*4.3. Time and Memory Consumption*

The partition and parallel processing strategy can reduce the memory consumption of every single process. Additionally, increasing the number of parallel processing can reduce the time consumed by the whole data processing. The partition strategy can be roughly divided into three stages: image partition, MT-InSAR processing, and correction. The total time consumed by partition depends mainly on the processing times of image blocks, that is, the total number of blocks divided by the number of blocks processed in a single parallel session. The traditional processing spends all its time on MT-InSAR processing.

The program running time and memory consumption of the two strategies are listed in Table 3. The traditional method costs about 20 h. The total time required for block processing is 46.6 h when the number of parallel processing is 1. However, when the number of parallel processing is greater than 2, the block processing needs less time than the traditional processing. Additionally, it only needs 8.7 h when the number of parallel processing is 5. In addition, the computer memory occupied by block processing is much lower than that of traditional processing. In this experiment, the single memory occupied by partitioning is only 1/20 of that occupied by the traditional processing. When the image coverage is large or the computer memory is small, the traditional processing may cause memory overflows and the data cannot be processed successfully, but this problem will not happen to our data partition processing.

**Table 3.** Time and memory consumption of traditional processing and block processing with the number of parallel processing 5.


#### **5. Discussion**

#### *5.1. Space Consistency Correction*

The precision of the deformation in each block can be calculated by Formula (9). In the results of Changzhou, the unit weight mean error after adjustment is 0.13 mm/yr, the precision of the block with the highest adjustment precision is 0.27 mm/yr, and the precision of the block with the lowest adjustment precision is 0.59 mm/yr. In the results of Qijiang, the unit weight mean error after adjustment is 0.28 mm/yr, the precision of the block with the highest adjustment precision is 0.40 mm/yr, and the precision of the block with the lowest adjustment precision is 0.81 mm/yr. The adjustment precision is plotted in Figures 6 and 7, which show that the precision of the center blocks is higher than that of the edge blocks.

**Figure 6.** (**a**) Deformation precision of each block after adjustment in Changzhou. (**b**) The histogram of the adjustment precision of each block after adjustment, (**c**) the histogram of the difference of the mean value of the overlapping area before and after adjustment of each block.

**Figure 7.** The same as Figure 6, but for Qijiang.

In theory, after adjustment, the deformation rates of the homonymy points in the overlapping region should be the same. As Figure 6c shows, before the adjustment, the difference between homonymy points in Changzhou is more than 5 mm/yr, with an STD of 5.1 mm/yr, and after the adjustment, the difference almost converges to 0, with an STD of 0.5 mm/yr. In Qijiang, the difference between homonymy points is more than 7 mm before adjustment, with an STD of 3.9 mm/yr, and it is reduced to 0.6 mm/yr after the adjustment. The precision of the block processing results in the two study areas was greatly improved by adjustment, indicating that adjustment can improve the consistency of the block deformation results.

We selected 4 deformation areas (A, B, C, D) in the overlapping area in Changzhou (Figure A2), and compared their deformation results before and after adjustment in Figure 8. The results of the two image blocks in region A have little difference, so the improvement of the result is not significant after the adjustment (Figure 8b,f). However, the results of regions B, C and D are improved obviously after adjustment. The mean values of the differences of deformation in these three regions change from 4.1 mm/yr, −5.6 mm/yr, and 5.6 mm/yr to 0.3 mm/yr, 0.1 mm/yr, and 0.2 mm/yr after correction, and the spatial consistency of the results improves more significantly.

**Figure 8.** The difference in the homonymy points before and after adjustment in regions A, B, C, and D of Changzhou. (**a**) shows the location of the four regions. (**b**–**e**) Are the difference of the overlapping areas before adjustment in regions A, B, C, and D, and (**f**–**i**) are the difference after adjustment. (**j**–**m**) Shows the statistical histograms of the four regions.
