To simplify the calculations, a convenient swath coordinate system was employed in the HSCAT, in which the nadir track of the satellite is used as the center and the designated positions are denoted by a pair of along- and cross-track coordinates. As shown in Fig. 1, the grid in this swath coordinate system has a resolution of 25 km×25 km and each cell of the grid corresponds to a WVC. The coordinate system’s origin is set at the boundary of the HSCAT orbit and corresponds to a location at the southernmost latitude of the satellite’s orbit. Since the HSCAT has a fairly narrow swath, it is acceptable to neglect the longitudinal compression on the swath’s edge. This model can be regarded as a strictly rectangular network for simplification of the calculations.

The resampling algorithm maps the time-ordered sigma0s and the related parameters into the subtrack aligned grid of the WVCs. In this study, the egg-shaped sigma0s are aggregated into the swath grids. If the center of a sigma0 footprint falls within the WVC, the sigma0s and the related parameters are grouped into the WVC; the location of each sigma0 can be described in terms of a pair of along- and cross-track bin coordinate indices. Therefore, the sigma0 locations that are originally in longitude and latitude coordinates must be accurately transformed to along- and cross-track subtrack swath coordinates. To calculate the subtrack coordinates for the sigma0s, a “triangle marking” resampling method was applied. As shown in Fig. 2, Point P_i corresponds to the position of the footprint’s center of the ith pulse and Point S_j corresponds to the position of the jth nadir point, while Point S₁ corresponds to the point at the southernmost latitude of the satellite. A line that runs through Point P_i and is perpendicular to the line of the nadir track intersects that line at Point D. The three points form a right triangle with the Points P_iDS₁. The length of the side P_iD corresponds to the cross-track coordinates in the subtrack coordinate system for Point P_i and the length of the side S₁D corresponds to the along-track coordinates. Once we obtain the position of the Point D, the right triangle P_iDS₁ can be determined, and this can be used to calculate the subtrack coordinates of the sigma0. To obtain the location of the intersection Point D, we use the nadir point position from the L1B data, which provides the geodetic latitude and longitude of the location on the spacecraft’s nadir track at the specified time for the first pulse of the frame. The PRF is 181 Hz for the HSCAT and each frame of data contains 96 pulses; therefore, the time interval of each frame is approximately 0.54 s. Considering that the velocity of the HY-2A satellite is approximately 7 km/s, under normal conditions, we can obtain a nadir point position every 3.78 km along the nadir track from L1B data. The position of the nadir point with the shortest distance to Point D was chosen to represent the approximate position of Point D. Considering that the resolution of the WVCs is 25 km, this approximation is acceptable for the purpose of binning HSCAT sigma0s.

By calculating the spherical distance l from P_i to every nadir Point S when the spherical distance l reaches its minimum, the corresponding nadir point can be determined as the point with the shortest distance to Point D, which is Point S_j in Fig. 2. The cross-track coordinate for P_i can be determined by calculating the spherical distance l_{i, j} between point P_i and point S_j, while the along-track coordinate d_j for P_i can be determined by as follows:

where S_i corresponds to the spherical distance between the two neighborhood nadir Points S_i and S_i-1, which can be calculated by

where R corresponds to the radius of the earth; α_i and β_i correspond to the longitude and latitude for the ith nadir point; and α_i−1 and β_i−1 correspond to the longitude and latitude for the i–1 nadir point, respectively.

Once the along- and cross-track coordinates are computed, they can be rescaled to the desired swath coordinate system in the along- and cross-track directions to obtain a pair of along- and cross-track bin coordinate indices for P_i.

When S_i is smaller than 12.5 km, which is a half of the resolution of the WVC, it is reasonable to use the position of the nadir point with the shortest distance to Point D to represent the position of Point D. When S_i is close to the resolution of the WVC, this type of approximation will induce a large error. Considering that the HSCAT operates at 24 h a day in a 971 km orbit, it is easily affected by high-energy particles in the universe and thus the electronic components may break down intermittently. In order to ensure a continuous operation, the HSCAT sensor refreshes automatically every day when it passes a specific region. During the resetting interval, the HSCAT cannot effectively acquire measurement data, resulting in an approximate 2 s lack of data such as the nadir positional information. To ensure that the regrouping algorithm works well when there are missing data for the nadir positional information, a nadir point interpolation correction is required.

Since the time interval of a frame is approximately 0.54 s, we assume that the nadir point data are discontinuous and require interpolation if the time interval of the neighboring sub-satellite points t_i+1−t_i is greater than 1.08 s. A linear interpolation is applied first to obtaining the time for the interpolated nadir point, as shown as follows:

where t_p corresponds to the time for which the nadir point data require interpolation; and t_i and t_i+1 correspond to the time for the ith and i+1 nadir points, respectively.

A cubic spline interpolation is used to obtain the longitude and latitude for the interpolated nadir points, as shows in Eqs (4)–(9):

where α_pand β_p correspond to the longitude and latitude for the interpolated nadir point; α_i and β_i correspond to the longitude and latitude for the ith nadir point, which is the nearest nadir point to the interpolated nadir point in the forward direction; α_i+1 and β_i+1 correspond to the longitude and latitude for the i+1 nadir point, which is the nearest nadir point to the interpolated nadir point in the backward direction; t_p, t_i and t_i+1 correspond to the observation time for the interpolated nadir point, the ith point, and the i+1 nadir point, respectively; {{\alpha }^{″}}_{i+1}^{} and {{\alpha }^{″}}_{i+1}^{} correspond to the second derivative for the longitudes of the ith point and the i+1 nadir point, which can be calculated by the longitude of the 100 nearest nadir points in the forward and backward directions; and {{\beta }^{″}}_i^{} and {{\beta }^{″}}_{i+1}^{} correspond to the second derivatives for the latitude of the ith point and the i+1 nadir point, which can be calculated by the latitude of the 100 nearest nadir points in the forward and backward directions.

Compared with the data regrouping methods for the QuikSCAT and ERS scatterometers, the proposed “triangle marking” resampling method makes good use of the geodetic information of the nadir point from the L1B data. In addition, the iteration computation in the projection transformation between the earth-located (geodetic) coordinates and the subtrack coordinates is avoided and only simple calculations such as the spherical distance calculation between two point is required, facilitating an easy implementation of the proposed regrouping method for the routine operations. Moreover, by introducing an interpolation correction for the nadir point, the proposed regrouping algorithm operates well when nadir point positional information is missing, which enhances the robustness of the regrouping algorithm.