A hydrological monitoring system, aimed at feeding a hydrological model (water balance), for a coastal lagoon should include at least the following:
Filtration exchanges with the sea and the nearby area can be neglected at first as the bottom of the lagoon is considered to be basically sealed by fine sediments (silt and clay) and the water head’s potential to generate filtration with the sea is very low, while the potential zone of exchange (the mouth) is quite limited.
It is spontaneous to think of a (possibly low-cost) technological system to measure several such variables. Low-cost systems are generally modular, thus allowing for the easy replacement of damaged parts. Modularity also allows for a wide freedom in the selection of components, including sensors, recorders, measurement algorithms, communication technology, the feeding source, and other characteristics [
15,
16,
17]. These systems are suited for a variety of contexts, e.g., water quality monitoring (nutrients, dissolved oxygen, etc.) in a river [
15]; environmental monitoring, e.g., the aquifer level, air quality, sediments, or the dynamics of wind-transported sands [
16]; management of urban waters [
17]; and measurement of water levels in a river [
18]. Therefore, we first evaluated this possibility for the different components of the target system.
2.1. River Inflow
No gauging station exists on the Tomarrazón-Camarones River; therefore, the first idea was clearly to set up a new, low-cost monitoring station of the river stage h with automatic measurements (by a pressure or distance sensor) and possibly tele-transmission of data [
18,
19]. In parallel, the goal was also to set up a suitable stage–discharge relationship Q(h), analogously to what has been achieved by [
5].
Unfortunately, our experience in the area led us to immediately discard this idea due to the high probability (or certainty) of rapid robbery of or damage to the devices: very poor people, who are quite numerous in the area, are prone to stealing anything that can reward them with amounts even smaller than a dollar.
Another option was to adopt satellite measurements of water level via-sensors, such as TOPEX/Poseidon, Jason-1, 2, and 3, ERS-1/2, Environmental Satellite-Envisat, ICESat, CryoSat-2, SARAL/AltiKa, and SWOT, with the support of software such as AlTiS, version 2.2.9 [
20]. However, analogously, this option was discarded for several reasons. The channel is almost invisible from above owing to the riparian vegetation cover, the channel is too narrow (10–20 m), and the frequency of survey of satellites is too low for our purposes. Nevertheless, these tools have produced data for two decades that can be very useful for studying, for instance, the evolution of water bodies in general [
21,
22,
23,
24].
As the rainy season was about to start when we were designing the system—and missing it would be a significant loss of data—a very simple but robust and even economic solution was set up. A rule was painted on a fixed vertical part of the wall of the foundation pillar of a bridge. Two rules were set up in the only two bridges in the area (
Figure 3b): Puente Troncal, the one selected for the routine measurements, and Puente Viejo, which is some distance upstream. This was conducted to perform a check on data coherence, explore the possible existence of a backwater effect, and provide an alternative calculation of flow based on the gradient of the water level (as explained later in this paper). Details of the rules on both bridges can be observed in
Figure 4.
The selected cross-sections are sufficiently stable because the riparian material is mainly organogenic claystone conglomerate, although, of course, a certain level of change in the sediment deposits may take place. With a survey during dry conditions, we determined the geometry of the cross-section so that the relationship A = A(h) was obtained, where A is the area of the wet cross-section for water level h (
Figure 5).
To simplify the routine measurement process as much as possible, we ensured that the water surface and the rule were visible in any condition from a selected site on top of the bridge and opted for just taking a picture, ideally a couple of times each day at (more or less) fixed times (early morning and mid-afternoon). Providentially, we could arrange a formal agreement with the local National Parks office, which immediately and enthusiastically shared the necessity of setting up such a project and accepted the pledge to have one of its staff members pass by the site and take the picture. In addition, we made a private agreement with a local person (muchacho) with a motorcycle, who, under a very reasonable payment, would commit to echar un ojo (having a look) at the river every day and who, in case of a flood or an emergency (e.g., the impossibility of a Parks servant passing by), would take an additional picture registering the time and immediately send it to us by WhatsApp® version 2.22.7.73. In spite of all these arrangements, in practice, however, most of the time, only one measurement a day was obtained (and some days even that was missing).
Still, the second part of the problem had to be faced: setting up a stage–discharge relationship. This meant gauging the flow rate, particularly during high flows, which is quite a dangerous task. The use of a classic current meter was out of the question because the bridge is too high over the water to hold the device with a long rigid arm, while an ADCP (Acoustic Doppler, Current-meter Profiler; [
25]) would require accessing the water body during high water to place a guiding rope/wire, a method that is physically very hard and dangerous owing to the dense vegetation one must cross and the slippery ground (
Figure 6). In addition, and most importantly, our University simply had no such device yet. On the other side, wading into the river would certainly be impossible because it becomes very deep and fast with a slippery bottom, while in low water, wading would interfere with the flow we had to measure. Other methods such as salt concentrations or dyes were not viable again because of the impossible access and because the river feeds a natural protected area. This is why we chose the old, somehow “primitive” method of throwing a floater and measuring the time elapsed to cover a fixed, known distance (of 11 m), and averaging amongst several (at least three) launches along different flow lines across the section. Only biodegradable objects such as fruits (ideal because they float but are almost fully immersed, meaning that the wind effect is minimal) or short wooden sticks were used. The flow rate was then obtained by just multiplying the average velocity by the area of the wet section, determined based on the water level h and the known geometric section A = A(h) previously determined in dry conditions (from
Figure 5). No correction to the velocity for the border effect was applied because the velocity field is quite complex and the main flow sometimes occurs on the side rather than in the central sector.
This approach would appear to most readers, as it does to us, as quite anachronistic and primitive. Nevertheless, the obtained data after more than one and a half years of observations behave surprisingly well, as shown in
Figure 7 (interpolated by a polynomial regression equation).
It is reasonable to wonder whether these 32 data pairs are sufficient to accurately calculate the relationship. The answer can perhaps be found in the final result, the reasonable consistency of modeled lagoon behavior (depending also on river inputs, estimated through this relationship) with respect to measurements of the lagoon state; however, this is dealt with in another forthcoming paper. Here, to confirm this positive result, and for scientific curiosity, a different method was adopted to estimate the flow rate based on classic hydraulics, i.e., by applying Chezy Manning’s Equations (1) and (2) [
26], which is often used in current research (e.g., [
27,
28,
29]). Here, we used the area A = A(h) and wetted perimeter p = p(h) relationships obtained from the cross-section geometry of
Figure 5 by points (
Figure 8), and for the slopes [m/m] of each measurement (at a given date and time), the water elevation difference Δy between the two bridges monitored (y
T at P.Troncal and y
V at P.Viejo, see
Figure 3), divided by their distance L along the stream axis (L = 1132 m from Google Earth
® images), establishes a reference elevation for both (the “IGAC 0”; with IGAC: Instituto Geográfico Agustin Codazzi, the official Colombian entity in charge of geographical issues and maps), so transforming the water depth h of our hydrometer into an elevation y. With this position, the stage–discharge curve at the P.Troncal station, as shown in
Figure 9, was obtained by estimating the Manning friction coefficient n from manual trial and error to fit the measured values as far as possible (obtaining the value n = 0.0447 in the SI system). Again, the results are surprisingly nice, as demonstrated by
Figure 9. Nevertheless, for the monitoring exercise, the empirical stage–discharge curve (based on our measurements) is preferred simply because this is closer to reality and because more data will be available in time, hopefully making it more reliable gradually.
with:
v [m/s]: average velocity in the cross-section
Q [m3/s]: flow rate
h [m]: water depth in the section
A [m2]: area of the wetted cross-section
R [m]: hydraulic radius, i.e., R(h) = A(h)/p(h)
s [-]: the river slope
p [m]: wetted perimeter
y [m.a.s.l]: water elevation: y = h + h0 with h0 IGAC reference [m.a.s.l].
Figure 8.
Analytic relationships (approximated) for the cross-section of the river at P.Troncal: (a) wetted area A = A(h); (b) wetted perimeter p = p(h).
Figure 8.
Analytic relationships (approximated) for the cross-section of the river at P.Troncal: (a) wetted area A = A(h); (b) wetted perimeter p = p(h).
Figure 9.
Matching between measured Q and Q estimated via the Chezy–Manning equation (R2 = 0.9738), with data from 23 April 2022–23 October 2023.
Figure 9.
Matching between measured Q and Q estimated via the Chezy–Manning equation (R2 = 0.9738), with data from 23 April 2022–23 October 2023.
However, things are never as smooth as they appear at first sight. While the data collection process was ongoing, we updated the exercise after an additional month (November 2023), when new significant flood events took place.
Figure 10 shows the stage–discharge curve with the new data, which shows significantly worse behavior, although it is still not bad.
The suspicion arose then that the deviations detected in the stage–discharge curve of
Figure 10 could depend on the backwater effect from the lagoon. Indeed,
Figure 11 plots the flow rate deviations (Q measured and Q estimated by the empirically found stage–discharge relationship) as a function of the water surface level of the lagoon, showing a certain tendency to over-estimate Q (i.e., a negative deviation) for increasing values of the water elevation in the lagoon (i.e., towards the right), as expected.
Therefore, we sought to improve the found power–law relationship by incorporating a dependency on the water elevation in the lagoon y
L, i.e., Q = Q(y
T, y
L). Specifically, a relationship with a term that would reduce the flow value Q for a lagoon elevation y
L closer to the river stage elevation y
T at P.Troncal (always higher than y
L) was set up, as shown by Equation (3), that simply mathematically expresses this concept. Its four parameter values have been calibrated by trial and error (y in [cm.a.s.l]):
where the parameters “a, b, y
0, and θ” were determined via calibration and assume the following values (Equations (4) and (5)):
The result is shown by the corresponding matching graph (
Figure 12), where a certain improvement is apparent both in terms of a closer position of the linear tendency line to the perfect matching line (solid red, 45 degrees) and in terms of dots that are closer, in general, to that line (indeed, R
2 = 0.9335 overcomes the value obtained by the mono-dimensional regression,
Figure 9). The improvement, however, is more relevant for low values of Q; however, this is clearly reasonable as the backwater effect of the lagoon vanishes for high flows as they are associated with high river stages while the lagoon water elevation moves in quite a limited range. It is certainly possible to better calibrate the set of parameters and even to find better functional relationships; however, for the moment, this is the stage–discharge relationship adopted hereafter.
2.2. Sea Level
The sea level is a fundamental variable because it determines the exchange relationship between the lagoon and the sea according to tides and the status of the mouth and, as such, it governs the annual life cycle of the same lagoon.
At first, the idea was to install a dedicated sensor to measure water levels with high frequency (as reported, for instance, in [
30]), but this idea was immediately discarded for the same security reasons already explained above and also for the absence of a suitable installation site. A much simpler solution was hence adopted by simply using existing reported sea elevation data (y
S); in our case, the tide gauge at Puerto Brisa (see
Figure 1a) located 39 km away from the mouth of the lagoon is the closest one.
However, getting the data—which are collected and owned by DIMAR (Colombian Dirección General Marítima y Portuaria)—is a process that requires administrative steps and time, as hourly data are not available online; hence, data are always obtained with a delay of a few months and only under explicit request. Another difficulty is the format of the data: they are delivered partly with a dd-mm-yyyy date format and partly with a mm-dd-yyyy format. A harsher difficulty is creating, within a continuous, hourly (Excel® 2019 MSO version 2406) data record sheet, an automatic reference to other sheets where our discontinuous time data of measurements of river flowrate and level, measurements of the lagoon level (usually bi-daily between 7–9 a.m. and 4–6 p.m., but with exceptions), and measurements of the area of the mouth and lagoon–sea exchange flow rate were recorded (by transcription of the physical field data formats). These data are collected at different, irregular times with many missing data (“holes”). To deal with this situation, we developed a specific Excel spreadsheet with suitable algorithms that proved to be indispensable, although far from trivial.
However, another, more serious problem concerned the altimetric consistency of data. According to common sense, the tide should oscillate most of the time around 0 with positive and negative values, although periods of higher or lower moving averages are possible due to particular combinations of astronomic and the meteorological drivers. However, here, almost all of the sea elevation data appeared to be much higher (about 60 cm) than the topographic IGAC 0 (
Figure 13), which is impossible because, in those conditions, no flow from the lagoon to the sea could occur, while it is clearly physically expected and was indeed observed in the field. We then asked for formal explanations from the relevant institutions (DIMAR, IGAC, and IDEAM—Instituto de Hidrología, Meteorología y Estudios Ambientales) and understood that the tide gauge, like the entirety of the Colombian national geodesic network managed by IGAC, is referenced to a sea level of 0 located on the Pacific coast in Bonaventura town, which is a completely different water body. Indeed, the average sea level in the Caribbean—where Puerto Brisas and Camarones are located—is in general 28 cm lower (in [
31,
32]). However, this fact does not solve the mismatch as both the river elevation and that of the lagoon are referenced to the same IGAC 0 and as such should be consistent. Therefore, the final explanation is that a deviation exists between the geoidal and the ellipsoidal models of the Earth’s surface in Colombia, as at present, there is no official update of the Colombian model. This is indeed consistent with the results found by [
33] when trying to validate geometric leveling points with classic topography and LIDAR data, finding an average difference of 0.63 m.
In order to proceed, it was therefore assumed that the P.Brisa tide gauge utilizes a different (unknown to us) local reference and without searching for that datum, the mismatching was solved by adopting a very operational criterion. As will be detailed in a forthcoming paper, we just imposed that the lagoon–sea flow exchange process was physically meaningful. This means that when the observed flow was outgoing (from the lagoon into the sea), the lagoon water elevation should be higher than that of the sea, and vice versa. Luckily, a meaningful value of a fixed vertical translation of the tide gauge data could be found by trial and error, which could fulfill this condition in all observed cases, except one. Considering that just a tiny level difference (a few centimeters) is involved, and that the tide gauge is 39 km apart and that there are often strong winds, the obtained result (denoted with yS*) can be considered fully satisfactory.
2.3. Lagoon: Water Level
The most spontaneous solution for the measurement of the water elevation of the lagoon seemed to be the installation of a rule fixed at a pole in a relatively central place in the lagoon so that even for low levels, the rule would still be immersed (as the lagoon is quite shallow and the bottom is characterized by a very gentle slope). Measurements would be taken by the personnel of National Parks from their headquarters on the shore using binoculars. However, this idea was soon discarded because the reflection of sunlight on the water would make remote reading impossible and because wind, the perennial companion of sunlight, generally provokes a high-frequency, irregular wave system with an amplitude of about 5–15 cm, which would severely affect the measurements. Even an automatic sensor, which might allow a certain degree of digital filtering of data, was discarded because of the usual security problem and in line with the idea of creating a very basic system.
Therefore, a manual device installed at the National Parks’ headquarters located on the lagoon shore (
Figure 1 for general location and
Figure 14) was chosen.
It is licit to wonder whether the foreseen monitoring frequency of twice a day is sufficient to capture the dynamics of the lagoon. According to Shannon [
34], in a linear (linearized) dynamic system, sampling should be carried out with a frequency not lower than T/10, where T is the minimum time constant. This criterion, as will be explained in a forthcoming publication, leads to a very wide range of values; specifically, when the mouth is open, the criterion provides values of 1.6 to about 5 h (depending on the filling status of the lagoon, being quicker when it is low); when the mouth is closed, and the core dynamics are governed by evaporation, values range from a couple of days to three months. Evidently, only in this second case is our monitoring definitely adequate, while, when the mouth is open, our measurements cannot capture the full oscillation process, which is tuned to the tide oscillations. Nevertheless, the data obtained can be very informative, as shown in the rest of this paper.
A mixed solution was eventually preferred, comprising a hydrometer (
Figure 15a) on the shore to be used for medium-high levels and a piezometer for situations with lower levels and, consequently, a dry hydrometer (
Figure 15b). As water levels can become lower than the local terrain, the hydrometer is enclosed in a typical yellow sewerage pipe set in a vertical position and the reading is performed by manually inserting a rule and reading the depth with respect to a horizontal reference set on the headquarters’ structure as indicated. To avoid the disturbance from the high-frequency wave oscillations, the device has a feeding tube with a reduced diameter section of Φ = 1.27 cm (which allows only a very small flow to pass through, according to the change of head from the lagoon surface and local, vertical movement), while the main vertical pipe section is proportionally much larger (Φ = 7.62 cm) so that the water volume input due to a flow increment translates into a much smaller vertical change, so fulfilling the dampening effect (“low-pass filter”). This function, on the other hand, is intrinsically guaranteed for the piezometer as the seepage across the soil cannot accelerate significantly; however, on the opposite side, this may dampen frequent (hourly) oscillations. The important difference is that water enters the hydrometer through the tiny tube, directly governed by the head on top of it, while water enters the piezometer by direct seepage across the soil matrix around it; therefore, the former cannot provide reliable data when the lagoon level drops below the sucking tip of the tube, while the piezometer works improperly when the level overcomes the ground level and even more so when it pours into the pipe from the top of the device.
The implementation of the devices is based on “home-made” very-low-cost technology, as shown in
Figure 16).
Measurements, after training, were taken by National Parks personnel twice a day, usually around 8 a.m. and 4 p.m. The operation consists of inserting a rigid rule inside the pipe with a block to set on the pipe edge and reading the wetted depth; from this, knowing the elevation of the reference beam edge, the elevation of water inside the pipe is determined. An alternative would have been a transparent cylinder with graduations marked on its external surface; however, the intense sunlight of the site would very soon deteriorate any plastic material, making reading impossible. The method adopted is more robust, but it requires corrections to be made because, as shown in
Figure 17, while inserting the rule (with a cross-section a*b), there is a Δh super-elevation that must be removed from the reading.
This correction is determined by imposing that the volume increment inside the pipe be equal to the volume of the piece of wet rule, according to Equation (6):
A
c = π D
2/4: area of the pipe’s internal cross-section
D [cm]: internal diameter of the pipe
h [cm]: depth of water inside the pipe already altered by the presence of the partly wet rule (which is the reading effectuated by Parks staff)
Δh [cm]: height difference generated by the rule
a, b: dimensions of the rule stick (specifically: width, a = 3.8 cm; thickness b = 1 cm).
From this, one obtains the correction to be applied to the reading, as shown in Equation (7):
2.4. Lagoon: Horizontality Hypothesis
A doubt arose about the fact that the water surface of the lagoon may not be horizontal all the time (or never), mainly because of the effect of wind or due to the hydraulic conditions governing the input and output of water flows when the mouth is open or when the river is flooding.
To ascertain whether the horizontality hypothesis can be acceptable, we adopted two criteria:
“instantaneous altimetry”: The digital elevation model (DEM) we utilized is based on photogrammetry and was generated from an aerial image dataset collected in 2017 (generously provided by a national government agency called ‘Fondo de Adaptación). The resulting orthophoto mosaic survey can be assumed to be instantaneous; therefore, the elevation of the water surface border, all around the lagoon, would be constant, were the hypothesis of horizontality verified. Unfortunately, the photo was taken in a dry period, and hence with a low level and small water surface, so that any structural difference cannot be very marked; nevertheless, from
Figure 18, a certain non-uniformity is seemingly evident, indicating that there might have indeed been a certain degree of tilt;
“synchronic monitoring”: By measuring the water elevation during the day with a relatively high frequency (every hour or so), both with our installed hydrometer and at the same time at an opposite point, namely the river at P. Viejo (so close to the lagoon that it can be assumed to coincide with its level at that point; see
Figure 1), it should be possible to detect any height difference. The results show a systematic elevation difference of approximately 15–20 cm between the two points (
Figure 19).
The outcome of these tests is not that straightforward to interpret. The first test is consistent with the conclusion that a certain tilt does exist, meaning that the horizontality hypothesis should be dropped. Possible reasons for this behavior are the sea outlet drawdown effect (as witnessed by our own data on the lagoon–sea exchange process according to hydraulics), the river input hydraulic load (e.g., [
35]), and wind seiches, although the latter are more relevant in large water bodies (e.g., [
36,
37]). However, the three campaigns were conducted on different days with different conditions at the mouth. Only in November was the river inflow really significant, eliminating the first two options; however, this would explain why the difference visible in
Figure 19b is smaller, as the river carrying a higher flood rose a bit. In turn, frequent moderate winds are a reality, which would explain the presence of seiches.
However, as visible from
Figure 20 and
Figure 1, it would be logical that the object with a higher elevation was the site at the river mouth (P.Viejo). However, from
Figure 19, it is evidently the reverse, as the river elevation is always lower than the lagoon elevation at its hydrometer. Several explanations can be conceived. One is that the seiche changes periodically and, by chance, all three campaigns resulted in the opposite situation to that of the satellite image taken in 2017. However, this is very unlikely as our (qualitative) record of wind direction says that the wind was more or less the same in the three conditions in terms of direction and intensity.
Another possibility is that the DEM assessment is not reliable; however, this hypothesis is likely to be dropped because we found a relatively reasonable consistency between the elevations and surface areas derived from the DEM and the lagoon elevation measurements (see the paragraph on morphometry). It may be possible, however, that it is reliable on average, but the differences it shows (higher and lower zones) are not, a possibility that could also be due to imprecisions in the definition of the water surface polygon. A third possible explanation is that there is a structural bias in the topographic survey fixing the “zeros” of the hydrometers. Indeed, this seems the most likely option as, although the survey started from official IGAC references in both cases, they were not physically coincidental owing to logistic constraints (absence of signal for the RTK equipment close to the lagoon mouth): an absolute difference between 20 and 40 cm is therefore possible. We cannot conclude whether there is a tilt or not because according to the instantaneous test (DEM), there seems to be; however, the time pattern test seems to contradict this conclusion as the detected difference has an opposite sign (however, this might be explained by a different topographic reference). More importantly, on all three days (with very different conditions at the mouth and the river), the results kept the same sign and even the same absolute value (which was a little lower in the third case, but this can be reasonably explained by the significant river inflow), which seems less likely and is more consistent with a hypothesis of identical levels (i.e., a horizontal surface or the absence of tilt) plus constant topographic bias.
However, it is important to observe that the measurement station of the lagoon level (hydrometer and piezometer) lies outside of the likely affected zones (
Figure 1 and
Figure 20). Therefore, assuming that the identified conditions (higher and lower zones) do not rotate around the lagoon, the water surface measurements can be considered representative of the real water surface elevation, which is extremely important for monitoring and modeling purposes. In principle, the possible bias between the river hydrometer and the lagoon hydrometer would not affect data acquisition because river elevation is used just to feed the local river stage–discharge relationship (and the relationship with the upstream station at P.Troncal that is using the same reference). However, there may be a subtle influence when the backwater effect is important through Equation (3), but this is negligible for high flows (which are the most important ones). On the other hand, lagoon elevation data are used just in relation to sea elevation data. However, the possible tilt of the lagoon water surface could be affecting hydrodynamic modeling.
In any case, our investigation still goes on to fully understand what is happening there. As a collateral note, it is important to note that the data on which this discussion is based cannot be considered exhaustive and fully representative as the DEM is quite imprecise (see par. 2.6), while our synchronic monitoring did not capture data at nighttime.
2.5. Lagoon: Exchange with the Sea
Being able to measure the exchange flow between the lagoon and the sea is key to setting up a water balance, even more so when water quality is dealt with. The key issue is finding a way to measure (or estimate) this flow in the simplest way.
Normally, there are one or two periods of the year when the lagoon mouth (“la boca”) opens and a significant flow of freshwater outgoing occurs; after a few days, the flow alternates twice a day between outgoing and incoming depending on the tide (and possible additional river floods) until the mouth closes again. This is quite a complex phenomenon, which is not easy to measure and hard to predict. Indeed, the opening date depends mainly on the arrival of the first significant river flood. This moment may vary greatly from year to year and may occur twice a year, owing to the bimodal hydrological regime with two wet seasons from April to June and September to November, respectively [
38]. Once the mouth is open, its geometry varies depending on the river inflow, the tide pattern, and the average sea level, usually approaching its maximum area in about two weeks. That configuration is usually held for one or two months and then the closure process starts, which is quite slow at the beginning and then accelerates, possibly during an interval of a month. The mouth area also varies during the day. However, this process is clearly different every year. In 2022, the opening lasted 36 days (from May 29 until July 4), and the second opening period, usually stronger, lasted a bit longer than 3 months (from 21 September 2022 until 5 January 2023, a total of 106 days).
The key problem to be addressed at this stage of monitoring is just measuring the water flow in several moments during the open mouth period, for both outgoing and incoming flows, and then setting up a relationship that determines the exchange flow Q
B (positive or negative) as a function of, possibly, the cross section area A, and the height difference between the lagoon water elevation (y
L) and that (y
S) of the sea, as shown in (Equation (8)):
Probably, the most suitable manner to carry out such a measurement is by using a digital current meter such as ADCP (Acoustic Doppler, Current-meter Profiler; [
25]) guided by a cable through the section. However, several reasons impeded this solution. Firstly, very pragmatically, our institution does not own such a device and the regional environmental authority was reluctant to lend us theirs because it can quickly become spoiled in brackish or salty water. Secondly, the wind can become quite strong and so does the wave surge. Therefore, the device, floating on the surface, would move significantly and irregularly and the data would become very noisy. On the other hand, consistently with the framework adopted, we chose to keep technology very simple and low-cost; therefore, we just pulled a rope as a reference across the section, and onboard a boat driven by hand through poles (an engine would alter measurements and easily get into trouble because of the vertical oscillations and irregular bottom), we measured the depth every 2 or 3 m (detected by colored knots on the rope) with a rule and the velocity with a current meter (two different devices to conduct a quality check on data: Manufacturer General Oceanics Environmental, Model 2030R Mechanical and manufacturer The Geography Specialists, Gepacks brand, model MFP126) at a depth of about 60% of the total depth to be hopefully more representative of the vertical averaged longitudinal velocity (see
Figure 21).
By adopting this method, several measurements have been carried out in different conditions, capturing both outgoing and incoming flows. This allowed us to set up a reliable relationship as expected and hoped (as will be described in a forthcoming paper). Therefore, actual systematic monitoring is reduced to components that have already been considered: the lagoon and sea water elevations.
2.6. Lagoon: Morphometric Relationships
As a basis for monitoring the storage changes, it is key to count the functional relationships: elevation–surface area and elevation–volume; these changes are indeed key components for the water balance. The water surface greatly varies with its elevation; this makes it imperative to merge a topographic representation of the zone with a bathymetric one.
We disposed of a set of aerial images taken during the dry season (17 March 2017) by a photogrammetric tripulated flight from an elevation of 1100 m above sea level (size of pixel: 20 cm). By using photogrammetric processing (via the Agisoft Metashape® software version 2.1.2), we generated an orthophoto mosaic of the whole scene and a dense points cloud describing the topography of the surrounding area around the lagoon. These points were manually edited and classified by using the ‘Auto-Classify ground points’ tool of Global Mapper Pro version 18®.
The bathymetry was reconstructed based on a set of 111 lagoon bottom elevation measurements directly determined by a GNSS-RTK from a boat, during 19, 20 of September 2022, together with an SIG-supported spatial analysis. The spatial pattern of measurement points reflects more operational navigation constraints (wind, depth, and distance) than logical planning (
Figure 22).
The adopted GIS strategy (ArcGIS Pro) to generate the bathymetric surface is articulated in 6 steps: (i) Geo-statistical interpolation (Kriging) of acquired points via GNSS-RTK. (ii) Extraction and smoothing of the resulting elevation curves. (iii) Interpolation of curves through TIN. (iv) Rasterization of the TIN generated. (v) Conversion of the bathymetric raster into a cloud points format “*.las” by utilizing the option “Export Layer to New File” in Global Mapper. (vi) Integration and interpolation of the photogrammetric points cloud and bathymetric surface to achieve their connection and thus the final topo-bathymetric DEM. For this last step, we adopted the tool “LAS Dataset to Raster” of ArcGIS Pro.
Independently, measurements of depth were taken by a sonar device (Garmin Striker) during the same bathymetric campaign, while in dry conditions, we carried out point GNSS-RTK measurements in the floodable zone of the lagoon (then in dry conditions). All of this was to create a database with which validation of the generated topo-bathymetric DEM would be possible.
The key outputs of this activity are the surface area–elevation S(y) and volume–elevation V(y) relationships which have been obtained by points (
Figure 23) by means of the r.lake.xy module, a spatial modeling tool hosted in GRASS GIS
® software version 7.8.7. This module fills the water body (topo-bathymetric DEM) from a given elevation until a specified elevation is achieved. This tool requires an input of a raster DEM, a maximum water elevation, and its location coordinates (x,y).
A final key check was to ascertain the coincidence of DEM elevation with measurements of the lagoon water surface. More precisely, the idea was to identify some dates with different conditions of the lagoon (at least 4); given the date, we would determine the area S of the water surface from satellite images (transformed into a polygon and eliminating possible “holes”). Then, from the inverse curve y = y(S), we would obtain the elevation y to be compared with the elevation measured that day by the hydrometer/piezometer devices and determine a fitting measure. We obtained an RMSE value of just 8.4 cm (
Table 1), which is considered satisfactory. Data from May 31 show the maximum deviation; however, a measurement error seems unlikely because both the piezometer and the hydrometer devices recorded a similar jump between morning and afternoon (a cause for this jump is a possible pressure effect due to stronger winds on the water surface). The significant deviation with respect to the DEM value is probably due to the fact that this value is not associated with a specific time instant, while the measurement is, so they are most probably out of phase.