Autumnal Beach Litter Identification by Mean of Using Ground-Based IR

environments

Article Autumnal Beach Litter Identiﬁcation by Mean of Using Ground-Based IR Thermography

Cosimo Cagnazzo 1,* , Ettore Potente 1 , Hervé Regnauld 2, Sabino Rosato 3 and Giuseppe Mastronuzzi 1

1 Department of Earth and Geo-Environmental Science, University of Bari, 70121 Bari, Italy; [email protected] (E.P.); [email protected] (G.M.) 2 Littoral, Environnement, Télédétection, Géomatique, University of Rennes 2, 35043 Rennes, France; [email protected] 3 Serveco Srl, 74020 Montemesola, Italy; [email protected] * Correspondence: [email protected]

Abstract: The progress of scientiﬁc research and technological innovation are contributing to an increase in the use of rapid systems for monitoring and identifying geo-environmental processes related to natural and/or anthropogenic activities. The aim of this study is identifying autumnal beach litter using ground-based IR thermography. Starting from quarterly autumn monitoring data of air temperature and sandy soil surface temperature, an empirical equation between the two environmental matrices (air and sandy soil) is obtained. This will allow the calculation of the sandy soil surface temperature knowing only the air temperature. Therefore, it will be possible to know in advance the thermal response of the sandy soil, thus creating a thermal blank of the beach. Using an IR thermal camera, it is possible for a quicker identiﬁcation of thermal anomalies of the coastal

area potentially connected to the presence of pollution due to the anthropogenic origin (particularly plastic material). The test area is located in the area of the Coastal Dunes Regional Natural Park of

Citation: Cagnazzo, C.; Potente, E.; Ostuni–Fasano in Apulia (southern Italy). Regnauld, H.; Rosato, S.; Mastronuzzi, G. Autumnal Beach Keywords: beach litter; infrared thermography; UAV; UGV; environmental monitoring; coastal pol- Litter Identification by Mean of Using lution Ground-Based IR Thermography. Environments 2021, 8, 37. https:// doi.org/10.3390/environments8050037 1. Introduction Academic Editor: Sílvia C. Goncalves Among the atmospheric and non-atmospheric parameters that influence the proper- ties of the climate system, temperature is a very relevant parameter in all the chemical, Received: 22 March 2021 physical and biological processes that affect the soil formation and its persistence in a Accepted: 22 April 2021 natural environment. Considering the same source of thermal radiation, each material is Published: 24 April 2021 heated differently according to its chemical-physical characteristics. Scientific research and technological innovation have developed new methodologies (IR thermography) and new Publisher’s Note: MDPI stays neutral investigation tools (thermal cameras) based on the use of temperature for the identifica- with regard to jurisdictional claims in tion of different materials, such as sandy soil. IR thermography allows for detecting and published maps and institutional affil- quantifying the infrared energy emitted by any object above absolute zero temperature iations. (−273.14 ◦C). Since every object characterized by a temperature above absolute zero emits thermal energy, it can be identified through infrared thermography. Thermal cameras allow for quick and remote collection of a large amount of data. Copyright: © 2021 by the authors. Over the last few years, the use of thermography has become widespread in various Licensee MDPI, Basel, Switzerland. fields such as: research and development, quality and process control, healthcare, con- This article is an open access article struction, industry and the mechanical field [1–4]. In the environmental monitoring field, distributed under the terms and thermography had a remarkable development, especially in wildfire detection [5], while is conditions of the Creative Commons only a first approach for flora and vegetation habitat monitoring, and sometimes also for Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ wildfauna, which necessarily require observations and confirmations directly in the field. 4.0/).

Environments 2021, 8, 37. https://doi.org/10.3390/environments8050037 https://www.mdpi.com/journal/environments Environments 2021, 8, x FOR PEER REVIEW 2 of 12

Environments 2021, 8, 37 also for wildfauna, which necessarily require observations and confirmations directly2 of in 12 the field. Moreover, the growing anthropogenic impact on the natural coastal environment is causingMoreover, a significant the growingincrease in anthropogenic coastal pollution impact [6] and on spills the natural of chemical coastal products environment in the marineis causing environment a significant [7]. increaseThe need in to coastalidentify pollution the areas [affected6] and spills by such of events chemical is affirming products thein theuse marine of thermography environment as [7 ].a The new need techni to identifyque, complementary the areas affected to other by such traditional events is methodologiesaffirming the use [8–10]. of thermography as a new technique, complementary to other traditional methodologiesIR thermal [ 8cameras–10]. are also installed on aerial (UAV) and terrestrial (UGV) platforms,IR thermal resulting cameras in effective are also equipment installed on in aerialterms (UAV) of time and efficiency terrestrial and (UGV) timeliness platforms, [11– 16].resulting in effective equipment in terms of time efficiency and timeliness [11–16]. TheThe aim aim of of this this study study is is identifying autumnal autumnal beach beach litter litter using using ground-based ground-based IR IR thermography.thermography. Starting Starting from from statistical statistical analysis analysis results results of of monitoring monitoring data data collected collected by by twotwo air air temperature temperature and and sandy-soil sandy-soil temperatur temperaturee sensors, sensors, installed installed in in the the Coastal Coastal Dunes Dunes RegionalRegional Natural Natural Park Park (Ostuni–Fasano, (Ostuni–Fasano, Italy) Italy) (Figure (Figure 11),), an empirical relation between thethe environmentalenvironmental matrices matrices (air (air and and sandy sandy soil) soil) will will be be found. found. This This will will allow allow to to estimate estimate the the sandy-soilsandy-soil surface surface temperature temperature kn knowingowing only only the the air air temperature. temperature.

Figure 1. (a) Study area localization in the Coastal Dunes Regional Natural Park (Apulia, Italy); Figure 1. (a) Study area localization in the Coastal Dunes Regional Natural Park (Apulia, Italy); (b) Polar(b) Polar plots plots of the of wind the wind speed speed distribution distribution and andwave wave height height distribution. distribution. By estimating the sandy-soil surface temperature it will be possible to rapidly detect any thermal anomalies along the sandy coast, such as anomalies deriving from plastic pollution, by using thermal cameras installed on aerial or terrestrial platforms. The results Environments 2021, 8, x FOR PEER REVIEW 3 of 12

Environments 2021, 8, 37 3 of 12 By estimating the sandy-soil surface temperature it will be possible to rapidly detect any thermal anomalies along the sandy coast, such as anomalies deriving from plastic pollution, by using thermal cameras installed on aerial or terrestrial platforms. The results of this studyof will this helpstudy with will the help optimization with the optimization of infrared thermographyof infrared thermography applications applications in the in the coastal environmentalcoastal environmental monitoring. monitoring.

1.1. Ecological1.1. Setting Ecological Setting The CoastalThe Dunes Coastal Regional Dunes Natural Regional Park Natural extends Park from extends the coast from towards the coast inland towards inland agriculturalagricultural areas, occupied areas, by centuries-oldoccupied by olive centuries-old groves. It includes olive thegroves. Site ofIt Community includes the Site of ImportanceCommunity (SIC) “Litorale Importance Brindisi”, (SIC) included “Litorale in the Brindi Europeansi”, included network in “Natura the European 2000”. network In the Park“Natura are many 2000”. species In ofthe ﬂora Park that are are many well species preserved, of flora such that as theare psammophilouswell preserved, such as the and Cakiletumpsammophilous maritimae [17 and–19 ].Cakiletum The system maritimae erosion and[17– the19]. relative The system vegetation erosion are and not the relative caused by thevegetation wind but are are not mainly caused due by to the people wind leveling but are themainly sand due in the to people summer leveling season, the sand in especially inthe the summer investigated season, site andespecially in general in the on theinvestigated whole psammophilous site and in coastalgeneral area on the whole that extendspsammophilous from the site study coastal and area far asthat Brindisi. extends from the site study and far as Brindisi. The Park wasThe established Park was established with the aim with of conserving the aim of co andnserving recovering and therecovering habitats the and habitats and animal andanimal plant species and plant indicated species in indicated Community in Community Directives 79/409/EEC Directives 79/409/EEC and 92/43/EEC. and 92/43/EEC. 1.2. Geological and Geomorphological Setting 1.2. Geological and Geomorphological Setting The area covered by this study is located on the Apulian Adriatic coast, in the Coastal The area covered by this study is located on the Apulian Adriatic coast, in the Coastal Dunes Park, stretched for 6 km from Torre Canne to Torre San Leonardo. The area hosts a Dunes Park, stretched for 6 km from Torre Canne to Torre San Leonardo. The area hosts coastal mobile system characterized by the presence of several lakes and ponds and by a a coastal mobile system characterized by the presence of several lakes and ponds and by a polyphasic dune belt parallel to the coast that reaches an altitude of about 17 m (Figure2). polyphasic dune belt parallel to the coast that reaches an altitude of about 17 m (Figure 2).

Figure 2. View Figureof the study 2. View area of (19 the July study 2018). area The (19 characteristics July 2018). The of characteristics the beach change of the quickly beach an changed dramatically quickly anddepending on the storm surge. dramatically depending on the storm surge.

The depositsThe consist deposits of medium consist of sand medium characterized sand characterized by carbonates, by carbonates, quartz and quartz other and other minerals inminerals very small in percentagesvery small percentages (pyroxene and(pyroxene feldspar); and there feldspar); are also there rare are fragments also rare fragments of siliciclasticof rockssiliciclastic and material rocks and of anthropogenicmaterial of anthropogenic origin [20–22 origin]. [20–22].

1.3. Climatological1.3. Climatological Setting Setting The averageThe annual average temperature annual temperature of the study of areathe study ranges area between ranges 15 between and 16 ◦15C and 16 °C according toaccording 1971–2000 to monitoring1971–2000 monitoring data [23]. data [23]. The whole coastalThe whole sector coastal is exposed sector tois aexposed wind regime to a wi characterizednd regime characterized by winds coming by winds coming from the north-westernfrom the north-western quadrants quadrants (Figure1), (Figure therefore 1), stormtherefore surges storm are surges very frequent are very frequent in in winter andwinter also and cause also the cause stranding the stranding of a large of amount a large of amount natural andof natural anthropogenic and anthropogenic material. Thematerial. main direction The main of direction the longshore of the transport longshore is transport NW-SE [24 is] NW-SE (Figure 1[24]). (Figure 1).

Environments 2021, 8, x FOR PEER REVIEW 4 of 12 Environments 2021, 8, x FOR PEER REVIEW 4 of 12 Environments 2021, 8, 37 4 of 12

2. Materials and Methods 2. Materials and Methods In order to create a thermal blank of the beach and consequently identify the thermal 2. MaterialsIn order and to create Methods a thermal blank of the beach and consequently identify the thermal polluting anomalies (attributable to the beach litter) with the thermal camera (FLIR C3), pollutingIn order anomalies to create (attributable a thermal blank to the of beach the beach litter) and with consequently the thermal identify camera the (FLIR thermal C3), two sensors (Elitech RC-4) equipped with a data logger were installed (Figure 3 and pollutingtwo sensors anomalies (Elitech (attributable RC-4) equipped to the with beach a litter)data logger with the were thermal installed camera (Figure (FLIR 3 C3),and Figure 4). Sandy soil surface temperature data and air temperature data were collected. twoFigure sensors 4). Sandy (Elitech soil RC-4) surface equipped temperature with adata data and logger air temperature were installed data(Figures were collected.3 and4). Soil moisture affects soil temperature just as air humidity affects air temperature. SandySoil moisture soil surface affects temperature soil temperature data and air just temperature as air humidity data were affects collected. air Soiltemperature. moisture Furthermore, the humidity of the air could be influenced by other atmospheric parameters affectsFurthermore, soil temperature the humidity just of as the air air humidity could be influenced affects air by temperature. other atmospheric Furthermore, parameters the (rain, wind, atmospheric pressure) and the humidity of the sandy soil by other parameters humidity(rain, wind, of atmospheric the air could pressu be influencedre) and the by humidity other atmospheric of the sandy parameters soil by other (rain, parameters wind, (vegetation, storm surges, rain). Since the thermal imager acquires only the temperature atmospheric(vegetation, pressure)storm surges, and therain). humidity Since the of thethermal sandy imager soil by acquires other parameters only the temperature (vegetation, data, the experimental monitoring activity concerned only the temperature data (air and stormdata, the surges, experimental rain). Since monitoring the thermal activity imager concerned acquires only only the the temper temperatureature data data, (air theand soil). This was done to make the thermographic technique a methodology to support other experimentalsoil). This was monitoring done to make activity the thermograp concerned onlyhic technique the temperature a methodology data (air to and support soil). other This new methodologies (spectral sensors, artificial intelligence algorithms) in emergency wasnew donemethodologies to make the (spectral thermographic sensors, techniqueartificial intelligence a methodology algorithms) to support in otheremergency new coastal pollution conditions where it is necessary to be quick in identification and methodologiescoastal pollution (spectral conditions sensors, where artificial it is intelligencenecessary to algorithms) be quick inin emergencyidentification coastal and mapping. pollutionmapping. conditions where it is necessary to be quick in identification and mapping.

Figure 3. Sensor localization; and IR thermography survey area. FigureFigure 3.3. SensorSensor localization;localization; andand IRIR thermographythermography surveysurvey area.area.

(a) (b) (a) (b) Figure 4. Air temperature sensor (a); sandy-soil surface temperature sensor (b). The pictures clearly show the presence of FigureFigure 4.4.Air Air temperaturetemperature sensorsensor ((aa);); sandy-soil sandy-soil surface surface temperature temperature sensor sensor ( b(b).). TheThe picturespictures clearly clearly show show the the presence presence of of vegetation typical of the dune environment (Cakiletum, Ammophiletum and Agropyretum). vegetationvegetation typical typical of of the the dune dune environment environment (Cakiletum, (Cakiletum, Ammophiletum Ammophiletum and and Agropyretum). Agropyretum).

Environments 2021, 8, 37 5 of 12

The first thermal sensor inserted in a meteorological screen (solar radiation protection), was installed 2 m from the ground to measure air temperature (Figure4a) as recommended by WMO (World Meteorological Organization) guidelines. The sensor is able to measure temperatures ranging from −40 ◦C to +80 ◦C with a 0.1 ◦C resolution and a 0.5 ◦C accuracy. The Elitech RC-4 sensor with probe was used for the sandy soil surface temperature. It has the same technical specs, and the data logger was positioned on the dune into a waterproof box, protected from bad weather and anthropic interference, and the probe (connected to the data logger) directly in contact with the sandy soil surface (Figure4b). Both sensors were set to a sampling rate of 5 min, the monitoring started on 7 October 2018 and ended on 15 December 2018. The monitoring period was chosen to avoid anthropogenic disturbances related to tourism, which could compromise the sensors installed. The collected data were downloaded and exported in .txt format, in a dataset gathering about 20,000 thermal data for each sensor. The statistical analysis was carried out using Microsoft Excel 2019 and R 4.0.3 software package. Daily average air and sandy-soil surface temperature was calculated, as well as Pearson’s correlation coefficient. This coefficient is the test that measures the statistical relationship between two continuous variables, giving information about the correlation and the direction of the relationship, and it ranges between −1 (strong negative linear relationship) and +1 (strong positive linear relationship). A null coefficient means there is no linear relationship between the variables. Moreover, by analysing the scatter plot, the line of best fit and the empirical equation were obtained. The coefficient of determination R2 was calculated to understand the accuracy of the regression model used to make predictions. The coefficient ranges between 0 and 1 and it is a statistical measure of how close the data are to the fitted regression line. Finally, the RMSE (Root Mean Squared Error) was calculated, and the daily average temperature values observed were compared with the ones obtained from the prediction model. Six thermal images 80 × 60 (4800 pixels) on the ground were collected in the study area, using an FLIR C3 thermal imaging camera (Figure3). Its sensor can detect and measure temperatures between −10 ◦C and +150 ◦C in a spectral band ranging from 7.5 to 14 µm, to a resolution of 0.1 ◦C and an accuracy of ±2 ◦C. The images were collected on 30 October 2018, twice in the morning (8:00 AM UTC and 9:00 AM UTC) and twice in the afternoon (12:00 PM UTC and 2:30 PM UTC); and on 31 October 31 in the morning (5:30 AM UTC) and in the afternoon (01:00 PM UTC). The observed air temperature was: • 16.5 ◦C ± 0.5 and 17.1 ◦C ± 0.5 (on 30 October 2018, respectively, at 08:00 AM UTC and at 09:00 AM UTC); • 24.8 ◦C ± 0.5 and 19.5 ◦C ± 0.5 (on 30 October 2018, respectively, at 12:00 PM UTC and at 02:30 PM UTC); • 14.4 ◦C ± 0.5 and 23.9 ◦C ± 0.5 (on 31 October 2018, respectively, at 05:30 AM UTC and at 01:00 PM UTC). The thermal images were processed with FLIR Tools software, setting the sand emis- sivity (ε = 0.90). For each radiometric thermal acquisition, knowing the air temperature, the empirical equation was applied in order to obtain a prediction of the sandy soil thermal range, taking into account the instrumental error. Finally, for each thermal image processed, the GIS software created 10 random points. Later the same points were imported into the corresponding RGB images to understand what they really identified. It was assessed that the degree of confidence of the results through the calculation of the kappa coefficient [25,26] allowed for the assessment of the degree of confidence of the results on a morning and afternoon scale. The kappa index measures the agreement between different assessments for the classification of the same object. Kappa index is defined as:

K = (po − pe)/(1 − pe) (1)

po = observed agreement; pe = hypothetical probability of agreement. Environments 2021, 8, x FOR PEER REVIEW 6 of 12

Environments 2021, 8, 37 6 of 12 K = (po − pe)/(1 − pe) (1) po = observed agreement; pe = hypothetical probability of agreement. TheThe calculation calculation of of the the index index is is carried carried out out with with the the construction construction of of a a matrix matrix called called the the confusionconfusion matrix. matrix. ThisThis index index ranges ranges from from 0 0 to to 1 1 and and expresses expresses the the correlation correlation between between the the homologues homologues pointspoints in in the the IRT IRT and and RGB RGB image. image. K K = = 1 1 indicates indicates perfect perfect agreement agreement while while K K = = 0 0 indicates indicates absenceabsence of of agreement. agreement. Generally, Generally, a a value value of of the the index index k k > > 0.750.75 indicatesindicates aa goodgood agreement.agreement.

3.3. Results Results and and Discussion Discussion DailyDaily averageaverage air air temperature temperature and and sandy-soil sandy-soil surface surface temperature temperature are shown are shown in Figure in Figure5, while5, whileTable Table 1 shows1 shows the themaximum maximum and and mi minimumnimum values values of ofthe the daily daily average average air airtemperature temperature and and sandy sandy soil soil surface surface temperature. temperature. Figure Figure 55 showsshows thatthat the the sandy-soil sandy-soil surfacesurface temperature temperature is is always always lower lower than than the the air air temperature. temperature. This This happens happens because because in the in autumnthe autumn the surface the surface of the of sandy the sandy soil, insoil, this in studythis study area, area, is always is always wet. wet.

FigureFigure 5. 5.Daily Daily average average air air and and sandy sandy soil soil surface surface temperature temperature during during the the monitoring monitoring period. period.

Table 1. Extreme values of average temperature.

Table 1. Extreme values of average temperature. Air Sandy-Soil Surface Maximum daily average temperature (◦C) 23.5Air Sandy-Soil 22.2 Surface ◦ MinimumMaximum daily daily average average temperature temperature ( C) (°C) 7.8 23.5 5.5 22.2 Minimum daily average temperature (°C) 7.8 5.5 The Pearson correlation coefﬁcient shows a strong positive correlation (0.97) between averageThe daily Pearson air temperature correlation coefficient and average shows daily a sandy-soilstrong positive surface correlation temperature (0.97) soil. between The bestaverage model daily for theair temperature study area, approximating and average da theily empiricalsandy-soil data, surface was temperature calculated from soil. theThe scatterbest model plot (Figure for the6 study) and itarea, results approximating in a non-linear the regression, empirical data, a power was function. calculated In from fact, inthe thescatter interpolation plot (Figure curve, 6) and a slight it results upwards in a concavitynon-linear can regression, be observed. a power function. In fact, in the interpolation curve, a slight upwards concavity can be observed.

Environments 2021, 8, 37 7 of 12 Environments 2021, 8, x FOR PEER REVIEW 7 of 12

Environments 2021, 8, x FOR PEER REVIEW 7 of 12

Figure 6. Scatter plot between average air temperat temperatureure and sandy soil surface temperature. Figure 6. Scatter plot between average air temperature and sandy soil surface temperature. The resulting model is represented by the equation: The resulting model is represented by the equation: Ts = 0.5572 × Ta1.1701 (2) Ts = 0.5572 × Ta1.1701 (2) 1.1701 Ts = sandy-soil surface temperature;Ts = 0.5572 Ta = air × Tatemperature. (2) TheTs = sandy-soilcoefficient surface of determination temperature; R2 Ta = 0.94 = air has temperature. validated the statistical model. The The coefﬁcient of determination R2 = 0.94 has validated the statistical model. The empiricalThe coefficientequation obtainedof determination was used R 2to = pr0.94edict has the validated daily average the statistical sandy-soil model. surface The empirical equation obtained was used to predict the daily average sandy-soil surface temperatureempirical equation knowing obtained the air temperature.was used to Therpredictefore, the the daily results average were comparedsandy-soil with surface the temperature knowing the air temperature. Therefore, the results were compared with the temperature knowingmeasured the on airsite temperature. (Figure 7). Therefore, the results were compared with the temperature measured on site (Figure7). temperature measured on site (Figure 7).

Figure 7. Comparison between measured and predicted sandy soil surface temperature during the same monitoring period. Figure 7.7. ComparisonComparison between between measured measured and and predicted predicted sandy sandy soil surfacesoil surface temperature temperature during during the same the monitoring same monitoring period. period. RMSE has been obtained through statistical processing. It is a statistical indicator that quantifiesquantiﬁesRMSE the has deviation been obtained between through the observedobservstatisticaled andprocessing. simulated It is data. a statistical The value indicator of RMSE that stoodquantifies atat 0.96.0.96. the The deviationThe very very low between low value value conﬁrms the confirmsobserv theed accuracy theand accuracysimulated of the empiricalof data. the Theempirical equation value of obtained.equation RMSE obtained.stoodFor statistical at 0.96.For statistical completeness, The very completeness, low a linearvalue regressionconfirms a linear regression ofthe the accuracy data of was the of alsodata the performed. wasempirical also performed. Althoughequation Althoughobtained. Forthe statisticalvalue of R completeness,2 = 0.94 was equal a linear to theregression non-linear of the regression, data was thealso value performed. of the Although the value of R2 = 0.94 was equal to the non-linear regression, the value of the

Environments 2021, 8, 37 8 of 12 Environments 2021, 8, x FOR PEER REVIEW 8 of 12

the value of R2 = 0.94 was equal to the non-linear regression, the value of the RMSE (0.97) wasRMSE slightly (0.97) higherwas slightly than the higher non-linear than the regression. non-linear This regression. statistical This difference statistical inﬂuenced difference the influencedchoice of the the regression choice of typologythe regression to obtain typology the empirical to obtain equation. the empirical equation. Table2 2 shows shows thethe airair temperaturetemperature valuesvalues obtainedobtained fromfrom thethe morningmorning andand afternoonafternoon thermal imaging survey. Sandy-soil surface te temperaturemperature values was obtained by usingusing these valuesvalues in in the the Equation Equation (2). (2). Moreover, Moreover taking, taking into account into account the thermal the camerathermal accuracy, camera accuracy,the thermal the range thermal of the range sandy of the soil sandy surface soil is shown.surface is shown.

Table 2. Temperature datadata (30–31(30–31 OctoberOctober 2018).2018).

Potential ThermalPotential Range Thermal of Sandy Range Soil of Sandy Surface Soil (for Air Temp. Sandy Soil Surface Temp. (°C) Predicted Sandy Soil Surface Temp.Each (◦ C)IR Image)Surface Considering (for Each the IR Image)Previous Considering Column and (°C)Air Observed Temp. (◦C) Observedby Applying (2) Predicted by Applying (2) the PreviousInstrumental Column Accuracy and the (±2 Instrumental °C) ◦ 08:00 AM UTC 16.5 14.8 12.8–16.8Accuracy (±2 C) 09:0008:00 AM AM UTC UTC 17.1 16.5 15.5 14.8 13.5–17.5 12.8–16.8 12:0009:00 PM AM UTC UTC 24.8 17.1 20.5 15.5 18.5–22.5 13.5–17.5 02:3012:00 PM PM UTC UTC 19.5 24.8 18 20.5 16–20 18.5–22.5 05:3002:30 AM PM UTC UTC 14.4 19.5 12.6 18 10.6–14.6 16–20 01:0005:30 PM AM UTC UTC 23.9 14.4 22.8 12.6 20.8–24.8 10.6–14.6 01:00 PM UTC 23.9 22.8 20.8–24.8 Figure 8 shows the thermal images collected in the morning of 30 October 2018 and 31 OctoberFigure 82018, shows processed the thermal using images FLIR collectedTools. Table in the 3 shows morning the of correspondence 30 October 2018 of and the material31 October detected 2018, processedin the thermal using image FLIR (predi Tools.ction) Table compared3 shows thewith correspondence the RGB image of(truth) the inmaterial the randomly detected selected in the thermal points. image (prediction) compared with the RGB image (truth) in the randomly selected points.

(a)

(b)

Figure 8. Cont.

Environments 2021, 8, x FOR PEER REVIEW 9 of 12

Environments 2021, 8, x FOR PEER REVIEW 9 of 12 Environments 2021, 8, 37 9 of 12

(c) Figure 8. Morning acquisition: thermal image (on the left) and RGB image (on the right). Randomly selected points (in red). The yellow pixels in the thermal image indicate the sandy soil surface identified by the application of the empirical Equation (2). Plastic bottles can be seen in (a), rubber soles in (b), glass bottles (andc) expanded polystyrene in (c).

Figure 8. Morning acquisition:TableFigure thermal 3. 8. Correspondence Morning image acquisition: (on the between left) thermal and the RGB material image image (on detected (on the the left) right).in and the RGB thermal Randomly image image (on selected theand right). the points RGB (in image Randomly selected points (in red). The yellow pixels in the thermal image indicate the sandy soil red). The yellow pixels in thein the thermal selected image points indicate (morning). the sandy soil surface identiﬁed by the application of the empirical surface identified by the application of the empirical Equation (2). Plastic bottles can be seen in (a), Equation (2). Plastic bottles can be seen in (a), rubber soles in (b), glass bottles and expanded polystyrene in (c). rubber soles in (b), glass bottles and expanded polystyreneMaterial in ( cin). the RGB Image (Truth) Natural and/or Table 3. CorrespondenceTable between 3. Correspondence the material detected between in the the material thermal detected image andin the the thermal RGB imageimage and in the theselected RGB image Sandy Soil Anthropic Total points (morning). in the selected points (morning). Anomaly

Sandy soil MaterialMaterial 15 in the in RGB the RGB Image 1 Image (Truth) (Truth) 16 Material detected in Natural and/orNatural Anthropic and/or Natural and/orSandy Soil Total the thermal image 0 Anomaly 14 14 anthropic anomaly Sandy Soil Anthropic Total (prediction) Material detected in the Sandy soilTotal 1515 1Anomaly 15 16 30 thermal image Natural and/or anthropic anomalySandy soil 0 15 14 1 14 16 (prediction) Material detectedTotal in 15 15 30 Natural and/or theThe thermal kappa image coefficient value was calculated0 an d resulted to be14 equal to 0.93. 14Figure 9 anthropic anomaly shows(prediction) the thermal images collected in the afternoon of 30 October 2018 and 31 October 2018 andThe kappaprocessed. coefﬁcient Table value4 Totalshows was the calculated corresp15ondence and resulted between to be 15 of equal the material to 0.93. Figure 30detected 9 inshows the thermal the thermal image images (prediction) collected compared in the afternoon with the of RGB 30 October image 2018(truth) and in 31 30 October random 2018 and processed. Table4 shows the correspondence between of the material detected in points.The kappa coefficient value was calculated and resulted to be equal to 0.93. Figure 9 showsthe thermal the thermal image (prediction) images collected compared in the with afternoon the RGB of image 30 October (truth) 2018 in 30 and random 31 October points. 2018 and processed. Table 4 shows the correspondence between of the material detected in the thermal image (prediction) compared with the RGB image (truth) in 30 random points.

(a)

Figure 9. Cont.

(a)

Environments 2021, 8, x FOR PEER REVIEW 10 of 12 Environments 2021, 8, 37 10 of 12

(b)

(c)

Figure 9. Afternoon acquisition:Figure thermal 9. Afternoon image acquisition: (on the left) thermal and RGB image image (on (on the the left) right). and Randomly RGB image selected (on the points right). (in red). The yellow pixels in theRandomly thermal imageselected represent points (in the red). sandy The soil yellow surface pixe identiﬁedls in the by thermal the application image represent of the empirical the sandy soil surface identified by the application of the empirical Equation (2). Plastic bottles and bags can be Equation (2). Plastic bottles and bags can be seen in (a,c) and rusty bottles in (b). seen in (a and c) and rusty bottles in (b).

Table 4. Correspondence between the material detected in the thermal image and the RGB image in the selected Table 4. Correspondence between the material detected in the thermal image and the RGB image points (afternoon). in the selected points (afternoon).

MaterialMaterial in the in RGB the ImageRGB (Truth)Image (Truth) Natural and/or Anthropic Sandy Soil Natural and/or Total Anomaly Sandy Soil Anthropic Total Material detected in the Sandy soil 13 1Anomaly 14 thermal image Natural and/or anthropic anomaly 3 13 16 Sandy soil 13 1 14 (prediction) Material detectedTotal in 16 14 30 Natural and/or the thermal image 3 13 16 anthropic anomaly (prediction)The kappa coefficient value was calculated and resulted to be equal to 0.74. The results Total 16 14 30 obtained from empirical equation application in the thermal images, the comparison between the material detected in the thermal images and the values of the kappa coefficient, allowedThe forkappa validation coefficient of the value thermal wa methodologys calculated used,and resulted which was to developedbe equal to to 0.74. identify The resultssandy soilobtained on the coastfrom usingempirical IR thermography, equation application distinguishing in itthe from thermal thermal images, anomalies the comparisonwhich can be between due to the the presence material of detected different in materials the thermal and pollutingimages and objects the andvalues material of the kappaof anthropogenic coefficient, originallowed such for as validation plastic. The of methodologythe thermal methodology described may used, have which problems was developedwith shaded to areas, identify because sandy they soil are on significantly the coast using cooler IR than thermography, the predicted soildistinguishing temperature it fromvalue. thermal However, anomalies using UAV which systems can with be due larger to scale the acquisitions,presence of thedifferent shaded materials areas would and pollutinghave less objects weight inand the material acquired of thermograms anthropogeni thanc origin the reallysuch as radiated plastic. beach The methodology areas.

Environments 2021, 8, 37 11 of 12

4. Conclusions This study represents an application of the IR thermography technology in the field of coastal environmental monitoring. The 70-day monitoring of the air and sandy soil surface temperature by means of two thermal sensors, carried out in the Coastal Dunes Regional Natural Park (Ostuni–Fasano), and the statistical analysis of a considerable amount of thermal data, allowed the development of an empirical equation used for calculating the sandy-soil surface temperature by knowing only the air temperature. The existence of a strong correlation between the two variables, the high value of the coefficient of determination R2 of the model and a low value of the RMSE value, confirmed the good quality of the empirical equation used. Moreover, the tests carried out applying the equation to the IR thermal survey and the value of the kappa coefficient calculated, validated this methodology, making it an effective tool for anthropogenic polluting material detection (plastic, glass, rubber) on the sandy coast in the study area. The results obtained can be used to rapidly process thermal images deriving from surveys carried out with sensors installed on UAV and UGV, quickly detecting the presence of anomalies due to potentially polluting objects (like plastic, glass, etc) or other pollutant materials. Further coastal environmental geology studies will be carried out in the future. In particular, remote multispectral methodologies will be tested in this study area for a more precise identification of pollutants.

Author Contributions: Conceptualization (C.C.), Data curation (C.C.), Formal analysis (C.C.; E.P. and S.R.), Founding acquisition (S.R. and G.M.), Investigation (C.C. and E.P.), Methodology (C.C.), Project administration (G.M.), Resources (S.R. and G.M.), Software (C.C.), Supervision (H.R.; S.R. and G.M.), Validation (C.C.; E.P.; S.R. and G.M.), Visualization (C.C. and E.P.), Writing—original draft (C.C. and E.P.), Writing—review and editing (C.C.; E.P.; S.R. and G.M.). All authors have read and agreed to the published version of the manuscript. Funding: Contributo Protezione Civile Regione Puglia–Progetto I-STORMS; Scientific Supervisor: Giuseppe Mastronuzzi. Institutional Review Board Statement: Not applicable. Informed Consent Statement: Not applicable. Data Availability Statement: The data presented in this study are available on request from the corresponding author. Acknowledgments: The authors would like to thank the reviewers for their contributions and constructive comments for the improvement of the manuscript and Regional Natural Park of the Coastal Dunes for their kind support. Work carried out as part of the research project “SIAT— Integrated UAV–UTV system with remote control for the pollutants dispersion assessment in the coastal environment for remediation” related to the XXXIII PON Cycle of the Ph.D. in Geosciences of the University of Bari, under the responsibility of Massimo Moretti. Conflicts of Interest: The authors declare no conflict of interest.

References 1. Bianchi, F.; Pisello, A.L.; Baldinelli, G.; Asdrubali, F. Infrared Thermography Assessment of Thermal Bridges in Building Envelope: Experimental Validation in a Test Room Setup. Sustainability 2014, 6, 7107–7120. [CrossRef] 2. Costanzo, A.; Minasi, M.; Casula, G.; Musacchio, M.; Buongiorno, M.F. Combined Use of Terrestrial Laser Scanning and IR Thermography Applied to a Historical Building. Sensors 2015, 15, 194–213. [CrossRef][PubMed] 3. Garduño-Ramón, M.A.; Vega-Mancilla, S.G.; Morales-Henández, L.A.; Osornio-Rios, R.A. Supportive Noninvasive Tool for the Diagnosis of Breast Cancer Using a Thermographic Camera as Sensor. Sensors 2017, 17, 497. [CrossRef][PubMed] 4. Kim, J. Non-Destructive Characterization of Railway Materials and Components with Infrared Thermography Technique. Materials 2019, 12, 4077. [CrossRef][PubMed] 5. Aranda, J.M.; Melendez, J.; De Castro, A.J.; Lopez, F. Forest ﬁre studies by medium infrared and thermal infrared thermography. In Proceedings of the Thermosense XXIII, Orlando, FL, USA, 23 March 2001; Volume 4360. Environments 2021, 8, 37 12 of 12

6. Cagnazzo, C.; Potente, E.; Rosato, S.; Mastronuzzi, G. Geostatistics and Structure from Motion Techniques for Coastal Pollution Assessment along the Policoro Coast (Southern Italy). Geosciences 2020, 10, 28. [CrossRef] 7. Cagnazzo, C.; Potente, E.; Rosato, S.; Mastronuzzi, G. Numerical modeling for coastal environmental monitoring: Oil spills simulation along the lower Adriatic coasts. Rend. Online Soc. Geol. It. 2020, 52, 19–24. [CrossRef] 8. Manfreda, S.; McCabe, M.F.; Miller, P.E.; Lucas, R.; Pajuelo Madrigal, V.; 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] 9. Stuart, M.B.; McGonigle, A.J.S.; Willmott, J.R. Hyperspectral Imaging in Environmental Monitoring: A Review of Recent Developments and Technological Advances in Compact Field Deployable Systems. Sensors 2019, 19, 3071. [CrossRef][PubMed] 10. Zanutta, A.; Lambertini, A.; Vittuari, L. UAV Photogrammetry and Ground Surveys as a Mapping Tool for Quickly Monitoring Shoreline and Beach Changes. J. Mar. Sci. Eng. 2020, 8, 52. [CrossRef] 11. Casana, J.; Kantner, J.; Wiewel, A.; Cothren, J. Archaeological aerial thermography: A case study at the Chaco-era Blue J community, New Mexico. J. Archaeol. Sci. 2014, 45, 207–219. [CrossRef] 12. Lega, M.; Kosmatka, J.; Ferrara, C.; Russo, F.; Napoli, R.M.A.; Persechino, G. Using Advanced Aerial Platforms and Infrared Thermography to Track Environmental Contamination. Environ. Forensics J. 2012, 13, 332–338. [CrossRef] 13. Lagüela, S.; Díaz-Vilariño, L.; Roca, D.; Lorenzo, H. Aerial thermography from low-cost UAV for the generation of thermographic digital terrain models. Opto-Electron. Rev. 2015, 23, 78–84. [CrossRef] 14. Ficapal, A.; Mutis, I. Framework for the Detection, Diagnosis, and Evaluation of Thermal Bridges Using Infrared Thermography and Unmanned Aerial Vehicles. Buildings 2019, 9, 179. [CrossRef] 15. Savelyev, A.; Sugumaran, R. Surface Temperature Mapping of the University of Northern Iowa Campus Using High Resolution Thermal Infrared Aerial Imageries. Sensors 2008, 8, 5055–5068. [CrossRef][PubMed] 16. Szrek, J.; Wodecki, J.; Błazej,˙ R.; Zimroz, R. An Inspection Robot for Belt Conveyor Maintenance in Underground Mine—Infrared Thermography for Overheated Idlers Detection. Appl. Sci. 2020, 10, 4984. [CrossRef] 17. Ciola, G.; Maiorano, F.; Massari, M.A. Il Parco Naturale Regionale delle Dune Costiere da Torre Canne a Torre San Leonardo: Il valore della biodiversità per ricostruire comunità solidali. Territori e comunità. In Le sﬁde dell’autogoverno comunitario. Atti dei Laboratori del VI Convegno della Società dei Territorialisti Castel del Monte (BA); Gisotti, M.R., Rossi, M., Eds.; Sdt Edizioni: Bari, Italy, 2018. 18. Perrino, E.V.; Ladisa, G.; Calabrese, G. Flora and plant genetic resources of ancient olive groves of Apulia (southern Italy). Genet. Resour. Crop Evol. 2014, 61, 23–53. [CrossRef] 19. Tomaselli, V.; Tenerelli, P.; Sciandrello, S. Mapping and quantifying habitat fragmentation in small coastal areas: A case study of three protected wetlands in Apulia (Italy). Monit. Assess. 2012, 184, 693–713. [CrossRef] 20. Mastronuzzi, G.; Palmentola, G.; Sanso, P. Evoluzione morfologica della fascia costiera di Torre Canne (Puglia adriatica). Studi Costieri 2001, 4, 19–31. 21. Mastronuzzi, G.; Sanso, P. Holocene coastal dune development and environmental changes in Apulia (Southern Italy). Sediment. Geol. 2002, 150, 139–152. [CrossRef] 22. Moretti, M.; Tropeano, M.; Van Loon, A.J.; Acquafredda, P.; Baldacconi, R.; Festa, V.; Lisco, S.; Mastronuzzi, G.; Moretti, V.; Scotti, R. Texture and composition of the Rosa Marina beach sands (Adriatic coast, southern Italy): A sedimentological/ecological approach. Geologos 2016, 22, 87–103. [CrossRef] 23. Marsico, A.; Caldara, M.; Capolongo, D.; Pennetta, L. Climatic characteristics of middle-southern Apulia (southern Italy). J. Maps 2007, 3, 342–348. [CrossRef] 24. Mastronuzzi, G.; Palmentola, G.; Sansò, P. Lineamenti e dinamica della costa pugliese. Studi Costieri 2002, 5, 9–22. 25. Cohen, J. A coefﬁcient of agreement for nominal scales. Educ. Psychol. Meas. 1960, 20, 37–46. [CrossRef] 26. Cohen, J. Weighed kappa: Nominal scale agreement with provision for scaled disagreement or partial credit. Psychol. Bulletin. 1968, 70, 213–220. [CrossRef][PubMed]