*3.1. Coherence Analysis*

As mentioned above, DInSAR can only be effectively applied in areas where the differential interferometric phase is characterized by high coherence. For a short spatial baseline interferometric pair, where two images are coregistered with high accuracy, the temporal decorrelation is the main factor of the coherence degradation.

To investigate the impact of the temporal baseline on the quality of interferometric results in the AOI, differential interferograms with a temporal baseline of 6, 12, 18, and 24 days and the common master image (12 February 2019) were processed, and their coherences were estimated (Figures 5 and 6). All image pairs had a short perpendicular baseline, to avoid the influence of spatial decorrelation on coherence degradation. All four analyzed image pairs covered relatively dry periods, without important or extreme precipitation (Figure 2) or floods. The parameters of interferometric pairs are presented in Table 3.

**Figure 5.** (**a**) Coherence and (**b**) wrapped differential interferograms for selected Sentinel-1 SLC image pairs (Table 3).

As it can be seen in Figure 5, the AOI is characterized by low coherence values, even for minimal possible temporal separation between images (6 days). The vegetation dominated areas, as for AOI, are especially likely to lose their coherence within a very short period. Moreover, an important loss of coherence is observed with a temporal baseline increase.

Figure 6 shows the coherence histograms for selected interferometric pairs (Figure 5; Table 3). As the temporal baseline increases, the frequency of pixels with low coherence also increases, whereas the mean coherence decreases from 0.13 (6 days) to 0.08 (24 days).

**Figure 6.** Coherence histograms for selected Sentinel-1 SLC image pairs (Table 3). Arrows indicate the mean coherence value.

**Table 3.** Parameters of image pairs selected for interferometric coherence analysis. Bp is the perpendicular baseline; Bt is the temporal baseline.


Figure 7 shows the average images for the ascending and descending orbital pass and associated coherence histograms. For each orbital direction, the average coherence image was obtained by averaging the coherence of all image pairs processed. These interferometric pairs correspond to the 2018–2019 period and meet the established baseline thresholds. The histograms (Figure 7c) show that the study area is dominated by low coherence (≤0.2), due to the land cover type (different types of vegetation). The highest values of average coherence (>0.3) correspond to urban areas and bare soil, reaching up to 0.99.

The average coherence values for different land cover classes are presented in Figure 8. In this case, the average coherence per class (ACC) value was calculated for each processed image pair.

As can be seen in Figure 8, the point cloud of ACC values is separated into two groups. The group with the highest ACC values corresponds to urban areas and bare soil, reaching a value of 0.5. The rest of the land cover classes belong to a group with lower ACC values, ranging from 0.05 to 0.2. The grassland, agriculture, and tule vegetation classes have the largest ACC values of this group (up to 0.2), whereas the lowest ACC values (≤0.05) were obtained for the mangrove class.

**Figure 7.** Average coherence estimated for the 2018–2019 period using image pairs from the (**a**) ascending and (**b**) descending pass; (**c**) coherence histograms for each average coherence. Arrows indicate the mean coherence value.

**Figure 8.** Average coherence for different land cover classes, estimated for image pairs obtained from the descending orbital pass during the 2018–2019 period.

### *3.2. Flooded Areas and Interferometric Coherence*

As the AOI is recurrently affected by floods, their influence on coherence degradation was also investigated. The coherence of the two temporary closed short-baselines interferometric pairs (Table 4) is compared in Figure 9. The 14 November 2018–20 November 2018 interferometric pair spans the flood events caused by strong precipitations (Figure 2), whereas the 14 December 2018–20 December 2018 interferometric pair spans the period with relatively dry climatic conditions. Figure 9 shows the important coherence loss due to flood occurrence. Coherence degradation was observed even for urban (e.g., Villahermosa) and bare soil areas. The mean coherence for the image spanning the flood event was 0.08, while the image pair with relatively dry conditions has a mean coherence of 0.12.

**Table 4.** Parameters of interferometric pairs selected for the flood impact on the coherence degradation analysis. Bp is the perpendicular baseline; Bt is the temporal baseline.


As shown above, the floods had a significant negative impact on the interferometric product quality, degrading considerably the coherence, even in short temporal baseline pairs (Figure 9). Floods are recurrent in the AOI, so the areas repeatedly affected by floods are very challenging for DInSAR applications. To identify the recurrently flooded areas, analysis of Sentinel-1 GRD images was performed.

Figure 10 shows the intensity data (dB) from Sentinel-1 GRD images acquired before (Figure 10a,c) and during flood events (Figure 10b,d). Dark areas (low negative intensity) correspond to the areas covered by water.

Recurrently flooded areas obtained using the Sentinel-1 GRD images and the methodology described in Section 2.5 are shown in Figure 11. The recurrently flooded areas are located south-southeast of the city of Villahermosa, in the towns of Gaviota del Sur, Parrilla, and Huapinol. These regions have recently been reported as vulnerable to flooding. Large recurrently flooded areas are also observed northwest of Comalcalco and north of Paraíso, where the Dos Bocas refinery is located. The analyzed flood events of February 2018 and November 2019 had an affected area of 6.92 ha and 11.37 ha in 2018 and 2019, respectively.
