Assessment of hydrologic connectivity in an ungauged wetland with InSAR observations

The Ciénaga Grande de Santa Marta (CGSM) is one of the world’s most productive tropical wetlands and one that has witnessed some of the greatest recorded dieback of mangroves. Human-driven loss of hydrologic connectivity by roads, artificial channels and water flow regulation appears to be the reason behind mangrove mortality in this ungauged wetland. In this study, we determined the CGSM’s current state of hydrologic connectivity by combining a remote sensing technique, termed as Wetland Interferometric Synthetic Aperture Radar (InSAR), with a hydrologic study of river water discharge. For this research, we processed 29 ALOS-PALSAR acquisitions taken during the period 2007–2011 and generated 66 interferograms that provide information on relative surface water level changes. We found that change in water discharge upstream on the main tributary of the CGSM could explain at most 17% of the variance of the change in water level in the CGSM. Fresh water inputs into the wetland were identified only when the mean daily water discharge in the river exceeded 700 m3 s−1, which corresponds to only 30% of the days during the period. The interferogram analysis also revealed that artificial channels within the wetland serve as barriers to water flow and contribute to the overall loss in hydrologic connectivity. We recommend increasing fresh water inputs from the Magdalena River by reducing water regulation of fresh water from the river and improving connectivity on either side of the artificial channels crossing the CGSM. This study emphasizes the potential of the application of wetland InSAR to determine hydrologic connectivity in wetlands that are completely or poorly ungauged and to define the necessary guidelines for wetland hydrologic restoration.


Introduction
Wetland hydrologic connectivity refers to the transfer of matter, energy, or organisms within or between wetland elements or their hydrologic fluxes (Pringle 2001, Foti et al 2012. The hydrologic fluxes include multiple water transport pathways such as surface run-off, groundwater, and shallow subsurface flow at various scales (Golden et al 2014, Cohen et al 2016. Assessments of hydrologic connectivity are necessary to understand water flow and hydrologic transport of sediments and other pollutants in wetland ecosystems. They also provide the knowledge on wetland functions required to develop the studies of environmental  (star). Yellow polygons depict the spatial coverage of the two scenes for which 29 acquisitions were collected (16 for Scene 1 and 13 for Scene 2). (b) Zoom (red in (a)) of the CGSM along with main channels of fresh water input from the Magdalena River (blue), and the two main roads (black) interrupting the flow of water between CGSM, river and sea. The road between Barranquilla and Santa Marta was initially constructed in the 1950's without adequate drainage systems, leading to the hydrologic isolation of the CGSM from the sea. The secondary road from Barranquilla to Guaimaro was constructed in the 1970's, restricting fresh water inputs from the Magdalena River. The Renegado channel is controlled at its mouth by a sluice gate for agricultural purposes. Mangrove coverage in 2016 is shown in green with dead mangrove areas that have not recovered to date represented in grey. Location of example transect of figure 2 is shown in purple.
impact, restoration and conservation (Freeman et al 2007, Quin et al 2015, Van Meter and Basu 2015, Thorslund et al 2017. Alteration in wetland connectivity is one of the major causes of wetland degradation globally (Jimenez et al 1985) and combined with macroclimatic changes, it can pose a highly detrimental threat to these ecosystems (Blanco et al 2006, Osland et al 2016. In the case of estuarine and coastal wetlands, such a combination has resulted in massive mortality episodes for mangroves (Barreto 2008, Cardona and Botero 1998, Cintron et al 1978. This is a cause for concern, since mangrove ecosystems are highly productive ecosystems, providing at least US $1.6 billion/year through ecosystem services and further support global coastal livelihoods (Polidoro et al 2010). The activities that are known to alter coastal wetland connectivity include flow regulation and consumption of water for agricultural purposes, road infrastructure, and urbanization, through reduction and even blockage of fresh water inputs into estuarine wetlands (Wemple et al 2017, Jaramillo and. The CGSM in Colombia presents an iconic case of alteration in wetland connectivity. This 1280 km 2 estuarine wetland complex forms the delta of the Magdalena River, the longest river in Colombia (figure 1) and is located between the two coastal cities of Barranquilla and Santa Marta. Over the last six decades, the wetland has experienced extended hydrologic modification through the construction of roads, artificial channels, and water flow regulation that have served to increase its vulnerability in relation to dry events of the El Niño-Southern Oscillation (Blanco et al 2006). These events augment surface water and pore water salinities resulting in prolonged hypersalinity conditions. Similar to many other areas in the world, hyper-salinity appears to be the main cause behind the CGSM's mangrove ecosystem's degradation and a significant controlling factor for mangrove development and rehabilitation (Botero and Salzwedel 1999, Cardona and Botero 1998, Rivera-Monroy et al 2011. From a historical mangrove coverage of 500 km 2 , mangrove coverage in the CGSM had been reduced to only 226 km 2 in 1996 (INVEMAR 2016). Previous studies reveal that the exposure to high salinity levels affects early mangrove seedling establishment, growth, eco-physiological proficiency and can further lead to biochemical function impairment and mortality (Flowers andYeo 1986, Krauss et al 2008).
Since 1996 coverage has recovered to 400 km 2 , but has recently become stagnant (INVEMAR 2016). The recovery is attributed to a rehabilitation program implemented between 1994 and 1998 to reestablish the hydrologic connectivity of the wetland. The program consisted in dredging five pre-existing channels connecting the CGSM with the Magdalena River (Botero and Salzwedel 1999) and improving drainage under the two roads that had decreased hydrologic connectivity between the wetland, its main tributary and the sea (figure 1). The program estimated that a total water flow of 163 m 3 s −1 (i.e. 2% of the mean annual discharge on the Magdalena River at its outlet into the Caribbean Sea) was the supply of fresh water needed to restore the wetland's mangrove ecosystems (Botero and Salzwedel 1999). After 2006, continued heavy siltation of material and channel clogging have necessitated additional dredging of 2.6 × 10 6 m 3 (Martínez 2005). Mangrove recovery has failed in several areas, implying that fresh water inputs are still too low and hyper-salinity continues.
The main objective of this investigation is to assess and quantify the current state of hydrologic connectivity within the CGSM and between the CGSM and its main tributary river. This will be accomplished with the knowledge that the roads, channel siltation, water flow regulation and hydrologic modifications restrict fresh water inputs and overall connectivity. Thus, we aim to identify the channels that currently control fresh water inputs into CGSM and the hydrologic conditions that enable these inputs. We also aim to assess the way in which landscape features in CGSM such as artificial channels and roads may affect the hydrologic connectivity of the wetland and threaten its sustainability. Unfortunately, due to its remoteness and spatial extension, the wetland lacks a hydrologic monitoring network. Hence, we employ Wetland Interferometric Synthetic Aperture Radar (Wetland InSAR), a technology based on remote sensing, to detect large-scale water level changes in time and space. We combine the results obtained from wetland InSAR with a hydrologic analysis of water levels and water discharge in the main tributary river to study the hydrologic connectivity of the wetland. We further suggest general guidelines for the improvement of the CGSM's hydrologic connectivity.

Study area
The CGSM comprises four large water bodies: the Main Lagoon, Salamanca island mangrove forest, located close to the city of Barranquilla, and the Northern and Southern Pajarales complexes located between the Main Lagoon and the Magdalena River (figure 1). Mangrove mortality is mostly concentrated in the Northern Pajarales complex. The Pajarales complex receives water from the Magdalena River, located west of the complex, through a series of natural and artificial channels among which Renegado, Aguas Negras, El Salado, and New Clarin Channels are the most significant. Flow regulation for agriculture and heavy siltation of the natural channels that feed the CGSM from the Magdalena River appear to restrict fresh water inputs into the CGSM and hamper river-to-wetland connectivity.
The annual precipitation over the CGSM experiences a bimodal seasonality of two dry seasons (roughly from December to April and July to August) and two wet seasons (roughly May to June and September to November); further, it is higher in the wet South (1330 mm yr −1 ) in comparison to the dry North (870 mm yr −1 ). The opposite takes place with potential evapotranspiration (2000 mm yr −1 and 500 mm yr −1 , respectively) (Blanco et al 2006). The arrival and duration of the season depends on whether the year is under El Niño or La Niña conditions. Pore water salinities are found within the range of 0 to 134 PSU.

Wetland InSAR
Common InSAR employs differences in the phase path length of two satellite acquisitions taken from the same orbital position, at the same resolution, and over the same region, to generate displacement maps representing the changes in the solid surface over time (Wdowinski et al 2004). These maps are referred to as interferograms and include visualization of phase differences indicated by coloration, and at times, contain shaky contour lines known as 'fringes', which can be either continuous or discontinuous (Ferretti et al 2007). Instead, Wetland InSAR can generate displacement maps of the water surface that have been previously employed to detect water movements in wetlands (Hong et  In this study, we obtained 29 ALOS-PALSAR satellite images from the Alaska Satellite Facility web portal, taken between the years 2007 and 2011 and covering the two scenes in the area of the CGSM (figure 1). We generated 66 interferograms (25 for Scene 1 and 41 for Scene 2, respectively) through selection of pairs of images with less than 365 days of gap between them, since differential interferograms over wetlands have low coherence beyond these thresholds (see supplementary materials and figure S1 available at stacks.iop.org/ERL/13/024003/mmedia). Coherence constitutes a unitless measure of the correlation between the two interferometric acquisitions that ranges from 0 (low) to 1 (high). The higher the coherence of an interferogram, the more possible it is to recognize a fringe pattern for water level change. We studied each interferogram to identify continuous and discontinuous fringe patterns that represent changes in water level difference between the two acquisitions. While continuous fringe patterns represent wetland water flow, discontinuous fringes or sharp phase changes represent a lack of connectivity between areas on both sides of the discontinuity. We analyzed the most persistent fringe patterns across the interferograms obtained from the Wetland InSAR technology The difference in water level between the two satellite acquisitions (Δh = WL 1 -WL 2 ) can be approximated from the difference in phase of the two satellite paths (see equation (1) in supplementary materials). The illustrated diagram corresponds to a hypothetical water profile difference when the daily discharge in the Magdalena River during the first acquisition (Q 1 ) is lower than in the second (Q 2 ) and under similar tide conditions. and attempted to relate them to water change and wetland hydrodynamics.

Relating wetland water level changes and Magdalena river discharge
The InSAR-derived water level change observations are relative in time and space. Over time, the observed changes represent water level changes between two acquisition dates (figure 2). In space, the measurements are in relation to a point of reference (e.g. point A' in figure 2). As we set the reference point in a hypothetical stable area with small water level changes, the observed measurements reflect the difference in water level between the two acquisition dates. Under natural conditions, water levels in each acquisition date (WL1, WL2) should relate to water changes in Magdalena River discharge during that acquisition (Q 1 , Q 2 ) as the CGSM forms a part of the Magdalena River Delta. To determine if this is the case even after the large hydrologic modifications suffered by the CGSM, we calculated the difference in mean daily discharge between the two acquisition dates (ΔQ = Q 2 -Q 1 ) at the nearest discharge station on the Magdalena River, located 30 km upstream of the mouth of the Renegado Channel ( figure 1(a)). For the sake of the multiinterferogram hydrologic analysis, we constructed the interferograms in such a way that the phase value from the acquisition date with the highest daily discharge Q (of the two dates) was subtracted from that of the lowest discharge. i.e. ΔQ > 0. Under the assumption that changes in the Magdalena River reflect differences in water level between the two satellite acquisitions, the difference in water level between acquisitions under similar tide conditions (Δh = WL 1 −WL 2 ) is expected to be positive and decreasing in the downstream direction (figure 2), following the typical profile of a riverine water discharge into an estuary (see Cai et al 2016 andBolla Pittaluga et al 2015).
We selected transects for the two regions where the most persistent continuous fringes occurred (figure 3) and extracted the wrapped phase change (ΔΦ) for all pixels in the interferogram that were intersected by these transects. The wrapped phase changes are expressed in the range from 0 to 2 . We unwrapped the phase change, calculated the magnitude of the corresponding change in water level, and repeated this procedure for all coherent interferograms (see supplementary materials).

Fresh water inputs
We conducted a hydrologic analysis of the interferograms in order to detect the degree of connectivity  (c)). Two main regions of continuous fringes were identified: (i) along the wedged formed by the Marquez and New Clarín channels, signaling fresh water input through the New Clarín Channel and (ii) the mouth of the Renegado Channel in the Southern Pajarales. The shape of the two fringe patterns (figures 3(b) and (c)) and the fact that the phase change decreased in the downstream direction when ΔQ > 0 inferred that the origin of the phase shift and the corresponding water level change was linked to fresh water inputs from the Magdalena River. A phase shift cycle of more than 2 (full colour range) could be identified in four out of 25 interferograms in (i) and 10 out of 41 in (ii). Although not as persistent among interferograms, two other regions showed marked fringe patterns in Scene 1: the flooding of the eastern plains from the Magdalena River but not exceeding beyond the Marquez Channel (south of (i)) and the mouth of the Salado Channel (west of (ii)) ( figure 3(b)).
For the acquisition pairs of 2010/07/29 and 2010/10/29 (Scene 1), spatial unwrapping of the phase shift map along the line of sight of the satellite (figure 4(a-c)) and conversion to water height change ( figure 4(d)) revealed a water level difference profile that decreased along the transects T-T' and S-S'. The total accumulated water level difference (Δh acc ) in each transect was −18 and −20 cm, respectively. Nevertheless, different hydraulic processes appear to occur in these two regions based on the differences in curvature of the Δh profile along the two transects.
For the purpose of comparison, we also analyzed the interferogram with the lowest corresponding mean Q (mean Q = 557 m 3 s −1 and ΔQ = 33 m 3 s −1 ) indicative of 'dry conditions' (figure 5 and S2). Although the interferogram presented high coherence, it did not exhibit continuous fringes such as those observed in the wet interferogram. During the dry season, levels in the Magdalena River drop considerably, to the point that fresh water inputs from this source into the CGSM appear to be negligible. The phase changes along the two transects S-S' and T-T' were found to be much less variable and did not complete a full phase change (i.e. < 2 ). When the phase changes along the two transects S-S' and T-T' were unwrapped (figure 5), phase difference profiles along the transects fell below 6 cm and did not present a particular trend that would suggest a constant fresh water input. Hence, under these hydrologic conditions, there appeared to be no important fresh water input from the Magdalena River via New Clarín, Salado, and Renegado Channels into the CGSM. However, since the transect T-T' is closer to the sea than the transect S-S', the influence of tide on the water level should be greater in the former transect under these 'dry conditions'. Indeed, the maximum water level difference in the transect near the sea T-T' (6 cm) doubles in comparison to that of S-S' (3 cm) (figure 5(d)), suggesting stronger tidal influence near the coast. However, in some interferograms even under 'wet conditions', observed water level change along the transect S-S' increases contrary to expected from fresh water inputs from the Magdalena River (see figure S3 for an example case). In this case, the extremely high accumulated water level difference (Δh acc = 25 cm) can neither be explained through tide, since tidal daily water level differences as far as the transect S-S' have not been found to exceed 6 cm ( figure S4). Rather, this case forms evidence of additional anthropogenic factors that control or alter the natural incoming of fresh water by regulation. In the case of the Renegado channel, the sluice gate that controls the flow at the mouth of the river may provide such a signal, as water does not enter the CGSM when the discharge is high due to flow control.
Using all available interferograms from both scenes yielded a statistically significant (p < 0.05) emerging relationship between ΔQ and Δh acc along the transect S-S' (figure 6(a)). However, the ΔQ merely explained 17% of the variability observed in Δh acc , probably due to the additional influence of tide and flow regulation. For all interferograms with large ΔQ (i.e. ΔQ > 200 m 3 s −1 ), Δh acc was found to be negative, implying a profile of decreasing water level difference between the two acquisitions in the downstream direction (as in figures 2 and 4). Evidence of fresh water inputs along the Renegado Channel was further confirmed in the analysis of the maximum discharge of the two acquisitions for each interferogram (Q max ; figure 6(b)); fresh water inputs into CGSM mainly occurred when Q exceeded 700 m 3 s −1 . The cumulative probability distribution of Q at Calamar station reveals that 700 m 3 s −1 is reached or passed only on 30% of the days during the period 2007-2011 ( figure S5).
A similar analysis for transect T-T' at the mouth of the New Clarin Channel into CGSM displays no significant relationship between Δh acc and ΔQ or Q max (figure S6). Although this may evidence the hydrologic dysfunction of the channel due to sediment blockage, it may as well relate to the fact that only 11 coherent interferograms were available to determine the existence of such a relationship.  Relationship between water level change in the CGSM and water discharge in the Magdalena River. For the 32 interferograms that reveal coherence in the area of the transect S-S', (a) change in total water level difference along transect (Δh acc ) against difference in mean daily discharge between the two acquisitions (ΔQ). (b) Relationship between the absolute value of Δh acc and the higher of the two mean daily discharge values corresponding to the two acquisition days (Q max ).

Impairment of hydrologic connectivity within CGSM
Both interferograms spanning over wet and dry conditions demonstrate discontinuous fringe patterns in particular in the Northern sector of CGSM that may represent barriers to hydrologic connectivity within the CGSM. In 15 out of 25 interferograms, abrupt changes in water level difference occur at both sides of the New Clarín Channel and can be observed along 10 km, suggesting low water connectivity between Figure 7. Sediment barriers to hydrologic connectivity. Sediment accumulated along the banks of the artificial channels that act as a barrier to fresh water input into neighboring wetland bodies.
wetlands on both sides of this channel (figures 3(b), 4(a) and 5(a)). Similarly, judging through the long discontinuous fringe, the Marquez channel also isolates water bodies on both sides of its course, interrupting the flood plain sheet flow from the Magdalena River at least during dry conditions. Interestingly, the interrupting effect of these channels appears to be stronger than the isolating effect of the road that connects Barranquilla and Ciénaga since the road-related fringe discontinuity is not as frequent in the interferograms.

Discussion
The CGSM faces major siltation and hydrologic connectivity loss issues threatening its sustainability. To support its management and propose adequate solutions, it is of major importance to improve our understanding of the factors that lead to increased siltation and hypersaline conditions. The persistent fringes of water level difference along the New Clarín, Renegado, and Salado Channels suggest the provision of fresh water to the CGSM from the Magdalena River. Fresh water enters the CGSM, at least during the period of available satellite imagery, despite the construction of the road along the river's eastern bank, the flow control of the sluice gate at the entrance of the Renegado channel and the impairment of the channels due to heavy siltation.
It can be argued that channel-reopening operations have indeed allowed once again fresh water entry from the river. This may be related to the gradual recovery of mangrove within the wetland during the last two decades, as hypersalinity episodes are reduced due to availability of fresh water and salt dilution. Our results show that the Magdalena River supplies fresh water to the CGSM through the Renegado channel when discharge at Calamar station is above 700 m 3 s −1 . This conclusion is supported by observing accumulative phase change along the two transects: all the phase changes greater than 2 (> 15 cm of water height) occurred in interferograms in which at least one acquisition was taken when the Magdalena discharge at Calamar station was above 700 m 3 s −1 . However, based on the mediocre recovery of mangrove populations observed to date (INVEMAR 2016), 700 m 3 s −1 may be too high a quantity to allow fresh water inputs during dry conditions. A possible solution could be to introduce environmental restrictions to the sluice gate located at the mouth of the Renegado channel to guarantee minimum environmental flows.
In the case of the CGSM, wetland features that are supposed to enhance hydrologic connectivity (such as artificial channels) actually exhibit the opposite effect. In the New Clarín Channel, dredging operations that remove sediment deposits from the channel bed and settle them along the channels' banks actually isolate the water passing along the channel from its surroundings. This practice has been recurrent during the rehabilitation program of the CGSM (figure 7). The isolating effect is even more noticeable during dry periods when no fresh water entering from the Magdalena can pass the artificial sediment barriers along the channels. This may be one of the reasons to which mangroves along the wetlands on both sides of the New Clarín channel still suffer from hypersalinity events and have not recovered (INVEMAR 2016). It is then necessary to create openings within the accumulated sediment barriers along both banks of the artificial channels such as the New Clarín, in order to support fresh water entry into the neighboring wetlands.

Conclusions
Regarding the hydrologic connectivity between the CGSM and the Magdalena River, we found evidence of recent fresh water inputs despite the fact that the connectivity between the river and CGSM had been his-torically impaired through the construction of a road, the control of flow by a sluice gate, and the heavy siltation in the channels. We discovered a negative relationship between change in mean daily discharge between the two acquisition dates of the interferogram and the accumulative difference between water levels along a transect at the mouth of an important channel, establishing fresh water entry. However, fresh water inputs into CGSM through this channel mainly occurred when Q exceeded 700 m 3 s −1 , which corresponds to only 30% of the days during the period 2005-2014. The study also demonstrated the way in which artificial channels that are supposed to enhance hydrologic connectivity within the wetland actually result in the creation of barriers to water connectivity within CGSM. In order to increase the effectivity of the restoration project of the CGSM, we first recommend the introduction of environmental restrictions for the sluice gate located at the mouth of the Renegado channel to guarantee fresh water inputs under both dry and wet conditions. Second, we advise the construction of openings within the accumulated sediment barriers along both banks of the dredged artificial channels in order to support fresh water entry into the neighboring wetlands. This remote sensing based analysis of wetland connectivity can be replicated in other wetlands worldwide that suffer from a combination of hydrologic modifications, ecological degradation, and lack of hydrologic data.