Ann. Limnol. - Int. J. Lim. 2020, 56, 6© EDP Sciences, 2020https://doi.org/10.1051/limn/2020004

Available online at:www.limnology-journal.org

RESEARCH ARTICLE

Monitoring spatiotemporal variability of water quality parametersUsing Landsat imagery in Choghakhor International Wetlandduring the last 32 years

Ahmad Reza Pirali Zefrehei1,*, Aliakbar Hedayati1, Saeid Pourmanafi2, Omid Beyraghdar Kashkooli2

and Rasoul Ghorbani1

1 Faculty of Fisheries and Environmental Sciences, Gorgan University of Agricultural Sciences and Natural Resources, Gorgan, Iran2 Department of Natural Resources, Isfahan University of Technology, Isfahan, Iran

Received: 15 January 2020; Accepted: 5 March 2020

*Correspon

Abstract – Use of Landsat is of importance in monitoring and assessment of long-term changes of waterquality in freshwater ecosystems, especially in small water bodies. In this study, over a 32-year period(1985–2017), the changes in water surface temperature (WST), secchi disk transparency (SDT), andchlorophyll-a (Chl-a) concentration were estimated at the Choghakhor wetland using Landsat imagery.Based on WST three detectable temperature zones can be observed within the wetland aquatic environmentwhere the highest amount was observed in thermal strips. The results showed Chl-a concentration volatilityin different periods in the wetland as well as its long-term increasing trend. The western part of the wetland,as compared to other areas, was affected by these changes, which could be due to the human activityconcentrated in this area. In contrast (SDT) showed a decreasing trend during this period that was consistentwith the observed changes in Chl-a concentration. This could be due to an increase in organic matter loadand suspended solids in the water body of wetland during this time. Comparison of the extracted satellitedata with the field data showed the least RMSE and high R2. Also, ANOVA results showed significantspatio-temporal differences between the studied parameters in Choghakhor wetland (p< 0.05). The presentstudy can help to detect long-term changes in Choghakhor wetland and help toward moving to optimalmanagement and protection of this wetland.

Keywords: Choghakhor International Wetland / chlorophyll-a / landsat imagery / spatio-temporal variations /water quality

1 Introduction

Wetlands play a key role in food production, water and airpurification, carbon storage and even changing local climate,affecting food turnover rate, prevention of flooding and overallprotection of the biodiversity in local freshwater ecosystems(Chen et al., 2011; Ryan et al., 2012; Keddy, 2010).Conventional measurements and monitoring of water qualityinvolve in situ sampling, which is a costly and time consuming(Ogilvie et al., 2018; El Masri et al., 2008). In addition, fortracking methods in the field, some areas are often inaccessibleor difficult to sample (Chen and Quan, 2012). Due to theprevalent logistic limitations, it is not possible to cover theentire area of water or frequent sampling at a site (Chen andQuan, 2012). Overall in situ sampling encountered with someproblems which inhibit coherent monitoring and estimation of

ding author: ahmadreza.pirali@gmail.com

water quality (Senay et al., 2001; El Masri and Rahman, 2008).In contrast, remote sensing techniques can overcome theseconstraints by providing alternative methods for monitoringwater quality across a wide range of time and spatial scales(Senay et al., 2001; El Masri and Rahman, 2008). Remotesensing images can pass even across inland water and hence,confirming in situ measured parameters (Chen and Quan,2012). Remote sensing analyzes the radiation data by a sensorin a specific area (here the wetland of water body). Achievedinformation, such as water transparency and chlorophyll-a(Chl-a) is in fact radiance within the visible and is near-infrared(Dörnhöfer and Oppelt, 2016). For example, Chl-a concentra-tion can be measured at spectral bands of 440–560 nm and670 nm (Matthews, 2011). In water column, the opticalproperties of water body such as suspended particulate matter(SPM) and Chl-a, may alter the radiation by absorption andscatter (Odermatt et al., 2012). For measuring water surfacetemperature (WST), sensors are needed that measure radiationat thermal infrared (13–17mm) (Dörnhöfer and Oppelt, 2016).

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

High resolution satellites such as Landsat are the available andefficient options for satellite-based measurements to monitorwater quality even in small waters bodies (Ogilvie et al., 2018;Brezonik et al., 2005). The Landsat data series are ideal for thispurpose based on a 30-meter spatial resolution, especially forsome of the relatively small waters bodies (Lee et al., 2016;Brezonik et al., 2005). Landsat is freely available and containsa collection of archived data that can provide us with insightsinto historical events or developments in a region andtherefore, can be used for monitoring purposes (Chen andQuan, 2012). Of course, estimation of qualitative parameterswith Landsat has various limitations. The most prominentlimitation for investigated properties within water is possess-ing inherent optical properties (IOPs) which can be measuredthrough satellite sensors (Brezonik et al., 2005). Landsatmeasures the radiance with the sensor and cannot be calibratedwith the intensity of the solar radiation, which usually varieswith factors such as latitude, day length, and season. Here theatmospheric interference, which can influence reflectedradiation level, is important (Brezonik et al., 2005).

In this study, we investigated the remote sensing of WST,Chl-a concentration and secchi disk transparency (SDT)parameters. WST is one of the most important abiotic factors indetermining environmental changes and ecological activitiesof the water body (Kang et al., 2014). The physical, biologicaland chemical processes depend on temperature. For instance,the temperature affects the dissolve oxygen content of water,the metabolic rate of aquatic organisms (Wetzel, 2001). Chl-ais a well-known indicator of ecological health in an aquaticenvironment that is widely used to represent water quality andtrophic status (Sun et al., 2014). Since phytoplankton containschlorophyll, its biomass is detectable by optical sensors; hence,Chl-a is an ecological indicator for the study of the impact ofnutrients and the ecosystem’s status in the water body (Shutleret al., 2007). SDT is one of the parameters that can show thelevel of opacity in the water. It can also be used in theevaluation of the eutrophication characteristics of water bodywith other parameters (Yüzügüllü and Aksoy, 2011). Forinland water ecosystems, SDT has been widely measured byremote sensing (e.g., Olmanson et al., 2008; Greb et al., 2009).

Permanent protection and monitoring of wetland changes,which are among the national natural assets of each country,are among the needs of sustainable development, andidentifying long-term wetland changes needs to be temporallyanalyzed using holistic approaches such as remote sensing.That’s why the present study aims to investigate and monitortemporal and spatial variations of water quality parameterssuch as WST, Chl-a and SDT at Choghakhor Internationalwetland using Landsat images in 1985–2017and the resultshave been evaluated.

2 Materials and method

2.1 Study area

The Choghakhor international wetland, as one of the mostimportant structural elements in the landscape of the regionand important bird area (IBA), is located in Chaharmahal andBakhtiari province, Borujen city and Boldaji district in thecentral Zagros Mountains, the basin of the Karun River. Thiswetland is located between 50° 520 to 50° 560E and 31° 540 to

Page 2 o

31° 560N (Fig. 1). The Choghakhor wetland as one of 23Ramsar sites in Iran with an area of 1600 ha is one of the mostimportant sites of Iran in terms of presence of endemicendangered fish species (Aphanius vladykovi) and the type ofwetland class is Lacustrine based on the Ramsar Convention(Behrouzirad, 2007; RSIS, 2010; Ebrahimi and Moshari,2006).

2.2 Satellite data acquisition and process

Landsat contains a large number of satellite images archiveof the studied area (path /row: 164/38) from 1985 to 2017. Inthis study, Landsat 5, 7 and 8 satellite images (TM, ETMþ andOLI sensors) were used. These images were selected based onavailable data from Landsat satellite archives. All images havebeen tried on a specific date (similar environmental conditions)to obtain better results. Most of the images are from May andOctober (spring and autumn, respectively). Eventually, only 18Landsat scenes (Tab. 1) were selected. The acquired imageswere Level-1 and obtained from earthexplorer.usgs.gov.

In this study, geometric and radiometric correction wasperformed on the required images. After studying of differentmethods, atmospheric correction was used by DOS (DarkObject Subtraction) method. Also, based on several studies(Hicks et al., 2013; Patra et al., 2016; Bonansea et al., 2015;Urbanskia et al., 2016), this method was found to be suitablefor atmospheric correction in aquatic ecosystems. DOSsearches for the darkest pixel value in any band (dark objectsreflect no light). This method assumes that nonzero values forwater bodies are due to atmospheric path radiance. Scatteringis eliminated by subtracting this value from each pixel in theband (Patra et al., 2016; Bonansea et al., 2015).

2.3 Field data collection

Due to the use of satellite image extraction data in thisstudy, field data need to be validated. For this purpose, the dataofWST, SDT, chlorophyll-a concentration on the wetland werecollected from 2006 (Ebrahimi, 2006) and 2010 (Esmaeili,2012). All parameters of this dataset were extracted accordingto the standard method (APHA-AWWA-WEF, 2000). Figure 1shows the position of the points on the map. Also based on theGPS coordinates of these points, the parameters were extractedin other years. For better comparison and proper evaluation,the data from the images were divided into two spring andautumn seasons (dry and wet) and three geographically pointswest, east and center.

2.4 Water discrimination

The surface of Choghakhor wetland water area wasdetermined through using the normalized differential waterindex (NDWI) and the proposed algorithm by McFeeters(1996) to identify the water body of wetland and separatingunrelated pixels (Alcantara et al., 2010). Ji et al (2009)suggested that NDWI can be successfully used to define andisolate the water bodies and monitoring the water rangechanges. According to McFeeters (1996), the threshold valuefor NDWI is set to zero. The NDWI is derived using green and

f 13

Fig. 1. Map and points of dataset in Choghakhor Wetland, Iran. The triangles represent dam.

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

near infrared (NIR) bands, Eq.1:

NDWI ¼ GREEN � NIRGREEN þ NIR

: ð1Þ

2.5 Water quality algorithm

After considering different algorithms studies, the follow-ing were selected because of suitable for small water bodies.

2.5.1 Chlorophyll-a concentration (Chl-a) and transparency(SDT) algorithms

The following relations which is suitable for small waterbodies, were used to calculate Chl-a and transparency

Page 3 o

(measured as SDT) (Chao Rodríguez et al., 2014):

Chl a½ � ¼ 0:0979e205926�RGRB ð2Þ

SDT ¼ �10:281RG� RBRGþ RB

þ 4:5753 ð3Þ

where Chl-a is the chlorophyll-a concentration measured inmilligrams per cubic meter (mg/m3), and SDT is the secchi disctransparency in meters.

RG and RB are the water reflectance measures in the greenand blue bands, respectively.

2.5.2 Water surface temperature

Landsat thermal band was used to calculate WST. For thispurpose, the digital number (DN) of thermal band must be

f 13

Table 2. TM and ETMþ and OLI thermal band calibration constants.

Constant K1 K2

Units W/(m2� sr� mm) K

TM 607.76 1260.56ETMþ 669.09 1282.71OLI 774.89 1321.08

Table 1. 18 Landsat scenes used in this analysis.

Row ID Image Sensor Date

1 LT05_L1TP_164038_19850802 TM 1985/8/2

2 LT05_L1TP_164038_19850412 TM 1985/4/123 LT05_L1TP_164038_19870504 TM 1987/5/44 LT05_L1TP_164038_19871011 TM 1987/10/115 LT05_L1GS_164038_19951017 TM 1995/10/176 LT05_L1TP_164038_19950526 TM 1995/5/267 LE07_L1TP_164038_20000515 ETM 2000/5/158 LE07_L1TP_164038_20001006 ETM 2000/10/69 LE07_L1TP_164038_20060414 ETM 2006/4/1410 LE07_L1TP_164038_20061007 ETM 2006/10/711 LE07_L1TP_164038_20101002 ETM 2010/10/212 LE07_L1TP_164038_20100527 ETM 2010/5/2713 LC08_L1TP_164038_20131221 OLI 2013/12/2114 LC08_L1TP_164038_20130511 OLI 2013/5/1115 LC08_L1TP_164038_20150517 OLI 2015/05/1716 LC08_L1TP_164038_20151125 OLI 2015/11/2517 LC08_L1TP_164038_20170522 OLI 2017/5/2218 LC08_L1TP_164038_20171114 OLI 2017/11/14

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

converted to the brightness temperature. Studies in otheraquatic ecosystems have also derived from this relationship(Bonansea et al., 2015; Syariza et al., 2015; Chao Rodríguezet al., 2014). Brightness temperature is calculated from sensordata using Planck’s equation calibrated for thermal infraredband (according to Chander and Markham, 2003):

T ¼ K2

LnK1

Llþ 1

� � ð4Þ

where T is temperature measured in Kelvin, Lʎ is the thermalband spectral radiance in watts/(m2� sr� mm), and K1 and K2are Planck’s equation coefficients which given in Table 2 foreach sensor. Finally, to convert the temperature to Celsius, thenumber was reduced to 273.15.

In the case of equation (4), it should be noted that theemission coefficient for water is near to 1 and can be used toestimate water surface temperature and compare it to thedataset (Chao Rodríguez et al., 2014).

2.6 Accuracy assessment

In order to validate the data obtained from the algorithmswith field data (in situ), root mean square error (RMSE) and

Page 4 o

coefficient of determination (R2) was also used (Eqs. (5) and(6)), in addition to drawing a scatter plot (Odermatt et al.,2012).

RMSE ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiPN

i¼1 xest;i � xmeas;i� �2

N

sð5Þ

where, xest are estimated water quality parameters of satellite;xmeas are measured water quality parameters of dataset; and Nis the number of samples.

R2 ¼ ðnP yð Þ � PxÞðP yð Þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

nðP x2Þ � ðP xÞ2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffinðP y2Þ � ðP yÞ2

qr0BB@

1CCA

2

; ð6Þ

where x is the estimated water quality parameters of satellite,while, y is the measured water quality parameters of datasetand n is the number of samples.

2.7 Data analyses

After extraction of parameters, normality of data wasevaluated by Kolmogorov–Smirnov test (p values > 0.05showed normality of data). The One-way ANOVA test wasused to determine any significant differences among meanestimated values of water quality parameters between the years1985 to 2017. Subsequently, the Duncan multiple range testwas performed if significant differences were found inANOVA. Differences were considered significant at p values< .05. All statistical analyses were performed using thestatistical package from SPSS Inc., released 2007 (SPSS forWindows, Version 16.0, and Chicago, SPSS Inc.). All datawere reported as mean ± SE (standard error).

3 Results

As mentioned in the previous section, the method used foratmospheric correction was well suited and similar to theresults of other studies in aquatic ecosystems (Hicks et al.,2013; Patra et al., 2016; Bonansea et al., 2015; Urbanskiaet al., 2016).

After generating the maps of each parameter, we used thescatter plots and the RMSE and R2 metrics to validate andcalibrate the data and compare it with the dataset. The resultsare presented in Figure 2. Accordingly, all the appliedalgorithms had suitable validation measurement. For WST, theRMSE of the satellite data and field data provides an error of4.2 °C and the value of determination coefficient (R2) with amaximum of 0.78 (y= 0.2024xþ 3.57). The determinationcoefficient indicates a positive relationship between thevariables used. Satellite measurements WST comparing tofield measurements typically showed below-surface temper-atures. Although there were no significant differences betweensatellite-based and field measurements, these differences mayoccur due to wind and local conditions of water. For Chl-aconcentration, the amount of determination coefficient (R2) isindicative of a high relation (0.83) among the used variables.The RMSE of the satellite data and field data provides an error

f 13

Fig. 2. Validation of Landsat estimated versus measured parameters (in situ) with 1:1 fit line, (a) WST, (b) SDT and (c) Chl-a.

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

of 0.47mg/m3 (y= 0.125xþ 0.521). For SDT, the amount ofdetermination coefficient (R2) is indicative of a high relation(0.69) among the used variables. The RMSE of the satellitedata and field data provides an error of 0.38m (y= 0.1603xþ 1.5391). The results of measured parameters variation inChoghakhor wetland from 1985 to 2017 is shown fromFigures 3–8. It should be noted that due to wetland drying andturbidity in the autumn of 2017, which led to dehydration, theparameter values were not reported in graph (Figs. 4, 6 and 8)at this time. Furthermore, due to the lack of available images inthe autumn of 1985, the nearest month (August) was chosen asthe course.

3.1 Water surface temperature (WST)

Figure 3 imply the spatio-temporal distribution of WSTjust for spring and autumn seasons of Choghakhor wetland.WST was calculated from Landsat thermal band converted todegrees Celsius. As expected, the WST map shows a spatial-temporal pattern that matches the local conditions. Thewarmest water of sampling time observed in high airtemperature (AT) (springs) and in the north and west part ofthe wetland. The coldest water of sampling time found in lowAT (autumns) accompanying with the center and south part ofthe wetland. In general, three detectable temperature zones can

Page 5 o

be observed within the wetland during studied years: lessvariable areas (blue in the map), moderate change areas, areasnear the shoreline (red in the maps), which are called thermalstrips (Sima et al., 2013). An overall view at zonationmaps (Fig. 3) of the WST within study period showed WSTfluctuations from 1985 to1987 (based on changes indistribution and spatial distribution on maps) were higher.Between 1995 and 2000, the wetland WSTwas in a relativelystable condition, which could be due to the high water volumefollowing the construction of the dam in output of the wetlandin the early 1990’s and increasing rainfall during this year.During the years from 2006 to 2017, a gradual increase in thewetland of WST based on the graph was revealed (rising to amaximum of 38.5 °C in spring 2015 (Fig. 3). However, thedifference of thermal band spatial resolution between differentLandsat sensors in the study period (such as ETMþ versus TMresolution of thermal band) can also be considered. WSTchanges become more pronounced in recent years, which canaffect water volume wetland as well as its margin fluctuations.This fact was observed at 38.5 °C on the margin of the wetlandknown as thermal string. Similar studies on other aquaticecosystems of Iran have also found that WST range. Forinstance, in the study of Lake Urmia (2007–2010) by MODISnoted the maximum WST as high as 30 °C (Sima et al., 2013).The analysis of the WST graphs shows the spatial variation indifferent parts of the wetland (Fig. 4), in general, spring of

f 13

Fig. 3. Choghakhor wetland of WST (°C) changes in spring(s) and autumn (a) Season, broken down annually.

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

2015 and 2017 have a higher WST than the same timeperiod (27.48–30.04 °C, Fig. 4). Also, WST fluctuations arehigher at the west of the wetland. The lowest WST rangewas observed in autumns 1995, (7.7 °C, Fig. 4) which couldbe due to water volume in these years. ANOVA results

Page 6 o

showed a significant difference between the WST of differentyears in the Choghakhor wetland (p< 0.05), in Table 3. Basedon maps WST (extractive data and its spatial distribution onthe map) ranged from 3.3 to 38.5 °C, with a mean value of20.9 °C.

f 13

0.00

5.00

10.00

15.00

20.00

25.00

30.00

35.00

East Center West

WST

(°C

)

Fig. 4. The comparison of WST (°C) in different parts of the Choghakhor wetland in different years, spring(s) and autumn (a).

Table 3. Statistical comparison of parameters calculated in this study using ANOVA (mean ± SE, n = 50).*

2017 2015 2013 2010 2006 2000 1995 1987 1985 F-value P-value

WST 29.66 ± 0.1a 22.91 ± 0.75b 13.03 ± 1.59d 16.62 ± 0.24c 14.10 ± 0.12cd 14.95 ± 1.48cd 13.53 ± 0.03d 16.49 ± 0.28c 24.45 ± 3.07b 43.42 p < 0.05

Chl-a 1.02 ± 0.52b 1.06 ± 0.12b 0.89 ± 0.02c 0.83 ± 0.03c 1.23 ± 0.03a 1.06 ± 0.008b 0.379 ± 0.003d 0.309 ± 0.04d 0.322 ± 0.007d 166.51 p < 0.05SDT 1.627 ± 0.04d 1.6 ± 0.02d 1.69 ± 0.03c 1.72 ± 0.07c 1.55 ± 0.1e 1.61 ± 0.07d 2.07 ± 0.08b 2.14 ± 0.24a 2.12 ± 0.09a 338.74 p < 0.05

* Different letters indicate statistical significance (Duncan, p < 0.05). F-value is the ratio of the mean squares between the groups to the intra-group squares, which indicates the magnitude of the difference between the groups and the comparison with the error.

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

Also, as expected, the method used to measure WST waswell suited and similar to the results of other studies in aquaticecosystems (Bonansea et al., 2015; Syariza et al., 2015;Chao Rodríguez et al., 2014).

3.2 Chlorophyll-a concentration (Chl-a)

The created images of the zoning of Chl-a concentration inChoghakhor wetland from 1985 to 2017 are shown in Figure 5,separately in spring and autumn seasons. Zoning maps of Chl-ain Choghakhor wetland, showed an increasing trend during1985 to 2017. The highest amount of Chl-a was observed in2006 and 2013 respectively. In general, the eastern and westernparts of the wetland have more fluctuations and higher Chl-aconcentration than the center (Fig. 6, which is more evident inthe western regions. ANOVA results showed a significantdifference between different Chl-a of different years in theChoghakhor wetland (p < 0.05), in Table 3. Based on mapsChl-a presented a mean value of 1.17mg/m3 and relativelyhigher values were found in spring season. The measuredChl-a fluctuations in the Choghakhor Wetland were between0.25 and 2.1mg/m3. The highest Chl-a value was observed inspring season of 2006 (1.31mg/m3).

3.3 Secchi disk transparency (SDT)

The created images of zonation of SDT condition ofChoghakhor wetland from 1985 to 2017 (separately in springand autumn seasons) are presented in Figure 7. The overalldepth of transparency decreased from 1985 to 2017, whichindicates an increase in turbidity and suspended as well as

Page 7 o

organic particles accumulation in the wetland. It confirms theresults of this study related to the Chl-a concentrationincreasing in recent years. Wetland of water contains highnutrient levels, suspended particulate matter and dissolvedorganic matter, some essential elements that cause phyto-plankton growth and turbidity and decrease in SDT andincrease in Chl-a were their consequences. The SDT changesin different parts of the wetland (Fig. 8) shows that the centralareas of the wetland are deeper. This trend is also seen inFigure 7. It suggests higher transparency and less effect offluctuations and wetland activities in this part of the wetland.Comparison of spring and autumn showed that the relativedepth in the autumn seasons is less than those of the springseasons. ANOVA results (Tab. 3) showed a significantdifference among the SDTof different years in the Choghakhorwetland (p < 0.05). The values of SDT (based on changes indistribution and spatial distribution on maps) varied between1.3 and 2.22m, with a mean value of 1.71m. This parameteralso showed a seasonal variation which was related to climaticseasons and rainfall. Thus, the wetland showed the lower SDTin autumn seasons, due to the inlet of high loads of suspendedsolids by the inflows. While higher SDT occurred in springseasons as inflows flows were lower. Landsat prediction,computed using equation (3) in the central pixels of the waterbody, provides values with a similar range.

4 Discussion

Suitable Monitoring of Choghakhor wetland changes isdefinitely essential for preserving this ecosystem, as a valuable

f 13

Fig. 5. Choghakhor wetland of Chl-a (mg/m3) changes in spring(s) and autumn (a) season, broken down annually.

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

natural heritage. Our study showed that remote sensing andprocessing of satellite images can help us to achieve this goaland this procedure could be the first study in the region thatpresents image of the historical changes throughout thewetlands in different periods. We also extracted several critical

Page 8 o

parameters of the aqueous environment using Landsatimagery.

Landsat is widely used to estimate water qualityparameters (Dörnhöfer and Oppelt, 2016). Also, in this study,the spatio-temporal dynamics results of the qualitative

f 13

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

East Center West

Chl

-a (m

g/m

3 )

Fig. 6. The comparison of of Chl-a (mg/m3) in different parts of the Choghakhor wetland in different years, spring(s) and autumn (a).

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

parameters showed the adequate ability of Landsat. This studyused three generations of Landsat (TM, ETMþ and OLI)based on different years. Of course, Landsat 8 has a particularfunction compared to the previous Landsat sensors (Patraet al., 2016). However, Landsat’s weakness is the lowertemporal resolution (16-day) compared to sensors such asMODIS and Sentinel. So, in small water bodies such asChoghakor wetland, the spatial resolution (30m) is asignificant factor in monitoring of these areas. MODIS takesless temporal resolution (1 day) but with a resolution of 250m,that it is not suitable enough for monitoring of these areas. Theunique Landsat feature comparing to other sensors (such as theSentinel with spatial resolution, 10m and 4-year datasets since2014) is a 40-year-old available dataset that allows long-termmonitoring so it was used in the present study.

Data obtained from comparing field samplings withsatellite data showed relatively good performance of RMSE(4.2 °C, 0.47mg/m3 and 0.38m). In a similar study at UrmiaLake, the RMSE of WST by MODIS varied between 2.59 and0.27 °C for different years and conditions (Sima et al., 2013).The RMSE and R2 in Arreo Lake forWSTand SDT by Landsatwere estimated to be 4.18 °C and 0.6m (Chao Rodríguez et al.,2014). Also, the RMSE of WST in the two Bimont andBariousses lakes varied between 1.75 and 2.39 °C for differentconditions (Simon et al., 2014). Simon et al. (2014) found thereasons for these errors as lags in field and satellite data,differences in depth and level of temperature and satellite fielddata, field measurement quality, and field scale differences andthermal band resolution. In study of Kemp Lake of Texas, theR2 of field data and Modis for chlorophyll-a were calculated0.3 and 0.8 in spring (June) and autumn (October), respectively(El Masri and Rahman, 2008). They confirmed that the numberof field samples, changes in chlorophyll-a absorption, less fielddata fluctuation of chlorophyll-a compared to October, aspotential contributors to the weakness of R2 with field andsatellite data in June.

In general, the difference in WST depends on the variousconditions. The WST measured by the satellite covers an area,while field data is a point value that can lead to a meaningfuldifference. Also, an error in the field measurement may occur.However, WST changes are much lower than land surfacetemperature. Another factor is water depth. Based on the water

Page 9 o

thermal properties, wherever the depth is greater, there is ahigher heat storage capacity, causing the WST to increasesignificantly during the warm seasons or decrease during thecold season (Fazelpoor et al., 2015), which in Choghakhorwetland, due to less fluctuations in depth and also less depthrange, is not very effective. In addition, WST has a significantrelationship with the characteristics of the ecosystem and isaffected by environmental and climate change (Adrian et al.,2009). The highest water WST based on the graph in bothseasons (spring and autumn) was observed in the westernregion of Choghakhor wetland (Fig. 4), which is likely to beaffected by input water temperature due to agriculturalactivities, depth and presence of dark-colored aquatic plants.In the case of comparing several parts of the wetland, thehighest WSTwas observed in the coastal areas of the wetland.The WSTof the wetland in the spring season is higher than theautumn season, which is justified by the warmer air temper-atures in the warm seasons. WST in parts with high variationcan be related to wetland margins and water volume changes orthe absence or presence of aquatic plant cover. According tothe satellite images in the autumn seasons of 1985, 2015, and2017, along with low water levels, the increase in WST isevident, and in winter 1995 the lowest WST is observed alongwith the high water status in the wetland. Interpretation theresults of quantitative parameters such as Chl-a needs tounderstand the ecology and optical patterns in the region(Stumpf et al., 2003). In a healthy ecosystem, naturally allecological and biological factors fluctuate due to seasonal andtemporal changes, and the intensity of this fluctuation varieswith respect to the geographical location, extent, depth,dominant flows and shape of the water source. TheChoghakhor wetland is no exception. Considering to thevariations of Chl-a among different seasons in the studiedyears, Chl-a concentration is higher in spring than the autumn,which is similar to the fluctuations of aquatic ecosystems intemperate regions and follows the natural pattern of primaryproduction in these areas. In the study by Esmaeili (2012) inthe Choghakhor Wetland, Chl-a patterns were similar to thosein this study, and this parameter was higher in spring. Althoughit increased slightly in the autumn, but still was lower than thespring (Esmaeili, 2012). In general, Chl-a concentration isdependent on biomass of phytoplankton and its mean

f 13

Fig. 7. Choghakhor wetland of SDT (m) changes in spring(s) and autumn (a) season, broken down annually.

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

concentration in aquatic ecosystems is a function of mean totalphosphorus and presents a log-log relationship. According tothe several studies, there is a positive correlation betweenChl-a concentration and nutrients (Wetzel, 2001; Souchu et al.,2010;Pereiraetal., 2010; Iqbaletal., 2017;Primpasetal., 2010).

Page 10

These changes vary in parameters such as Chl-a and SDT indifferent ecosystems andmay be related to the lack of nitrogen athighphosphorus levels, the impact of organic solid particles, andtheir rate of washing. (Kufel et al., 1997; Carpenter, 2005;Covinoetal., 2009;Zoriasateinetal., 2013).Regarding theyears

of 13

0

0.5

1

1.5

2

2.5

East Center West

SDT

(m)

Fig. 8. The comparison of SDT (m) in different parts of the Choghakhor wetland in different years, spring(s) and autumn (a).

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

when Chl-a concentration was higher in autumn than in spring(1985, 1987, and 2013), environmental conditions in the areashould be studied and environmental factors, especially waterturbulence and water temperature, are usually influenced bywater volume fluctuations, especially in the years 1985–1987(years before dam construction). However, further study ofaquatic ecosystems shows that the relationship of Chl-a withphosphorus and transparency in sediment reservoirs is moreunstable thansimilar lakesandecosystems(Boyntonetal., 1982;Filstrup and Downing, 2017; Murrell et al., 2007). Stumpf et al.(2003) found that Chl-a concentration and algae density in theeast of the Gulf of Mexico increased from late summer to earlywinter. SDT changes can be observed in relation to Chl-afluctuations and these changes must be evaluated from twoaspects: (i) Inflows into the wetland (organic and inorganicsuspensions and associated particles), as observed in thewetland center at most of the years it was more transparent,(ii)Densityofphytoplankton (basedonChl-a concentration) thatincreases or decreases as a result of transparency changes.However, the results of this study showed that SDT was moreinfluenced by the second factor (phytoplankton density).Residential and agricultural uses of the Choghakhor wetlandare dispersed in the southern part of the wetland, where thewaste water is transferred from the sources into the wetland. Inthe northern half of the wetland, the use of pollutants orwastewater and the source of input that is derived from it, is verylow and almost contained bare land (Samadi, 2015). Therefore,the amount of nutrients in this area of Choghakhor wetland canbe attributed to the reasons for the increase of Chl-aconcentration. Also, in autumn, with the lowest degree oftransparency, the highest density of phytoplankton and Chl-awas observed. It seems that the existing parameters areinsufficient to investigate the dominant relationships withinwetland ecosystem, and some factors make the analysis difficultto complete, such as: (1) highly dependent variables of thestudied variables (nutrients, chlorophyll and phytoplankton,transparency (SDT)), (2) Seasonal changes of nutrients inphytoplankton densities, 3) difficulty in separating extrinsicfactors (phytoplankton and nutrients are instinct factors ofaquatic ecosystems structures), 4) Nutrition cycle dynamicand identifying factors resulting from human activities shouldbe added.

Page 11

5 Conclusions

Our method helps to have a rapid detection of ecologicalchanges in the wetland by using some environmentalindicators (through extraction of satellite images). The resultsshowed Chl-a and SDT volatility in different periods. Chl-a.during 1985 to 2017 showed an increasing trend that wasconsistent with changes in SDT. The western part of thewetland, as compared to other areas, was affected by thesechanges, which could be due to the human activityconcentrated in this area. Although it was not possible todraw profiles from the temperature of wetland depth with themethod used in this study, it provides valuable informationbased on the epilimnion layer, as well as patterns and spatial-temporal distribution of WST in different years (1985–2017).Furthermore, combining in situ data and satellite data can fillthe gaps of our information in this ecosystem and help ustoward comprehensive management of these valuable resour-ces and represent an outlook from the past to the present. Theinformation obtained from the present study could be helpful inthe future optimal management of wetlands (exposed toclimate change, pollution, etc.). Mapping the spatial distribu-tion for Chl-a, SDT, andWSTwith remotely sensed data wouldbe helpful for management of water bodies by determining thepoint and non-point sources of pollutions that are responsiblefor such spatial variability. Thus, we conclude that remotesensing can potentially be used as a tool for monitoring waterquality throughout the seasons and can provide naturalresource managers and decision makers with crucial informa-tion. Future research should focus more on the use of remotelysensed data to estimate viability seasonal variations in waterquality parameters in long-term and their potential impact as anapproach to mitigate climate change.

Funding

This study was funded by the Iran National ScienceFoundation (INSF) [grant number 97008261]. Also, thisresearch was financially supported by Gorgan University ofAgricultural Sciences and Natural Resources (GAU), Gorgan,Iran.

of 13

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

Acknowledgments. We thank Iran National Science Foundation(INSF) and Gorgan University of Agricultural Sciences andNatural Resources (GAU), Gorgan, Iran for their support.

References

Adrian R, O’Reilly CM, Zagarese H, et al. 2009. Lakes as sentinelsof climate change. Limnol Oceanogr 54: 2283–2297.

APHA–AWWA–WEF (American Public Health Association–Ameri-can Water Works Association–Water Environment Federation).2000. Standard methods for the examination of water andwastewater (18th ed.). APHA–AWWA–WEF: Washington, DC.

Alcantara EH, Stech JL, Lorenzzetti JA, et al. 2010. Remote sensingof water surface temperature and heat flux over a tropicalhydroelectric reservoir. Remote Sens Environ 114: 2651–2665.

Bonansea M, Rodriguez MC, Pinotti L, Ferrero S. 2015. Using multi-temporal Landsat imagery and linear mixed models for assessingwater quality parameters in Río Tercero reservoir (Argentina).Remote Sens Environ 158: 28–41.

Brezonik P, Kevin D, Bauer M, Bauer M. 2005. Landsat-based remotesensing of lake water quality characteristics, including chlorophylland colored dissolved organic matter (CDOM). Lake ReservManag 21: 373–382.

Behrouzirad B. 2007. Wetlands of Iran, Tehran, Iran.Boynton WR, Kemp WM, Keefe CW. 1982. A comparative analysisof nutrients and other factors influencing estuarine phytoplanktonproduction. In: V.S. Kennedy (ed.), Estuarine Comparisons. NewYork: Academic Press, p. 69–90. DOI: 10.1016/B978-0-12-404070-0.50011-9.

Carpenter SR. 2005. Eutrophication of aquatic ecosystems:Bistability and soil phosphorus. Proc. Natl Acad Sci 102: 10002–10005.

Covino T, Golden HE, Li HY, Tang J. 2009. Aquatic carbon-nutrientdynamics as emergent properties of hydrological, biogeochemical,and ecological interactions: scientific advances. Water Resour Res54:7138–7142.

Chander G, Markham BL. 2003. Revised Landsat-5 TM radiometriccalibration procedures and postcalibration dynamic ranges. IEEETrans. Geosci. Remote Sens. 41: 2674–2677.

Chao Rodríguez Y, Anjoumi A, Domínguez Gómez JA, RodríguezPérez D, Rico E. 2014. Using Landsat image time series to study asmall water body in Northern Spain. Environ Monitor Assess 186:3511–3522.

Chen J, Quan W. 2012. Using Landsat/TM imagery to estimatenitrogen and phosphorus concentration in Taihu Lake, China. IEEEJ Stars 5: 273–280.

Chen ZH, Mao ZH, Chen JY. 2011. Coastline change monitoringusing 4 periods’ remote Sensing data in Zhejiang Province from1986 to 2009. Appl Remote Sens Technol 26: 68–73.

Dörnhöfer K, Oppelt N. 2016. Remote sensing for lake research andmonitoring � recent advances. Ecol Indic 64: 105–122.

Ebrahimi E. 2006. Using GIS Techniques in Aquatic EnvironmentStudies: Developing Databases and Map for ChoghakhorWetland. University project research, No, 2887, Departmentof Natural Resources, Isfahan University of Technology, 80 p.

Ebrahimi S,Moshari M. 2006. Evaluation of the ChoghakhorWetlandstatus with the emphasis on environmental management problems.Publications of the Institute of Geophysics. Polish Acad Sci E-6(390) 8pp.

Esmaeili AR. 2012. Trophic Status of Choghakhor Wetland, Thesis ofMaster of Science, Department of Natural Resources, IsfahanUniversity of Technology, 87 p.

Page 12

ElMasri B, Rahman AF. 2008. Estimation of water quality parametersfor Lake Kemp Texas derived from remotely sensed data. Availableonline at twri.tamu.edu/funding/usgs/2006-07/el-masri_manu-script.pdf

Fazelpoor K, Dadolahi Sohrab A, Elmizadeh H, Asgari HM, KhazaeiSH. 2015. Evaluating the Efficiency of the Use of Satellite Imagesin Measuring the Sea Surface Temperature and Carbon Fixation inthe Persian Gulf. Tech J Eng Appl Sci 5: 242–254.

Filstrup CT, Downing JA. 2017. Relationship of chlorophyll tophosphorus and nitrogen in nutrient-rich lakes. Inland Waters 7:385–400.

Greb SR, Martin AA, Chipman JW. 2009. Water clarity monitoring oflakes in Wisconsin, USA using Landsat, in Proceedings of 33rdInternational Symposium of Remote Sensing of the Environment.Stresa, Italy.

Hicks BJ, Stichbury GA, Brabyn LK, Allan MG, Ashraf S. 2013.Hindcasting water clarity from Landsat satellite images ofunmonitored shallow lakes in the Waikato region, New Zealand.Environ Monitor Assess 185: 7245–7261.

Iqbal M, Billah M, Haider N, Islam Sh, Payel HR. 2017. Seasonaldistribution of phytoplankton community in a subtropical estuaryof the south-eastern coast of Bangladesh. Zool Ecol 27(3-4), https://doi.org/10.1080/21658005.2017.1387728.

Ji L, Zhang L, Wylie B. 2009. Analysis of dynamic thresholds for thenormalized difference water index. Photogram Eng Remote Sens75: 1307–1317.

Kang K, Kim SH, Kim D, Cho YK, Lee SH. 2014. Comparison ofcoastal sea surface temperature derived from ship, air, and space-borne thermal infrared systems. Int Geosci Remote Sens Symp:4419–4422.

Keddy PA. 2010. Wetland Ecology: Principles and conservation.Cambridge University Press.

Kufel L, Prejs A, Rybak JI (Eds.). 1997. Shallow Lakes ’95: TrophicCascades in Shallow Freshwater and Brackish Lakes (Develop-ments in Hydrobiology) Dordrecht: Springer, Vol. 119

Lee Z, Shang S, Qi L, Yan J, Lin G. 2016. A semi-analytical scheme toestimate Secchi-disk depth from Landsat-8 measurements. RemSens Environ 177: 101–106.

Matthews MW. 2011. A current review of empirical procedures ofremote sensing in inland and near-coastal transitional waters. Int JRemote Sens 32: 6855–6899.

McFeeters SK. 1996. The use of the Normalized Difference WaterIndex (NDWI) in the delineation of open water features. Int JRemote Sens 17: 1425–1432.

Murrell MC, Hagy JD, Lores EM, Greene RM. 2007. Phytoplank-ton production and nutrient distributions in a subtropicalestuary: importance of freshwater flow. Estuar Coasts 30:390–402.

Odermatt D, Gitelson A, Brando VE, SchaepmanM. 2012. Review ofconstituentretrieval in optically deep and complex waters fromsatellite imagery. Rem Sens Environ 118: 116–126.

Olmanson LG, Bauer ME, Brezonik PL. 2008. A 20-year Landsatwater clarity census of Minnesota’s 10, 000 lakes. Remote SensEnviron 112: 4086–4097.

Ogilvie A, Belaud G, Massuel S, Mulligan M, Le Goulven P, CalvezR. 2018. Surface water monitoring in small water bodies: potentialand limits of multi-sensor Landsat time series. Hydrol Earth SystSciences 22: 4349–4380.

Patra P, Dubey SK, Kumar Trivedi R, Kumar Sahu S, Keshari Rout S.2016. Estimation of Chlorophyll a concentration and trophic statesfor an inland lake from Landsat 8 OLI data: a case of Nalban Lakeof East Kolkata Wetland, India. doi: 10.20944/preprints201608.0149.v1.

of 13

A.R. Pirali Zefrehei et al.: Ann. Limnol. - Int. J. Lim. 2020, 56, 6

Pereira HC, Allott N, Coxon C. 2010. Are seasonal lakes asproductive as permanent lakes? A case study from Ireland. Can JFish Aquatic Sci 2010: 1291–1302.

Primpas I, Tsirtsis G, Karydis M, Kokkoris GD. 2010. Principalcomponent analysis: development of a multivariate index forassessing eutrophication according to the european waterframework directive. Ecol Indic 10: 178–183.

Ramsar Sites Information Service (RSIS). 2010. ChoghakhorWetland. https://rsis.ramsar.org/ris/1939?language=en.

Ryan PP, Geoffrey JH, Gang C. 2012. How wetland type and areadiffer through scale: a geobia case study in Alberta’s Boreal Plains.Remote Sens Environ 117: 135–145.

Samadi J. 2015. Survey of spatial-temporal impact of quantitative andqualitative of land use wastewaters on Choghakhor wetlandpollution using IRWQI index and statistical methods. Iran WaterResour Res 11: 157–191.

Senay GB, Shafique NA, Autrey BC, Fulk F, Cormier SM. 2001. Theselection of narrow wavebands for optimizing water qualitymonitoring on the Great Miami River, Ohio using HyperspectralRemote Sensor Data. J Spat Hydrol 1: 1–22.

Shutler JD, Land PE, Smyth TJ, Groom SB. 2007. Extending theMODIS 1 km ocean colour atmospheric correction to theMODIS 500 m bands and 500 m chlorophyll-a estimationtowards coastal and estuarine monitoring. Rem Sens Environ107: 521–532.

Sima S, Ahmadalipour A, Tajrishy M. 2013. Mapping surfacetemperature in a hyper-saline lake and investigating the effect oftemperature distribution on the lake evaporation. Rem Sens Environ136: 374–385.

Simon RN, Tormosa T, Danisb PA. 2014. Retrieving water surfacetemperature from archive LANDSAT thermal infrared data:application of the mono-channel atmospheric correction algorithm

Page 13

over two freshwater reservoirs. Int J Appl Earth Observ Geoinfor30: 247–250.

Souchu P, Bec B, Smith VH, et al. 2010. Patterns in nutrient limitationand chlorophyll a along an anthropogenic eutrophication gradientin French Mediterranean coastal lagoons. Can J Fish Aquatic Sci67: 743–753.

Sun D, Hu C, Qiu Z, Cannizzaro JP, Barnes BB. 2014. Influence of ared band-based water classification approach on chlorophyllalgorithms for optically complex estuaries. Rem Sens Environ155: 289–302.

Stumpf RP, Culver ME, Tester PA, et al. 2003. Monitoring Kareniabrevis blooms in the Gulf of Mexico using satellite ocean colorimagery and other data. Harmful Algae: 147–160.

Syariza MA, Jaelania LM, Subehie L, Pamungkasb A, KoenhardonocES, Sulisetyonod A. 2015. Retrieval of Sea Surface TemperatureOver Poteran Island Water of Indonesia with Landsat 8 TIRSImage: A Preliminary Algorithm, The International Archives of thePhotogrammetry, Remote Sensing and Spatial Information Sciences,Volume XL-2/W4, 2015, Joint International GeoinformationConference, 28–30 October 2015, Kuala Lumpur, Malaysia.

Urbanskia JA, Wochnaa A, Bubakb I, et al. 2016. Application ofLandsat 8 imagery to regional-scale assessment of lakewaterquality, Int J Appl Earth Observ Geoinfor 51: 28–36.

Wetzel RG. 2001. Limnology: Lake and river ecosystems. San Diego,CA: Academic Press, 3rd ed. p. 1006.

Yüzügüllü O, Aksoy A. 2011. Determination of Secchi Disc depths inLake Eymir using remotely sensed data. Proc Social Behav Sci 19:586–592.

Zoriasatein N, Jalili S, Poor F. 2013. Evaluation of ecological qualitystatus with the trophic index (TRIX) values in coastal area ofArvand, northeastern of Persian Gulf, Iran.World J Fish Mar Sci 5:257–262.

Cite this article as: Pirali Zefrehei AR, Hedayati A, Pourmanafi S, Beyraghdar Kashkooli O, Ghorbani R. 2020. Monitoring spatiotemporalvariability of water quality parameters Using Landsat imagery in Choghakhor International Wetland during the last 32 years. Ann. Limnol. -Int. J. Lim. 56: 6

of 13