Validation of Satellite Rain Rate Estimation with Ground-Based Observing Systems

Abstract

This chapter illustrates an example of the use of ground-based observing systems as a validation tool for satellite-based observations and retrievals. Clearly, satellite observations offer the advantages of larger spatial coverage with respect to ground-based observations. On the other hand, ground-based observations are usually more accurate and represent more closely the situation near the surface. This chapter presents an experimental product for the estimation of convective precipitation rain rates from satellite observations and the validation framework used to test and to evaluate the achieved performances. The convective precipitation product, derived by an algorithm combining a precipitation product with a convection detection algorithm, is intended to be used by hydrologists for civil protection purposes. The described validation procedure, using a combination of weather radars and raingauges, was not only relevant to the overall product accuracy evaluation, but also represented a critical component for the development of the merging algorithm.

Appendix A Implementation Details on Data Projections

This appendix is included simply to clarify the methodology used for the inter-comparisons of the various products. In order to compare values derived at different resolutions, several procedures were implemented to remap products:

• from AMSU to SEVIRI grid;

• from radar to AMSU grid;

• from radar to SEVIRI grid;

• from gauges to SEVIRI grid.

In particular the SEVIRI grid was chosen as reference, since it provides a fixed grid at a convenient resolution.

1.1 A.1 The AMSU-to-SEVIRI Remapping Process

This process was designed to allow the comparison of current PC products with SEVIRI-derived rainfall products and with radar and rain gauge data on a fixed grid. However, it was also used to compute the background BTs for the PC algorithm (as described in Sect. 3.2). An example of the AMSU-SEVIRI remapping is shown in Fig. II.7.27.

Fig. II.7.27

Example of PC remapping onto SEVIRI grid

The basic concepts of the implementation are hereafter described.

Given the location of a SEVIRI pixel, the bounds of the corresponding AMSU pixel are found. Routine begins by determining if it lies within the bounds of the AMSU grid, and if so, then searches for the closest AMSU observation. Having found the nearest AMSU, the routine then looks to the left and right (on the same AMSU row) to find which of these two is closer to the SEVIRI, and then the closest and its neighbor to the left or right are chosen as two of the four AMSUs surrounding the SEVIRI pixel. A check is performed to make sure the selected pixel is not on the lateral edge of the AMSU array, so that it makes sense to look left or right. If the pixel is on the edge, then the algorithm simply uses the two points on the row nearest the edge. A weight is assigned to these two based on their distance from the SEVIRI pixel. The distances are determined by computing the angle between the SEVIRI position vector and either of the two chosen AMSU position vectors. This part of the algorithm, using vector algebra, is quite efficient.

Once the distances are calculated there are two different strategies to derive the BT on the SEVIRI grid. The first one makes simply use of the nearest neighbors, while the second interpolates (weighted average) among the four closest AMSU FOVs according to the following procedure: returning to the previously found nearest AMSU, the routine looks at the two AMSU points above and below, in adjacent rows, to find which of these two is closer to the SEVIRI point. The chosen row (above or below the row in which the closest AMSU lies) is then used, and the four points used for interpolation are the two found in the first row, and the two in the next-best row.

An interpolated estimated is made in each row ($W_{i}$) between the two chosen points on that row, using weights

$$\begin{array}{lll} W_{i1} & = & 1-s_{1}/(s_{1}+s_{2})\\ W_{i2} & = & 1-W_{i1} \end{array}$$

((3))

where s ₁ and s ₂ are the distances (in latitude degrees) between SEVIRI and the first and second closest AMSU points on that row. In the above case, if $s_{1}=0$ (i.e., an AMSU coincides with SEVIRI), then that AMSU has weight 1 and its neighbor has weight 0. A similar weighting scheme is then used vertically (W _j ) along the AMSU column containing the closest AMSU to arrive at the weights which are used to interpolate as many of the AMSU channels as are needed to the SEVIRI location.

$$\begin{array}{lll} W_{j1} & = & 1-d_{1}/(d_{1}+d_{2})\\ W_{j2} & = & 1-W_{j1} \end{array}$$

((4))

where d ₁ and d ₂ are distances of two AMSUs along the minimal column from SEVIRI.

The horizontal weights and the vertical weights are then used to average the AMSU BTs (or any other quantity defined on the AMSU grid) on the two rows (e ₁ and e ₂):

$$\begin{array}{lll} e_{1} & = & W_{i1}*{\textrm{BT}}(1,1)+W_{i2}*{\textrm{BT}}(2,1)\\ e_{2} & = & W_{i1}*{\textrm{BT}}(1,2)+W_{i2}*{\textrm{BT}}(2,2)\\ e & = & W_{j1}*e_{1}+W_{j2}*e_{2}\end{array}$$

((5))

to get the final estimate e of the BT.

The co-location scheme based on the nearest neighbors was used for the derivation of background BTs (as described in Section 3.5), while the second one (based on weighted averages) was used for validation purposes.

1.2 A.2 The RADAR-to-AMSU Remapping Process

Radar data were convolved to the AMSU footprint using the methods described in Bennartz (1999), Bennartz and Michelson (2003), and Bennartz et al. (2002). The methods used in this study were initially derived for Baltex radar Data Center (BRDC) composites but were adjusted to account for the radar composites provided by CNMCA. The convolution takes into account the actual spatial sensitivity of AMSU-A and AMSU-B as outlined in Bennartz (2000). A fixed Z–R relation of $Z=200R^{1.6}$ was used in this study. Due to missing information about the actual position of the radar in the composite imagery, a parallax correction could not be performed. Also, the radar data used in this study were not gauge adjusted.

1.3 A.3 The RADAR-to-SEVIRI Remapping Process

The observations available for surface rain intensity (SRI) were on a different scale (spatial resolution and projection) compared to satellite grids and since the variability of precipitation fields strongly depends on the scale at which the fields were considered a meaningful comparison was not trivial. The upscaling technique (fine to course resolution) used to remap radar data onto MSG grid, here described, is very simple but numerically effective. National mosaic of SRI generated by radars and MSG product were composed of two static grids, each radar cell was linked to the SEVIRI pixel which contains the center of radar pixel. Therefore, radar data were remapped onto geostationary grid through the mean value of SRI calculated on radar cells linked to each satellite grid (Fig. II.7.28). An example of the RADAR-SEVIRI remapping is shown in Figs. II.7.29 and II.7.30.

Fig. II.7.28

Radar-to-SEVIRI remapping scheme

Fig. II.7.29

Example of radar-derived RR on SEVIRI grid

Fig. II.7.30

Example of radar-derived precipitation classes on SEVIRI grid