SAR and InSAR data linked to soil moisture changes on a temperate 1 raised peatland subjected to a wildfire 2

Alexis Hrysiewicz1,2,*, Eoghan P. Holohan1,2, Shane Donohue1,3 and Hugh Cushnan4 3 1SFI Research Center in Applied Geosciences (iCRAG), University College Dublin, Belfield, Dublin 4 4, Ireland 5 2 UCD School of Earth Sciences, University College Dublin, Belflied, Dublin 4, Ireland 6 3 UCD School of Civil Engineering, University College Dublin, Belflied, Dublin 4, Ireland 7 4 RPS Group, Elmwood House, 74 Boucher Road, Belfast, BT12 6RZ Northern Ireland, United 8 Kingdom 9 Corresponding author: Alexis Hrysiewicz (alexis.hrysiewicz@ucd.ie) 10


54
Peatlands are one of the largest carbon sinks on Earth: an estimated 20-30 % of global soil carbon is 55 to be stored in peat, despite fens and bogs covering only a small percentage of the world's land surface 56 (Drösler et al., 2008;Gorham, 1991 Ballynafagh bog occurs at the eastern limit of the range of raised bogs in Ireland (Cross, 1990)  where impacted, almost all surface vegetation on the high bog was removed (Figure 1c,d). It was 193 noted, however, that although the bog surface appeared completely scorched, the fire did not appear 194 to affect the peat below 3-5 cm depth. The period of the wildfire in 2019 (green bar in Figure 2) is 195 coincident with the summer peak of temperature and a period of rapidly increasing soils moisture 196 deficits as estimated at the regional Casement MET station. It is worth noting that the summer of 197 2019 was not the warmest or driest in the period from 2017-2021. Ostensibly, conditions may have 198 been more favourable for wildfires in 2018 and 2020, but ignition did not occur. 199 installed. The sensors were planted at 15 cm depth below the peat surface, following excavation of a 205 shallow hole, which was subsequently back filled with the excavated material. Data were logged by 206 using METER EM50 data loggers from 2017 to 2020, and subsequently using METER ZL6 data 207 loggers. The change in data logger type was made to reduce the risk of power loss, as occurred on a 208 few occasions in 2018 and 2019. The ZL6 data loggers are solar powered and, since their installation, 209 data has been continuously monitored every 30 minutes. In March 2019, a piezometer pinned to 1.5 210 m-depth (with 0.5 m screen) was installed at the Sub-marginal station, where peat thickness is about 211 8m. The piezometer comprises a 40 mm internal diameter PVC tube, the whole underground section 212 of which is screened, with a HYDROS 21 water level sensor connected to a METER ZL6 data logger. 213

Methods and Data
The piezometer continuously logged the water table depth until end of October 2020. 214 The post-fire inspection on the 19 th of July 2019 revealed that the two monitoring stations had escaped 215 any significant damage by the wildfire. The Central station was located at the southern extremity of 216 a c. 20 m by 10 m "island" of preserved or lightly damaged vegetation ( Figure 1c). The Sub-marginal 217 station was located within the domain of intact vegetation a few metres from the border of main burn 218 area ( Figure 1d). Consequently all sensors continued to function during and after the wildfire. 219

Multispectral data processing 220
To map fire-related vegetation changes at Ballynafagh, we used Sentinel-2 multispectral images at 221 L1C level (without atmospheric correction on radiance measurements) that were acquired before and 222 after the wildfire event. The multispectral bands were cropped and False-Colour, Normalised 223 Difference Vegetation Index (NDVI) and Infrared (IR) images were created (Red: 665 nm, Green: 224 560 nm, Blue: 490 nm; NDVI: 655 nm and 842 nm; IR: 842 nm, 665 nm, 560). Without changing the 225 coordinate reference system, the spatial resolution of the optical images is 10 metres. From the post-226 fire NDVI image, we extract the outlines of burnt areas by using segmentation with a minimum 227 threshold of NDVI = 0.2 and a maximum threshold of NDVI = 0.4. Only burnt areas with an area of 228 at least 25 pixels and non-burnt areas of a minimum of 5 pixels (respecting 4-connected pixels) are 229 selected. 230

SAR data processing 231
Synthetic Aperture Radar (SAR) data from the Sentinel-1 satellite mission were used to map changes 232 related to the wildfire also. In addition, InSAR processing of these data were used also to map surface 233 displacements at Ballynafagh bog for the 2017-2021 study period and to examine if the wildfire 234 affected the trends in and/or the quality of InSAR measurements. In the next paragraphs we give a 235 brief overview of the SAR data and InSAR processing to explain the nature, origin and significance 236 of the key parameters analysed later in the study. 237 During a SAR acquisition, the satellite emits radar waves of a given wavelength that reflect 238 The phase information within a single image is not usable because of spatial randomness of the pixel 247 phase, but the difference of phases within two SAR images of the same target area can be calculated 248 to obtain the changes in propagation time. In this case, the phase difference is directly linked to any 249 ground surface displacement that occurred between the two image acquisition dates, as well as to 250 other contributions from topography, satellite orbits, changes in atmospheric conditions, noise, etc. 251 The image obtained by phase differencing is called an interferogram. The stability (i.e. similarity) of the pixel phases between the two SAR acquisitions is termed the coherence (Zebker & Villasenor, 253 1992). Loss of coherence can be called decorrelation. 254 The InSAR method for calculating surface displacements consists of firstly accurately repositioning 255 the one image with respect to the other image (coregistration), and then subtracting or minimising the 256 contributions of all the other sources of phase variation, especially topography and atmosphere 257 (Massonnet & Feigl, 1998). The resultant image is termed a differential interferogram, and hence the 258 method is commonly termed D-InSAR. To obtain the time series of surface displacements -i.e., the 259 evolution of displacements for consecutive SAR acquisitions -an inversion can be done upon a 260 network of differential interferograms. The interferograms can computed either relative to one 261 reference image (single reference network) or relative to several reference dates (multi-reference 262 network) (e.g., Casu et al., 2006;Ferretti et al., 2001). The time elapsed between the acquisitions of 263 the SAR images used to generate each interferogram is termed the temporal baseline. A good network 264 design usually minimises the temporal baselines to maximise coherence (i.e. minimise temporal 265 decorrelation). 266 For this study, InSAR coherence and displacement estimations were derived by processing the 267 Sentinel-1 SLC images acquired, in IW mode, in the Ascending pass and in VV polarisation. All 264 268 available acquisitions between 4 th January 2017 and 18 th June 2021 (~4.5 years) were used: 130 from 269 S1A and 134 from S1B. The time interval between acquisitions was typically 6 days (250 images) or From our coregistered SLC stack, the conversion of radar phase to displacement was achieved by 276 using the GAMMA ® Interferometric Point Target Approach (IPTA) with a multi-reference network 277 of interferograms (Werner et al., 2003). This interferogram network included both single-look and 278 multi-look images. The latter are derived by a kernel-based image averaging of 10 pixels in range and 279 2 pixels in azimuth (i.e. a 10/2 multi-look factor) to increase signal-to-noise ratio at the lower spatial The target points were selected in the following steps and with the following criteria. Firstly, single-284 look points were selected based on phase stability (i.e., coherence) and amplitude deviation (from the 285 mean) (Werner et al., 2003). Secondly, these single-look points were merged with all multi-looked 286 point data inside the same data stack. Thirdly, the phase of the merged data stack is modelled for 287 unwrapping, under the assumption that the contributions from atmosphere and topography greatly 288 exceed those due to displacement, i.e.: 289 , where, (for each interferogram), !"# is the observed phase, ℎ is the SAR geometry constant, $ is 290 the perpendicular baseline of the interferogram and is the residual phase (interpreted as 291 atmosphere). Where a point displays a phase uncertainty value with respect to the modelled value of 292 greater than 1.3 radians, it is removed from the stack. This uncertainty threshold value is based on 293 trial and error; increasing this threshold, we can select more points/pixels but with lower confidence. 294 The key parameters and values used in the IPTA processing are listed in Table S1 (see Supplementary  295 material). 296 In parallel, coherence maps were computed by using a 10/2 multi-look (as for the IPTA processing) 297 and a 5 x 5 pixel estimation kernel (in radar geometry). Geocoding of images was done with a spatial 298 resolution compatible with the SAR resolution (~ 30 x 30 metres). To investigate the variation of 299 coherence around the two in-situ monitoring stations, we used the same estimation parameters, 300 regarding the multi-look kernel and kernel for estimating the coherence, and the coherence was 301 filtered by using a mean kernel of 3 x 3 pixels, centred on the pixels containing each station. Thus, 302 the coherence around each in-situ stations represents an average value for a range/azimuth area of 303 dimension ~ 150 x 150 m. 304

SAR backscatter intensity maps and wildfire duration 320
SAR backscatter intensity maps also enable delimitation of the wildfire and estimation of its duration. 321 Figure 4 shows the maps of mean SAR backscatter intensity in VV polarisation acquired over 322 Ballynafagh bog for the pre-fire and post-fire periods. On the pre-fire map (Figure 4a), the bog is 323 characterised by a relatively low SAR backscatter intensity (~ -11 dB) with low spatial variation (i.e., 324 -13 dB to -10 dB). On the post-fire map (Figure 4b area. This step-like increase in intensity is not observed in the non-burnt area (Figure 5c). In addition 344 the annual SAR intensity fluctuations could be higher for burnt areas compared to non-burnt areas 345 but the descripted time series of SAR intensity contain a single post-fire oscillations. Qualitatively, it 346 seems that the minimal peak of SAR intensity (in summer 2020) remains equal for the burnt and non-347 burnt areas. 348 The SAR backscatter intensity is also affected by annual oscillations of soil moisture and groundwater 349 level ( Figure 5). Soil moisture is highest -typically at saturation (or at sensor detection limit) -during 350 the winter and early spring months. Soil moisture decreases to its lowest values during the summer 351 months. Average groundwater level at the Sub-marginal station is 8 cm below the ground surface (see 352 Figure 5a). In winter, the groundwater levels reach up 4 cm below the ground surface, and declines 353 up to 32 cm in summer. Groundwater and soil moisture changes are positively correlated in time.  Table 1 gives the Pearson's correlation coefficient (r) between SAR intensity and soil moisture. These 355 parameters are well-correlated (r > 0.5) at the Sub-marginal station, but poorly correlated at the 356 Central station (r < 0.2). However, the intensity at Central station correlates well with soil moisture 357 measured at the Sub-marginal station (r=0.76) meaning that the poor correlation is likely a local effect 358 and caused by the temporal evolution of soil moisture at the Central station. 359 The corresponding time series of SAR backscatter intensity are given in Supplementary Materials for 361 the VH polarisation. Overall, the VH times series are noisier than the VV results and are also 362 correlated with soil moisture, but there is not increase in SAR intensity after the wildfire. The SAR 363 data in VH polarisation are therefore not affected by the wildfire. The InSAR velocity data indicate that during the observation period most of the high bog area, 379 straddling the Central, Sub-central, and Sub-marginal ecotopes, has undergone subsidence at average 380 vertical rates of up to -9 mm.yr -1 (LOS rates of -6.9 mm.yr -1 ). Several other areas within and just 381 outside the SAC boundary are apparently affected by uplift at average vertical rates of up to +5 mm.yr -382 1 (LOS rates of +3.8 mm.yr -1 ). These areas include a northern part of the high bog classified mainly 383 as Marginal ecotope, as well as zones of cut-over (i.e., harvested) bog to the west. The obtained 384 InSAR-derived velocities are thus dichotomous and somewhat heterogenous, but they overall display 385 a broad consistency in space across the bog. The temporal evolution InSAR-estimated surface displacement at Ballynafagh bog tracks the 396 temporal evolution of soil moisture and groundwater levels measured in-situ (Figure 7a-b). The 397 oscillations in the InSAR-derived displacements are near synchronous with both groundwater and 398 soil moisture variations. For the sub-marginal station, soil moisture data is positively and significantly 399 correlated (p_value < 0.001) with detrended InSAR displacement (see Table 2). The Pearson's 400 coefficient for linear regression (r), between soil moisture and InSAR-derived displacement is 0.69 401 for sensor 1 and 0.54 for sensor 2. For the Central station, the soil moisture is poorly correlated with 402 InSAR displacement immediately around that station. However, the soil moisture data at the Central 403 station are positively and significantly correlated with the InSAR displacement at the submarginal 404 station (see Table 2). Although the timescale of seasonal soil moisture and groundwater level 405 decreases is similar to the timescale of seasonal subsidence estimated from InSAR, the recovery of 406 soil moisture and groundwater to high levels may be much sharper for some oscillations (e.g., 2020) 407 -i.e. occurs over a much shorter timescale -than the seasonal upswing in surface displacement. 408 Finally, the magnitudes of changes in groundwater and ground surface levels are in ratio of roughly 409 10:1. 410  Figure 7a-b). The long-term LOS velocities appear to be lower at these points than those observed at the in-situ stations (-2.5±0.2(1 ) mm.yr -1 and -0.4±0.2(1 ) mm.yr -1 respectively), while the annual 414 oscillations are very similar (Figure 7). The variations in long-term velocity and in the magnitude of 415 annual oscillations further show that the InSAR-derived displacements are dichotomous and 416 heterogenous within the bog. However there is no shift or variation in the burnt area displacement 417 time series that is coincident with the wildfire. 418

Evolution of InSAR Coherence 419
Consistent with the InSAR-derived surface displacement evolution, there is not a systematic pattern 420 of spatial or temporal change in the coherence that one can relate to the wildfire. Figure 8 shows the 421 changes in coherence over Ballynafagh bog in the days before and after the wildfire. Overall, the 422 coherence on the bog is high to moderately high for the relatively short temporal baselines considered 423 here. The maps with lowest coherence are formed when one SAR image of the pair was acquired on 424 a rainy day -for example, the coherence maps spanning June 23 rd -July 5 th , June 23 rd -July11 th , 425 July 23 rd -August 4 th and July 29 th -August 4 th . Low coherence thus appears to be simultaneous with 426 differences in precipitation, in groundwater levels, and hence differences in soil moisture, between 427 the pair of SAR image acquisitions. Conversely, high coherence is associated with similar 428 precipitation and soil moisture conditions for the SAR acquisition pair. of the burnt area seems to become slightly higher for longer temporal baselines (>1 year) compared 448 to that of the non-burnt area, but it is unclear if this is a significant change. The seasonality of soil moisture changes creates an annual oscillation within InSAR coherence decay 459 in time on the bog. Figure 11 shows the relationship between observed coherence and temporal 460 baseline at the Sub-marginal station, when the first acquisition (reference image) is acquired in a 461 different season. Also plotted for each season is the probability of having a coherence higher than 0.5 462 for a given temporal baseline. The main trend seen in each graph is the well-known decrease in 463 annually. The oscillation of coherence is strongest when the reference image is acquired in the winter, 465 and it is weakest when the reference image is acquired in the summer. 466 The probability of high coherence (>0.5) is thus linked to the season in which the reference image is 467 acquired. For a reference image acquired in winter, high coherence can be found with temporal 468 baselines of up to three years. For a reference image acquired in summer, on the other hand, high 469 coherence is very unlikely be found with temporal baselines of more than one year. 470 471 Figure

472
The relationships between changes in soil moisture (and vegetation) and SAR intensity, InSAR phase, 473 coherence or closure phase have been well documented in previous works, (e.g., Barrett

Link between InSAR-derived displacements, soil moisture and groundwater level 492
In the absence of human interference, the peat-condition at raised bogs is controlled mainly by short-493 term seasonal and long-term climatic variations (temperature, rainfall and insolation), which control 494 evapotranspiration, soil moisture and water table levels (Heikurainen et al., 1964).  (Table 2). On the other 510 hand, the correlation between the same parameters is poor at the Central station. However, the soil 511 moisture data from the Central station show a moderate correlation (r = 0.35) with InSAR 512 displacement at the sub-marginal station and elsewhere on the bog. From visual inspection of the 513 Central station time-series (Figure 7b), it is apparent that correlation at the station itself is lost firstly in the second half of 2018 and secondly in the second half of 2020. These periods follow very dry 515 summers with drought conditions of up to several weeks in length as represented by: (1) the large 516 falls in soil moisture and groundwater levels locally at Ballynafagh bog (Figure 2a-b); (2) the periods 517 of high temperature and low rainfall regionally (Figure 2a) and (3) the high soil moisture deficits 518 regionally (Figure 2b). Furthermore, the relative recovery of soil moisture following the drought 519 periods is much greater and more rapid at the Central station than at the Sub-marginal station. 520 Consequently, we suggest that a rapid change in hydrogeological conditions around the Central 521 station following the end of the drought periods in 2018 and 2020 led to an underestimation of the 522 true ground displacement there, and hence a locally poor correlation between soil moisture and InSAR 523 displacement. This is because if rapid change in soil moisture is linked with large and rapid ground 524 displacement, then phase ambiguity may occur such that InSAR underestimates the true displacement 525 of the peat surface (Marshall et al., 2022;Tampuu, 2022). Additionally, as demonstrated here ( Figure  526 11), large soil moisture change can reduce InSAR coherence such that interferograms spanning the 527 period of rapid change are inaccurate. 528

The 2019 wildfire and implications for C-band radar penetration and backscattering at 529 raised peatlands 530
A striking result of our study if Ballynafagh bog is that the average SAR backscatter intensity 531 increases in a step-like manner after the 2019 wildfire (see Figure 4 and Figure 5). In contrast, the 532 InSAR coherence and displacement at Ballynafagh shows no clear effect from the 2019 wildfire. The 533 areas of increased SAR intensity after the wildfire correspond closely to the areas of reduced NDVI 534 on the bog (Figure 3), which we attribute to the removal of the mossy vegetation layer by wildfire 535 (Figure 1c-d). In support of this interpretation, we note that outside the SAC area containing the bog, 536 similar reductions of NDVI are seen also in fields within which grass or cereal crops were recently 537 harvested (Figure 3). Error! Reference source not found. shows a schematic interpretation to 538 explain these observations, and the general variation of SAR and InSAR data on the raised peatland, 539 in terms of the propagation and backscattering of the C-band radar beams. 540 541 Figure  likely to be so small, given the high level of groundwater and soil moisture in the wet periods (Nolan 544 and Fatland (2003). Thus most of the radar backscattering occurs at the peat soil surface. The 545 combination of peat and vegetation properties causes attenuation of the SAR backscatter intensity to 546 average values of -10 to -12 dB in VV (see Figure 5) and 9-10 dB in VH (Supplementary material). 547 During dry summer periods, the radar waves penetrate further into the upper few cm of the peat 548 because the groundwater levels and soil moisture are lower. It is difficult to give an absolute value of 549 the penetration depth into the peat, but given the generally high soil moisture content in peat (a 550 minimum of 0.5 at 15 cm depth during drought - Figure 5 and Figure 7), it is unlikely to be more than a few centimetres (Ayalew et al., 2007;Nolan & Fatland, 2003;Toca et al., 2022). The radar 552 backscatter intensity decreases because of the decreased dielectric permittivity related to the 553 decreased soil moisture content, especially during periods of drought and high temperature. In 554 addition the declining groundwater level leads to subsidence of the peat surface, which is seen as 555 displacement-related phase in InSAR. However, the change in soil moisture between winter and 556 summer periods also decreases InSAR coherence, making accurate detection of such ground 557 displacements more difficult. 558 After the wildfire, the backscatter intensity for the VV polarisation increases abruptly on average by 559 2-3 dB, as the attenuation related to 10-20 cm mossy vegetation layer is removed. The SAR 560 backscatter intensity in VH polarisation is not affected by the vegetation removal caused by the 561 wildfire (see Supplementary Materials). Moreover, after the vegetation is removed, the intensity of 562 VV polarised SAR backscatter is the same as the VH polarised backscatter (both average around 9-563 10 dB). Thus the mossy vegetation structure represents a partial polarised filter attenuating the 564 returning SAR waves in the vertical direction. 565 Finally, the InSAR phase is unaffected by the vegetation removal due to the wildfire, because the 566 main backscattering level is the peat soil surface. Moreover, if the soil moisture content does not 567 change much between image acquisitions and the severity of the burn is limited, the coherence 568 remains stable and shows no effect from the wildfire. 569

Transferability to other peatlands 570
Our findings are suitable for temperate raised peatlands and further studies should be focused on 571 temperate blanked bog (perhaps temperate fens if the InSAR coherence is enough to produce InSAR 572 time series of displacements). In terms of transferability of these results to other peatlands, we must 573 consider several issues: 574 1. the extent of raised bogs, as opposed to other peatland types such as blanket bogs and fens. 575 Raised bogs are relatively common in Ireland, making up about 20% of the total current 576 peatland area. In other European countries, however, largely destroyed by human activities; 577 2. different vegetation and hydrological dynamics at other types of bog. Blanket bog vegetation 578 is similar to that of raised bogs although the hydrological dynamics may differ. Nonetheless 579 InSAR has proven capable to mapping apparent displacement of temperate blanket bogs 580 (Alshammari et al., 2020;Marshall et al., 2022). Fens are characterised by different 581 hydrological dynamics and by more vascular vegetation than raised bogs, with the latter factor 582 making them much more difficult targets for InSAR in our experience; 583 3. condition of the bog -InSAR works well on relatively intact bogs; in our experience it does 584 not work so well on highly degraded bogs, afforested bogs, or bogs with bare peat. 585 4. the role of climate: InSAR has apparently worked well to detect displacement at tropical raised 586 bogs (Hoyt et al., 2020). Peatland in boreal or continental climates is likely to be more difficult 587 targets, in part due their being commonly afforested and partly because of long annual periods 588 of snow cover and/or ice which will decrease or destroy coherence and partly because there 589 are much larger changes in temperature and precipitation during the year. For such bogs, such 590 as shown by Tampuu (2022), the underestimation of InSAR-derived displacements during 591 summer is the main challenge for C-band InSAR. 592 For InSAR coherence, our conclusions on the link between soil moisture changes and coherence-593 related-moisture value should be transferable to other temperate bogs as we observe minor links 594 between coherence and vegetation types. However, (1) the dielectric constants, which control changes 595 in coherence, may be slight different through peatlands and (2) the temporal decorrelation may vary. 596 Further study could consist of an investigation of these in-situ constants to get a clear picture of the 597 variability of InSAR coherence. Regarding the effects of wildfire, a more intense fire than that at 598 Ballynafagh in 2019, whereby a significant depth of the peat layer is burned, will likely lead to a loss 599 of coherence as the radar properties of the materials could be modified. 600

Implications of soil moisture changes for InSAR computations on peatlands 601
Another important observation in our study is that the coherence on a raised peatland can increase 602 over time. This partly compensates for typical temporal decorrelation on longer temporal baselines 603 (> 1-2 years), and, to our knowledge, this is only observable on peat targets for these durations. The 604 coherence oscillates with an annual frequency with respect to the first coherence value (Figure 11). 605 Indeed, the coherence remains high several months after the reference acquisition (about 3 months), 606 decreases for durations of about 6 months, then increases 1 year after the first acquisition, and so on 607 ( Figure 11). Thus, it is possible to observe medium or high coherence for 1-or even 2-years temporal 608 baselines. 609 We can define, after simplifications, that: 610 with γ the InSAR coherence (Zhang et al., 2008). With a coherence of 0.7 on the 1-2-years 611 interferograms and equation 2, we can interpret that γ Temporal is also higher than 0.7, which 612 demonstrates that temporal decorrelation is extremely low on peatlands: probably the lowest 613 compared to other vegetation targets (Tampuu, 2022). On Figure 13 and for any observed coherence 614 value, InSAR coherence is a product of the three previous terms. Each coherence component tends to 615 decrease the observed coherence. The most visible trend (in red) is a decrease in coherence over time: 616 i.e., the coherence varies from 0 to 0.5 over 4.5 years. This trend is strictly the temporal decorrelation 617 and is characterised by an irreversible decrease. The second variation is shown by the blue line. This 618 decorrelation is characterised by oscillated decorrelation over time. However, it is not temporal decorrelation because the coherence can increase, if the soil moisture changes are low. Then, the last 620 component is the noise decorrelation represented by the purple segment. In our study case, we show 621 that soil-moisture-related coherence (γ Soil Moisture ) is the main factor controlling the recovery of 622 coherence on interferograms with long temporal baselines (>1-2 years), (see Figure 11 and Thirdly, InSAR coherence negatively correlates with changes in soil moisture. Consequently, InSAR 660 coherence is low for large soil moisture changes, and is high for small soil moisture changes between 661 two SAR acquisitions. Moreover, the designing of InSAR stack should take into account the 662 relationship to optimise the coherence of the InSAR stack, and avoid coherence loss due to sharp soil 663 moisture changes especially across dry periods. 664 Fourth, the wildfire highlighted how SAR and InSAR estimates relate to different attributes for raised 665 peatlands: (1) SAR intensity is affected by both changes in soil moisture and vegetation; (2) InSAR 666 coherence is affected by only soil moisture changes. Consequently, SAR and InSAR data from C-667 band radar sensor reveal information on different levels in the peat column.