the high-precision software GAMIT. The difference between the zenith total delay and zenith dry delay is referred to as the zenith wet delay, which can be used to obtain the GPS PWV. By considering the water vapor calculated by sounding as the true value, the comparison of GPS PWV and sounding PWV denotes that the water vapor retrieved by ground-based GPS is close to the sounding PWV and can reveal the characteristics of high water vapor in summer and low water vapor in autumn and winter. By calculation, we determined the MRE, RMSE, PBIAS, and R of the two methods to be 6.96%, 2.79 mm, -2.37%, and 0.98, respectively, indicating that ground-based GPS inversion is feasible and that the changes in local water vapor can be used to revise the MODIS near-infrared water vapor products. The obviously underestimated water vapor value inverted by MODIS is considerably different from the sounding PWV and GPS PWV. By calculation, the MRE between MODIS PWV and GPS PWV is 21.12%, RMSE is 15.29 mm, PBIAS is 19.22%, and R is 0.63. The proposed model is constructed to revise the MODIS water vapor products monthly by fitting the linear relation between GPS PWV and MODIS PWV monthly. The results denote that the MODIS PWV and GPS PWV exhibit a certain linear fitting each month. The fitting degree in January, March, May, July, August, and December is slightly better than that in the remaining six months. The MODIS water vapor products obtained by ground-based GPS exhibit improved MRE, RMSE, and PBIAS every month, and their accuracy has significantly improved. Therefore, they can be reliably used for water-vapor monitoring in Hong Kong.

CLC Number:

• P228.4