remote sensing Article
Automatic Measurement of Water Height in the As Conchas (Spain) Reservoir Using Sentinel 2 and Aerial LiDAR Data Higinio González-Jorge 1,2, * ID , Luis Miguel González-deSantos 1,2 ID , Joaquin Martínez-Sánchez 1,2 ID , Ana Sánchez-Rodríguez 1 and Henrique Lorenzo 1 1
2
*
Department of Natural Resources and Environmental Engineering, University of Vigo, As Lagoas s/n, 36310 Vigo, Spain;
[email protected] (L.M.G.-d.);
[email protected] (J.M.-S.);
[email protected] (A.S.-R.);
[email protected] (H.L.) Galician Aerospace Centre (CINAE), Porto do Molle, 36350 Nigrán, Spain Correspondence:
[email protected]; Tel.: +34-986-818-752
Received: 4 May 2018; Accepted: 5 June 2018; Published: 8 June 2018
Abstract: A methodology for the measurement of height in water reservoirs is developed. It is based on Sentinel 2 imagery and aerial LiDAR data. The methodology is automatized using Matlab software and focused on image processing techniques (equalization, binarization, and edge detection) combined with LiDAR data processing (near neighbour search and height averaging). It is applied in a region of interest selected by the user characterized by a water–land interface. Results are validated in the As Conchas water reservoir (Spain) using an in situ sensing system provided by the Hydrographic Miño-Sil Confederation. The duration of the experiment was one year. The Sentinel 2 bands B2, B3, B4, and B8 were tested during this study. The best results for water height evaluation were obtained for band B8 (842 nm) with an error of 0.20 m and a standard deviation of 0.17 m. The time resolution of the technique depends on the Sentinel 2 revisit time. The time resolution and height accuracy could be improved using complementary satellite systems. Keywords: remote sensing; earth observation; sentinel satellites; climate change; drought; water reservoir
1. Introduction Due to climate change, drought episodes are becoming more frequent in regions that traditionally did not face this problem. For example, Galicia, a typically rainy Spanish region, experienced drought phenomena during the summer and autumn of 2017. These facts, together with other environmental issues, such as water pollution, are of concern to the authorities and aroused interest in using techniques to monitor water bodies. One example in Galicia is the Hydrographic Miño-Sil Confederation [1], which is responsible for the management of one of the main hydrological basins of the region. Their data acquisition is based on in situ sensors that provide parameters as water level, flow, or chemical components. In this sense, the initiative described by Tauro et al. [2] must be noted, as it describes the works of the measurements and observation activities carried out by the XXI Century Working Group of the Association of Hydrological Sciences. This group combines hydrologists with experts in other fields such as robotics and electronics. George [3] and Dekker et al. [4] report on the limitations of conventional single point sampling methods to address heterogeneity and temporal changes in water quality. In these cases, remote sensing techniques based on different platforms (i.e., manned airborne systems, unmanned airborne systems (UAS) or satellites) appear as useful tools that provide the measurement of a number of water parameters [5]. Remote Sens. 2018, 10, 902; doi:10.3390/rs10060902
www.mdpi.com/journal/remotesensing
Remote Sens. 2018, 10, 902
2 of 9
Bandini et al. [6] used UAS for the measurement of water levels in rivers and lakes using the Global Navigation Satellite System (GNSS) receiver and payloads for range measurement (RADAR, SONAR, and laser distance meter). Manfreda et al. [7] conducted a review of the existing research studies and applications of UAS and concluded that they could be a bridge for the existing gap between the existing in situ field observations and the manned airborne or satellite observations. There are many important water masses that cannot be monitored in situ or for which the use of unmanned aerial vehicles is not recommended due to weather conditions or high operational cost. In addition, coastal and inland waters require representative spatial and temporal information that cannot be always provided by in situ sensing or UAS. The satellite market presents a good opportunity in this sense. For example, ESA [8] (Sentinel satellites) and NASA [9] (Landsat satellites) provide freely accessible and periodic information services to its users. Other non-free services are based on high-resolution imagery (i.e., GeoEye, Worldview, Spot, and Deimos). Applications of remote sensing to water monitoring cover several related aspects with, for example, chlorophyll concentration [10], oil spill detection [11], flood monitoring [12], the evaluation of the pathways of marine debris [13], and water level studies. The effort of the current work focus is on the latter and some highlighted works are next described. Tarpanello et al. [14] coupled MODIS and radar altimetry from ENVISAT data for discharge estimation in poorly gauged river basins. The discharge estimation accuracy is validated using in situ measurements along the Po river (Northern Italy). Birkett et al. [15] applied the radar satellite altimeter from NASA operating on-board the Topex-Poseidon satellite to the monitoring of large rivers and wetlands level. The methodology showed a water level monitoring geometric accuracy of 11 cm. Other examples of radar altimetry measurements are provided by the ENVISAT satellite. Frappart et al. [16] compared the performance of different algorithms applied to ENVISAT data to deliver reliable water levels for land hydrology using in situ sensing stations. Crétaux et al. [17] developed a database to monitor, in near real time, water level and storage variations using remote sensing data from altimetry missions (Topex-Poseidon, GFO, ERS-2, Jason and Envisat). It includes information for 150 lakes and reservoirs. Hall et al. [18] used ICESat laser altimetry observations to geodetically level gauge stations within the Amazon Basin. The method allows for calculation of slope and the discharge of rivers. Zang et al. [19,20] and Creataux et al. [21] also used ICESat laser altimetry to examine the water level of lakes, mass changes, and groundwater storage variations in the Tibetan Plateau to illustrate critical aspects and issues linked to the observation of climate change impact on surface waters. As can be observed in the above research, water level monitoring is mainly based on satellite altimetry missions. Currently, the Sentinel 3 [22] from ESA also includes an active topography package system that allows for the measurement of the height of sea and inland waters. Its measurements present a resolution of 300 m. Thus, other remote sensing methodologies are necessary for small- and medium-size inland water reservoirs. In the present work, the authors present a methodology for water level measurement in areas where a higher resolution than that provided by Sentinel 3 satellites is required. It is based on Sentinel 2 data combined with digital elevation models obtained from aerial LiDAR. The manuscript is structured in three main sections. Section 2 includes the materials and methods. It describes the area of study (As Conchas water reservoir in Spain), the data sources (aerial LiDAR and satellite imagery), and the data processing algorithm. Section 3 presents the results and discussion, where the accuracy of the procedure is verified with in situ water level data. Section 4 shows the conclusions of the manuscript. 2. Materials and Methods 2.1. Area of Study The area of study is the water reservoir “As Conchas” (Figure 1), located at the south of Ourense province in Spain. The reservoir is formed by the dam that holds the water of the Limia river at the
Remote Sens. 2018, 10, 902
3 of 9
Remote Sens. 2018, 10, x FOR of 9 a Baixa Limia district. ThePEER As REVIEW Conchas dam was constructed in 1949 and has a height of 48 m3 and length of 170 m. The geodetic height of the water level is approximately 500 m. The river basin reaches m and a length of 170 m. The geodetic height of the water level is approximately 500 m. The river 978 km2 between the municipalities of Muiños and Bande. Its main purpose is the production of 2 between the municipalities of Muiños and Bande. Its main purpose is the basin reaches 978 Remote Sens. 2018, 10,km PEER REVIEW 3 of 9 electricity, although itx FOR is also used for tourism, leisure, and agriculture. production of electricity, although it is also used for tourism, leisure, and agriculture. This reservoir monitored byofthe Miño-Sil Confederation which m This and areservoir length isofpermanently 170permanently m. The geodetic height theHydrographic water level is approximately 500 m. The [1], river is monitored by the Hydrographic Miño-Sil Confederation [1], provides daily hydrographic data as river temperature, dissolved oxygen, pH,isturbidity, 2 between basinprovides reaches 978 theheight, municipalities of Muiños Bande. Its dissolved main purpose the which dailykm hydrographic data asflow, height, flow, riverand temperature, oxygen, pH, etc.turbidity, Height water data provided the Confederation areagriculture. used are in this study asstudy ground production electricity, it isHydrographic alsobyused tourism, leisure, and etc. of Height wateralthough databy provided the for Hydrographic Confederation used in this truth to validate the data from the methodology here described. Thistruth reservoir is obtained permanently monitored bythe themethodology Hydrographic Miño-Sil Confederation [1], as ground to validate the data obtained from here described.
which provides daily hydrographic data as height, flow, river temperature, dissolved oxygen, pH, turbidity, etc. Height water data provided by the Hydrographic Confederation are used in this study as ground truth to validate the data obtained from the methodology here described.
Figure 1. Location of As Conchas water reservoir. Image source is Spanish Geographic Institute [23]. Figure 1. Location of As Conchas water reservoir. Image source is Spanish Geographic Institute [23]. Authors highlight with a red square the area of interest of As Conchas water reservoir. Authors highlight withofaAs red squarewater the area of interest As Conchas water reservoir. Figure 1. Location Conchas reservoir. Imageofsource is Spanish Geographic Institute [23]. Authors highlight 2.2. Aerial LiDAR Data with a red square the area of interest of As Conchas water reservoir. 2.2. Aerial LiDAR Data The Spanish Geographic Institute provides open LiDAR data for all the territory [23], consisting 2.2. Aerial LiDAR Data The Spanish Geographic Institute provides open LiDAR data2).for allground the territory [23],system consisting of digital files of true RGB-coloured LiDAR point clouds (Figure The reference is Spanish Geographic Institute provides open LiDAR data for all the territory [23], consisting of ETRS89 digital The files of true RGB-coloured LiDAR point clouds (Figure 2). The ground reference and units are divided in cells of 2 × 2 km, with a resolution of 0.5 points per squaresystem metre. is of digital files of true RGB-coloured LiDAR point clouds (Figure 2). The ground reference system is ETRS89 andsome unitsunderwater are dividedpenetration, in cells of 2and × 2the km, withunder a resolution 0.5 points square metre. Laser has terrain shallowofwater zones per can be detected ETRS89 and units are divided in cells of 2 × 2 km, with a resolution of 0.5 points per square metre. Laser has some underwater penetration, and the terrain under shallow water zones can be detected and 3D modelled by aerial LiDAR. Laser has some underwater penetration, and the terrain under shallow water zones can be detected and 3D by aerial LiDAR. andmodelled 3D modelled by aerial LiDAR.
Figure 2. LiDAR point cloud around the As Conchas water reservoir. Underwater terrain zones with Figure 2. LiDAR point cloud around the As Conchas water reservoir. Underwater terrain zones with Figure 2. LiDAR cloud around As Conchas water reservoir. Underwater terrain software. zones with no-data appear point as black points in thethe figure. The image is obtained from Cloud Compare no-data appear as black points in the figure. The image is obtained from Cloud Compare software. no-data appear as blackinpoints figure. The image is obtained from Cloud Compare software. The scale is presented metresinatthe bottom right. The scale is presented in metres at bottom right. The scale is presented in metres at bottom right.
Remote Sens. Sens. 2018, 2018, 10, 10, 902 x FOR PEER REVIEW Remote
44 of of 99
2.3. Satellite Imagery 2.3. Satellite Imagery Imagery (Figure 3) from the Sentinel 2 [24] satellite is used for the methodology. Sentinel 2 is an (Figure 3) fromdeveloped the Sentinel [24]European satellite isSpace used for the methodology. Sentinel 2 EarthImagery observation mission by 2the Agency inside the Copernicus is an Earth observation mission developed theforest European Space Agency inside the Copernicus Programme [25] to support services relatedby with monitoring, land cover changes detection Programme [25] to support services related with forest monitoring, cover2A changes detection and natural disaster management. It consists of two identical satellites,land Sentinel and Sentinel 2B. and natural management. of covering two identical satellites, Sentinel They providedisaster multi-spectral imageryItinconsists 13 bands the visible, nearSentinel infrared2A andand short wave 2B. They electromagnetic provide multi-spectral imagery in 13 covering thea visible, and short infrared spectrum. Sentinel 2 bands satellites provide revisit near time infrared of five days. The wave electromagnetic spectrum. Sentinel 2 satellites provide revisit time of five spatialinfrared resolution changes from 10 m (B2—blue, B3—green, B4—red, and aB8—near infrared) to days. 60 m The spatial resolution changes from 10 mB10—shortwave (B2—blue, B3—green, B4—red, and B8—near infrared) to (B1—coastal aerosol, B9—water vapour, infrared). 60 mLevel (B1—coastal B9—water vapour, B10—shortwave infrared). from the Copernicus Open 1C (topaerosol, of atmosphere reflectance) imagery is downloaded Level (top of twelve atmosphere reflectance) imagery is downloaded from the hydrologic Copernicuscycle. Open Access Hub1C [26]. The months of year 2017 were used to obtain a complete Access Hub [26]. The twelve months of year 2017 were used to obtain a complete hydrologic cycle.
Figure 3. Sentinel Figure 3. Sentinel 22 image image from from As As Conchas Conchas water water reservoir reservoir (red (red highlighted) highlighted) at at 24 24 December December 2017. 2017. Band B4. In the imagery, water appears as as black black and and sand sand as as white. white.
2.4. 2.4. Data Data Processing Processing The algorithmisisdescribed describedininFigure Figure The data processing is focused a region of interest The algorithm 4. 4. The data processing is focused on a on region of interest (ROI) (ROI) manually selected the user (Figure The ROI is focused on the land-water manually selected by the by user (Figure 5). The 5). ROI is focused on the land-water interface.interface. The use The of a use of a ROI prevents the necessity of large computing resources and makes the image processing ROI prevents the necessity of large computing resources and makes the image processing algorithm algorithm more robust. input to the algorithm of LiDAR the aerial LiDAR and2 more robust. Data inputData to the algorithm consists ofconsists the aerial point cloudpoint and cloud Sentinel Sentinel in (490 bands B2 (490 nm), B4 (560 nm),nm), B5 (665 andnm). B8 (842 nm). These are imagery 2inimagery bands B2 nm), B4 (560 nm), B5 (665 and nm), B8 (842 These bands arebands selected selected for two main reasons. They present the maximum spatial resolution (10 m) and good for two main reasons. They present the maximum spatial resolution (10 m) and good reflectance reflectance for the interface. water-landFigure interface. Figure 6 shows aprofile reflectance profile the contrast forcontrast the water-land 6 shows a reflectance between the between top-left and top-left and bottom-right corners of the As can be observed, band showscontrast. the higher contrast. bottom-right corners of the ROI. As can ROI. be observed, band 8 shows the 8higher Reflectance Reflectance data are obtained using SNAP software from ESA [27]. data are obtained using SNAP software from ESA [27]. QGIS is used used for for the the conversion conversion of of the the satellite satellite imagery imagery from from JPEG2000 JPEG2000 format format to to QGIS software software [28] [28] is GEOTIFF. On the other hand, Cloud Compare software [29] is used to convert point clouds from GEOTIFF. On the other hand, Cloud Compare software [29] is used to convert point clouds from LAS LAS format to Thus, PLY. Thus, the imagery andclouds point clouds in formats available for format to PLY. the imagery and point provideprovide outputsoutputs in formats available for Matlab Matlab processing [30],iswhich is the software used to implement the following algorithmic. processing [30], which the software used to implement the following algorithmic.
Remote Sens. 2018, 10, x FOR PEER REVIEW
5 of 9
Remote Sens. 2018, 10, 902 Remote Sens. 2018, 10, x FOR PEER REVIEW
5 of 9 5 of 9
Remote Sens. 2018, 10, x FOR PEER REVIEW
5 of 9
Figure 4. Algorithm workflow. Figure4.4.Algorithm Algorithm workflow. workflow. Figure Figure 4. Algorithm workflow.
Figure 5. of interest (ROI) (red square). Xand andYYare UTM (universal trasverse 5. Region of interest (ROI) (red square). Band Band B8. B8. XX (universal trasverse FigureFigure 5. Region (ROI) (red square). and areUTM trasverse Mercator) coordinates (H29). Mercator) coordinates (H29). Mercator) coordinates (H29).
Figure 5. Region of interest (ROI) (red square). Band B8. X and Y are UTM (universal trasverse Mercator) coordinates (H29).
Figure 6. Reflectance change in bands B2, B3, B4, and B8.
Figure 6. Reflectance change in bands B2, B3, B4, and B8. Figure 6. Reflectance change in bands B2, B3, B4, and B8. Figure 6. Reflectance change in bands B2, B3, B4, and B8.
Remote Sens. 2018, 10, x FOR PEER REVIEW
6 of 9
Satellite images are equalized (Figure 4) to increase the global contrast of the image and make the Remote edge Sens. detection robust. This step allows a better distribution of the intensities on the histogram 2018, 10, 902 6 of 9 and the areas of lower local contrast gain a higher contrast. The next step of image processing after equalization is a binarization process. The binarization process used the Otsu method, which Remote Sens. 2018, 10, x FOR PEER REVIEW 6 of 9 Satellite images are equalized (Figure 4) to increase the global contrast of the image and make minimizes the intra-class variance of the thresholded black and white pixels based on the threshold the edge detection robust. step (Figure allows 4) a better distribution of the intensities on the Satellite images are This equalized to increase the of the[31]. image and histogram make value. The 8-bit image histogram is used to compute theglobal Otsucontrast threshold After binarization, and the areasdetection of lower local This contrast gain aa higher contrast. of The step of processing the edge robust. step allows better distribution thenext intensities on image the histogram edge detection process is applied using the Sobel filter [32]. Edge detection finds the edges in the after and equalization islower a binarization process. binarization process used the Otsu method, which the areas of local contrast gain a The higher contrast. The next step of image processing after image byequalization approximating thevariance gradient of the image. It used is a pixels discrete differentiation operator is a binarization process. The binarization process the Otsu method, minimizes the intra-class of magnitude the thresholded black and white based on thewhich threshold thatvalue. approximates the gradient of is the intensity function. Figure 7based shows thethreshold binarization minimizes intra-class variance ofimage thetothresholded black andthreshold white pixels onbinarization, the The 8-bitthe image histogram used compute the Otsu [31]. After edge and value. The 8-bit image histogram is used to compute the Otsu threshold [31]. After binarization, edgedetection detection on theis ROI image. process applied using the Sobel filter [32]. Edge detection finds the edges in the image edge detection process is applied using the Sobel filter [32]. Edge detection finds the edges in the by approximating the gradient magnitude of the image. It is a discrete differentiation operator that image by approximating the gradient magnitude of the image. It is a discrete differentiation operator approximates the gradient of the image intensity function. Figure 7 shows the binarization and edge that approximates the gradient of the image intensity function. Figure 7 shows the binarization and detection the ROI edge on detection on image. the ROI image.
Figure 7. Edge detection process that finds the limit between water and sand. Figure 7. Edge detection process that finds the limit between water and sand.
Figure Edge detection process that findsimage the limitwith between watercorrelated and sand. with coordinates The ROI image is 7.developed from a satellite pixels (Figure 8). Thus, a set of 21 coordinates delimit the with water–land interface obtained from the The ROI image is developed fromthat a satellite image pixels correlated withiscoordinates (Figure 8).image Thus, aissetdeveloped of 21 coordinates the of water–land interface is obtained the The ROI from athat satellite image with pixels correlated with coordinates image. The number of coordinates depends ondelimit the area the region of interest andfrom the morphology image. The number of coordinates depends on the area of the region of interest and the morphology (Figure 8). Thus,interface. a set 21More coordinates that delimit the are water–land interfaceto is enable obtainedaveraging from the and of the water-land than ten coordinates recommended of The the water-land interface. More than tenon coordinates are recommended to enable averaging and of image. number of coordinates depends the area of the region of interest and the morphology obtain accurate results. These UTM–ETRS89 coordinates are correlated with the point cloud data obtain accurate results. These UTM–ETRS89 coordinates are correlated with the point cloud data the water-land interface. More than ten coordinates are recommended to enable averaging and obtain from aerial to find theirtheir corresponding heightvalues. values. neighbour algorithm fromLiDAR aerial LiDAR to find corresponding height TheThe nearnear neighbour algorithm is used is used accurate results. These UTM–ETRS89 coordinates are correlated with the point cloud data from aerial to determine the abovementioned XYXYcoordinates andthethe corresponding value. to determine the abovementioned coordinates and corresponding height height value. The final The final LiDAR to find their corresponding height values. The near neighbour algorithm is used to determine height of the water reservoir is obtained by averaging the 21 height data values previously height of the water reservoir is obtained by averaging the 21 height data values previously the abovementioned XY coordinates and the corresponding height value. The final height of the water calculated. calculated. reservoir is obtained by averaging the 21 height data values previously calculated.
Figure 8. Water-land limit interfacing (green) in an ROI (red) satellite image. Band B8.
3. Results and Discussion Figure 9 shows the height values obtained from Sentinel 2 bands B2, B3, B4, and B8. Although the authors attempted to cover all the year 2017, there are some data gaps mainly due to cloud cover, which is quite frequent in the Galicia region. The satellite data were compared with the in situ Results andand Discussion 3. Results Discussion sensing values provided by the Hydrographic Miño-Sil Confederation, which were used for ground truthing. Table 1 shows the error and standard deviation of the height values in comparison with the
Figure 8. Water-land (green)ininananROI ROI (red) satellite image. Figure 8. Water-landlimit limitinterfacing interfacing (green) (red) satellite image. BandBand B8. B8.
3.
Figure 9 shows the heightvalues values obtained obtained from 2 bands B2,B2, B3, B3, B4, and B8. Although Figure 9 shows the height fromSentinel Sentinel 2 bands B4, and B8. Although the authors attempted to cover all the year 2017, there are some data gaps mainly due to cloud cover, the authors attempted to cover all the year 2017, there are some data gaps mainly due to cloud cover, which is quite frequent in the Galicia region. The satellite data were compared with the in situ sensing which is quite frequent in the Galicia region. The satellite data were compared with the in situ
sensing values provided by the Hydrographic Miño-Sil Confederation, which were used for ground truthing. Table 1 shows the error and standard deviation of the height values in comparison with the
Remote Sens. 2018, 10, 902
7 of 9
values provided by the Hydrographic Miño-Sil Confederation, which were used for ground truthing. Table 1 shows the error and standard deviation of the height values in comparison with the in situ sensing data. is calculated sensing Remote Sens.The 2018,error 10, x FOR PEER REVIEWas the difference between the height data from the in situ 7 of 9 and the height data obtained using the methodology presented in this study. To obtain a global result, in situ sensing error is calculated as the difference between thefor height data from the inThe situ error the error values fordata. eachThe Sentinel 2 band are averaged using the results all the time series. sensing and the height data obtained using the methodology presented in this study. To obtain a does not present a clear trend for the month under study. In addition, to provide dispersion results, global result, the error values for each Sentinel 2 band are averaged using the results for all the time the standard deviation is also calculated. series. The error does not present a clear trend for the month under study. In addition, to provide The results obtained from the methodology presented in this work correlate greatly with the dispersion results, the standard deviation is also calculated. ground truth data. The errorfrom andthe standard deviation ranges 0.39correlate ± 0.23 m in thewith B2 band The results obtained methodology presented in from this work greatly the to 0.20 ± 0.17 m in the B8 band. As can be observed in Figure 6, the B8 band shows the higher difference ground truth data. The error and standard deviation ranges from 0.39 ± 0.23 m in the B2 band to 0.20 between themwater–sand reflectance which makes edge during ± 0.17 in the B8 band. As can be observed in the Figure 6, detection the B8 band showsthe theimage higherprocessing difference part the water–sand reflectance which are makes the edge detection during theobtained image processing of thebetween methodology robust. The error results relatively close to other values from satellite part of the methodology robust.between The error10–20 resultscm are[15,18]. relatively close to other values obtained from altimetry techniques which range satellite9altimetry techniques which range between cm [15,18]. of peaks and overestimation of Figure shows that the technique gives some 10–20 underestimation Figure 9 shows that the technique gives some underestimation of peaks and overestimation of be minimum values. There is no compelling reason why this may have occurred, although it could minimum values. There is no compelling reason why this may have occurred, although it could be due to the low resolution of the Sentinel 2 imagery in the horizontal plane (10 m). This means that due to the low resolution of the Sentinel 2 imagery in the horizontal plane (10 m). This means that the movements of the water mass at the edge lower than its magnitude cannot be detected, with the the movements of the water mass at the edge lower than its magnitude cannot be detected, with the consequent impact on the height measurement. consequent impact on the height measurement.
Figure 9. Height values obtained from B2, B3, B4, and B8 Sentinel imagery.
Figure 9. Height values obtained from B2, B3, B4, and B8 Sentinel imagery.
The methodology shows some limitations derived from the cloud cover, making it impossible to obtain the water-land interface imagery from the Sentinel satellite. could improve to The methodology shows some limitations derived from2the cloudFuture cover,works making it impossible the methodology a multi-satellite approach. This approach couldFuture combine the current obtain the water-landusing interface imagery from the Sentinel 2 satellite. works could Sentinel improve the 2 imagery using with Sentinel 1 data, which provide robust information independent of the cloud coverage. In methodology a multi-satellite approach. This approach could combine current Sentinel 2 addition, the spatial resolution and the consequent accuracy of height evaluation can be improved imagery with Sentinel 1 data, which provide robust information independent of cloud coverage. using satellites with higher spatial resolution (non-free imagery). Thus, higher temporal frequency In addition, the spatial resolution and the consequent accuracy of height evaluation can be improved of data and lower height value error can be achieved. Future research could also estimate the using satellites with higher spatial resolution (non-free imagery). Thus, higher temporal frequency of sensitivity of the method to the grey level threshold calculation during the image binarization data and lower height value error can be achieved. Future research could also estimate the sensitivity process. of the method to the grey level threshold calculation during the image binarization process. Table 1. Average error and standard deviation of the results.
Table 1. Average error and standard deviation of the results. B2 B3 B4 B8 Average error (m) 0.39 0.36 0.33 0.20 B2 B3 B4 B8 Standard deviation (m) 0.23 0.28 0.25 0.17 Average error (m) 0.39 0.36 0.33 0.20 Standard deviation (m) 0.23 0.28 0.25 0.17
Remote Sens. 2018, 10, 902
8 of 9
4. Conclusions A methodology for the evaluation of height value of water reservoirs was developed. It is based on the combination of free data from Sentinel 2 imagery and digital elevation models from aerial LiDAR, provided by national cartographic institutes. The methodology combines image processing and LiDAR data processing. Four Sentinel 2 bands are used for the study (B2, B3, B4, and B8). The best results appear for the infrared band B8, which depicts an error of 0.20 m and standard deviation of 0.17 m. The methodology could be easily automatized to provide continuous monitoring. Its results are of special interest for remote water reservoirs, where it is difficult to establish and maintain in situ sensing systems. In addition, a spatial resolution of 10 m is more adequate for medium to small water reservoirs than other satellite tools, which use an active topography instrument (i.e., the Sentinel 3 satellite with a resolution of 300 m). Author Contributions: H.G.-J. participate in the conceptualization of the manuscript, writing and review, L.M.G.-d., J.M.-S. and A.S.-R. in the development of the methodology and software implementations and H.L. in the formal analysis, discussion and conclusions. Funding: This work was partially funded by the Spanish Ministry of Interior (Grant SPIP2017-02122), Spanish Ministry of Economy, Industry and Competitiveness (Grants: EUIN2017-87598; TIN2016-77158-C4-2-R, RTC-2016-5257-7). The support of Xunta de Galicia through grant ED431C2016-038 is acknowledged. Acknowledgments: The authors specially acknowledge the support provided by the Hydrographic Miño-Sil Confederation. Conflicts of Interest: The authors declare no conflict of interest.
References 1. 2.
3. 4. 5.
6. 7.
8. 9. 10.
11.
12.
Hydrographic Miño-Sil Confederation. Available online: https://www.chminosil.es/es/ (accessed on 28 May 2018). Tauro, F.; Selker, J.; van de Giesen, N.; Abrate, T.; Uijlenhoet, R.; Porfiri, M.; Manfreda, S.; Caylor, K.; Moramarco, T.; Benveniste, J.; et al. Measurements and Observations in the XXI century (MOXXI): Innovation and multidisciplinarity to sense the hydrological cycle. Hydrol. Sci. J. 2018, 63, 169–196. [CrossRef] George, D.D. The airborne remote sensing of phytoplankton chlorophyll in the lakes and tarns of the English Lake District. Int. J. Remote Sens. 1997, 18, 1961–1975. [CrossRef] Dekker, A.G.; Vos, R.J.; Peters, S.W.M. Comparison of remote sensing data, model results and in situ data for total suspended matter (TSM) in the southern Frisian lakes. Sci. Total Environ. 2001, 268, 197–214. [CrossRef] Sawaya, K.E.; Olmanson, L.G.; Heinert, N.J.; Brezonik, P.L.; Bauer, M.E. Extending satellite remote sensing to local scales: Land and water resource monitoring using high-resolution imagery. Remote Sens. Environ. 2003, 88, 144–156. [CrossRef] Bandini, F.; JAkobsen, J.; Olesen, D.; Reyna-Gutierrez, J.A.; Bauer-Gottwein, P. Measuring water level in rivers and lakes from lightweight Unmanned Aerial Vehicles. J. Hydrol. 2017, 548, 237–250. [CrossRef] Manfreda, S.; McCabe, M.; Miller, P.; Lucas, R.; Madrigal, V.P.; Mallinis, G.; Ben Dor, E.; Helman, D.; Estes, L.; Ciraolo, G.; et al. On the use of unmanned aerial systems for environmental monitoring. Remote Sens. 2018, 10, 641. [CrossRef] European Space Agency. Available online: http://www.esa.int/ESA (accessed on 28 May 2018). NASA. Available online: https://www.nasa.gov/ (accessed on 28 May 2018). Zibordi, G.; Mélin, F.; Berthon, J.F. A time-series of above-water radiometric measurements for coastal water monitoring and remote sensing product validation. IEEE Geosci. Remote Sens. Lett. 2006, 3, 120–124. [CrossRef] Leifer, I.; Lehr, W.J.; Simecek-Beatty, D.; Bradley, E.; Clark, R.; Dennison, P.; Hu, Y.; Matheson, S.; Jones, C.E.; Holt, B.; et al. State of the art satellite and airborne marine oil spill remote Sensing: Application to the BP Deepwater Horizon oil spill. Remote Sens. Environ. 2012, 124, 185–209. [CrossRef] Feng, L.; Hu, C.; Chen, X.; Cai, X.; Tian, L.; Gan, W. Assessment of inundation changes of Poyang Lake using MODIS observations between 2000 and 2010. Remote Sens. Environ. 2012, 121, 80–92. [CrossRef]
Remote Sens. 2018, 10, 902
13. 14.
15. 16. 17.
18. 19. 20.
21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32.
9 of 9
Maximenko, N.; Hafner, J.; Niiler, P. Pathway of marine debris derived from trajectories of Lagrangian drifters. Mar. Pollut. Bull. 2012, 65, 51–62. [CrossRef] [PubMed] Tarpanelli, A.; Brocca, L.; Barbetta, S.; Faruolo, M.; Lacaba, T.; Moramarco, T. Coupling MODIS and Radar altimetry data for discharge estimation in poorly gauge driver basins. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2015, 8, 141–148. [CrossRef] Birkett, C.M. Contribution of the TOPEX NASA radar altimeter to the global monitoring of large rivers and wetlands. Water Resour. Res. 1998, 34, 1223–1239. [CrossRef] Frappart, F.; Calmant, S.; Cauhopé, M.; Seyler, F.; Cazenave, A. Preliminary results of ENVISAT RA-2-derived water levels validation over the Amazon basin. Remote Sens. Environ. 2006, 100, 252–264. [CrossRef] Crétaux, J.F.; Jelinski, W.; Calmant, S.; Kouraev, A.; Vuglinski, V.; Bergé-Nguyen, M.; Gennero, M.C.; Nino, F.; Abarca del Rio, E.; Cazenave, A.; Maisongrande, P. SOLS: A lake database to monitor in the Near Real Time water level and storage variations from remote sensing data. Adv. Space Res. 2011, 47, 1497–1507. [CrossRef] Hall, A.C.; Schumann, G.J.P.; Bamber, J.L.; Bates, P.D.; Trigg, M.A. Geodetic corrections to Amazon river water level gauges using ICESat altimetry. Water Resour. Res. 2012, 48, W06602. [CrossRef] Zhang, G.; Yao, T.; Xie, H.; Zhang, K.; Zhu, F. Lakes state and abundance across the Tibetan Plateau. Chin. Sci. Bull. 2014, 59, 3010–3021. [CrossRef] Zhang, G.; Yao, T.; Shum, C.K.; Yi, S.; Yang, K.; Xie, H.; Feng, W.; Bolch, T.; Wang, L.; Behrangi, A.; et al. Lake volume and groundwater storage variations in Tibetan Plaeau’s endorheic basin. Geophys. Res. Lett. 2017, 44, 5550–5560. [CrossRef] Crétaux, J.F.; Abarca del Río, R.; Mergé-Nguyen, M.; Arsen, A.; Drolon, V.; Clos, G.; Maisongrande, P. Lake volumen monitoring from space. Surv. Geophys. 2016, 37, 269–305. [CrossRef] Sentinel 3. Available online: http://www.esa.int/Our_Activities/Observing_the_Earth/Copernicus/ Sentinel-3/Instruments (accessed on 28 May 2018). Spanish Geographic Institute. Available online: http://centrodedescargas.cnig.es/CentroDescargas/locale? request_locale=en (accessed on 28 May 2018). Sentinel 2 Satellite. Available online: http://www.esa.int/Our_Activities/Observing_the_Earth/ Copernicus/Sentinel-2 (accessed on 28 May 2018). Copernicus Programme. Available online: http://www.copernicus.eu/ (accessed on 28 May 2018). Copernicus Open Access Hub. Available online: https://scihub.copernicus.eu/ (accessed on 28 May 2018). Snap Software. Available online: http://step.esa.int/main/toolboxes/snap/ (accessed on 28 May 2018). QGIS Software. Available online: https://www.qgis.org/en/site/ (accessed on May 28 2018). Cloud Compare Software. Available online: http://cloudcompare.org/ (accessed on 28 May 2018). Matlab Software. Available online: https://www.mathworks.com/products/matlab.html (accessed on 28 May 2018). Otsu, N. Threshold selection method from gray-level histrograms. IEEE Trans. Syst. Man Cybern. 1979, 9, 62–66. [CrossRef] Duda, R.; Hart, P. Pattern Classification and Scene Analysis; John Wiley Sons: Hoboken, NJ, USA, 1973; pp. 271–272. © 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
