Assessing the effectiveness of potential best management practices for science-informed decision support at the watershed scale: The case of the Mar Menor coastal lagoon, Spain

Abstract

Coastal lagoons are ecosystems of high environmental importance but are quite vulnerable to human activities. The continuous inflow of pollutant loads can trigger negative impacts on the ecological status of these water bodies, which is contrary to the European Green Deal. One example is the Mar Menor coastal lagoon in Spain, which has experienced significant environmental degradation in recent years due to excessive external nutrient input, especially from non-point source (NPS) pollution. Mar Menor is one of the largest coastal lagoons of the Mediterranean region and a site of great ecological and socio-economic value. In this study, the highly anthropogenic and complex watershed of Mar Menor, known as Campo de Cartagena (1244 km²), was modelled with the Soil and Water Assessment Tool (SWAT) to analyse potential options for recovery of this unique system. The model was used to simulate several best management practices (BMP) proposed by recent Mar Menor regulations, such as vegetative filter strips, shoreline buffers, contour farming, removal of illegal agriculture, crop rotation management, waterway vegetation restoration, fertiliser management and greenhouse rainwater harvesting. Sixteen scenarios of individual and combined BMPs were analysed in this study. We found that, as individual measures, vegetative filter strips and contour farming were most effective in nutrient reduction: approximately 30 % for total nitrogen (TN) and 40 % for total phosphorus (TP). Moreover, waterway vegetation restoration showed the highest sediment (S) reduction at approximately 20 %. However, the combination of BMPs demonstrated clear synergistic effects, reducing S export by 38 %, TN by 67 %, and TP by 75 %. Selecting the most appropriate BMPs to be implemented at a watershed scale requires a holistic approach considering effectiveness in reducing NPS pollution loads and BMP implementation costs. Thus, we have demonstrated a way forward for enabling science-informed decision-making when choosing strategies to control NPS contamination at the watershed scale.

Graphical abstract

Highlights

• •Mar Menor has experienced a significant environmental degradation in the last years.
• •Several BMP scenarios proposed by legislations are assessed with the SWAT model.
• •VFS and CF were found the most effective BMPs at the watershed scale.
• •An effective combination of BMPs could reduce the nutrient exports by 70 %.
• •The research findings can guide decision-makers to improve coastal lagoon situation.

1. Introduction

Coastal lagoons are shallow water bodies generally separated from the open sea by a sandy bar, characterised by low water renewal ratios due to their limited water exchange with the sea and reduced freshwater inputs (Soria et al., 2022). According to the Habitats Directive of the European Union (EU, 1992), coastal lagoons are threatened sites declared as a priority for environmental protection, as they support ecosystems that are very vulnerable to hydrological alterations, water pollution and habitat loss (Pérez-Ruzafa et al., 2005). Coastal lagoons and their surroundings are often under high anthropogenic pressures and undergo significant socio-economic and environmental changes over the years (Upadhyay et al., 2022). The coastal lagoon areas exemplify the conflict of interest between the development of human activities and the ecological requirements of aquatic ecosystems (Flower and Thompson, 2009). Soria et al. (2022) identified thirty-seven coastal lagoons with a surface area larger than 10 km² in the Mediterranean region, and concluded that most of them showed eutrophication problems due to the pollution of their inflows. To cope with this issue, the European Green Deal developed a list of goals and actions to preserve biodiversity, reverse the environmental degradation and protect ecosystems (EC, 2019). Since eutrophication has been one the main environmental issues reported for coastal lagoons (Viaroli et al., 2010; Rodríguez-Gallego et al., 2017), this study focuses on the case of the highly anthropised Mar Menor coastal lagoon as a representative case to assess potential solutions for controlling contamination in coastal lagoons at the watershed scale.

Mar Menor is one of the largest coastal lagoons of the Mediterranean region (135 km²), with significant cultural, socio-economic and ecological value. The lagoon and its adjacent areas are protected at the national and international levels. Mar Menor is included in the Ramsar wetland sites of international importance and the Specially Protected Areas of Mediterranean Importance (SPAMI; Boletín Ofical del Estado [BOE], 2020). Moreover, Perni et al. (2011) estimated the total economic value of the Mar Menor coastal lagoon under good ecological conditions to approximately 45M € per year. During the last few decades, the surroundings of Mar Menor have undergone significant changes because of tourism intensification and intensive agricultural expansion (Álvarez-Rogel et al., 2020). These changes have led to an increase in non-point source (NPS) pollution loads into the lagoon, negatively impacting its ecological status, exemplified by increasing algal levels, more frequent hypoxia events, and subsequent fish kills. Among major NPS sources, agriculture has been recognised as a main source of nutrients (Liu et al., 2013), such as nitrogen and phosphorus. Eutrophication is the main cause of water quality degradation in coastal lagoons (Le Moal et al., 2019), and its consequences adversely impact the local economy, mainly in the fishery and tourism sectors (Jimeno-Sáez et al., 2020). In the Mar Menor watershed, surface runoff can be very high due to common torrential rainfall events (Senent-Aparicio et al., 2021a), generating massive inflows of water and pollution loads into the coastal lagoon (García-Pintado et al., 2007) and further aggravating its vulnerability. Therefore, reducing NPS pollution from anthropogenic activities is required to avoid and reduce the severe environmental degradation of the Mar Menor coastal lagoon.

Throughout recent years, Spanish and regional governments have developed several legislations (BOE, 2020; BORM. Boletín Oficial de la Región de Murcia, 2019, BORM. Boletín Oficial de la Región de Murcia, 2018, BORM. Boletín Oficial de la Región de Murcia, 2017) to counteract the degradation of Mar Menor. Thus, the most recent Mar Menor laws have suggested implementing specific measures to achieve environmental goals, including a range of best management practices (BMP) in the complex and highly anthropogenic Mar Menor watershed known as Campo de Cartagena (CC). The CC is characterised by a semi-arid climate and ephemeral streams combined with intensive agriculture, mainly supported by groundwater pumping and the Tagus-Segura water transfer scheme (Alcolea et al., 2019). Vegetative filter strips, shoreline buffers, contour farming, removal of illegal agriculture, crop rotation management, waterway vegetation restoration, fertiliser management and greenhouse rainwater harvesting are some of the BMPs proposed by the regional government to limit NPS loading into the coastal lagoon. However, no efficiency assessment of the above-mentioned actions has yet been conducted at the whole watershed scale due to the complexity of the study area and the scarcity of reliable gauging data (Senent-Aparicio et al., 2021a). Therefore, there is a great need to analyse the potential effectiveness of the different BMPs in reducing NPS pollution at the watershed scale to support science-informed decision-making. Thus, a comprehensive understanding of cost-effectiveness and environmental benefits is essential for decision-makers and farmers to select the most appropriate BMPs (Wu et al., 2022).

Watershed modelling is a useful and practical approach for evaluating BMPs since it can assess their efficiency with few spatial and temporal limitations (Lee et al., 2020). Hydrological models, such as the Soil Water Assessment Tool (SWAT; Arnold et al., 1998), have been widely adopted as powerful science-based tools to analyse the role of BMPs in reducing sediment and nutrient losses (Martin et al., 2021; Shi and Huang, 2021; Liu et al., 2019; Uniyal et al., 2020). Moreover, a few studies on surrounding areas of the Mar Menor coastal lagoon have assessed the effectiveness of certain BMPs in reducing NPS pollution with the above-mentioned approach (Puertes et al., 2021; López-Ballesteros et al., 2019). However, these studies have only addressed specific sub-basins of the CC, not considering the entire watershed.

This study is the first to analyse BMPs at the entire watershed scale of the highly important Mar Menor lagoon. The study builds further on the model developed in the study by Senent-Aparicio et al. (2021a), who analysed the different components of the hydrological cycle of the Mar Menor using a combination of the SWAT and the QGIS Water Ecosystems Tool (QWET; Nielsen et al., 2017; Nielsen et al., 2020). Following the recommendations of Senent-Aparicio et al. (2021a), a new line of research based on evaluating the efficiency and cost-effectiveness of BMPs at the watershed scale was conducted using an enhanced and more accurate SWAT model of the CC for agricultural practices.

The main objectives of this research are to (1) improve hydrological modelling of the CC, including more accurate agricultural information, (2) evaluate the effect of individual and combined BMPs on NPS pollution loads that flow into the Mar Menor coastal lagoon at the watershed scale and (3) assess the cost-effectiveness of BMP implementation for controlling NPS pollutants. Additionally, this work intends to help stakeholders and policymakers achieve a better science-based understanding of the Mar Menor environmental issue and identify an effective management strategy at the watershed scale.

2. Materials and methods

2.1. Study area

The Mar Menor coastal lagoon watershed, CC, is located in the Segura River Watershed, a semi-arid region of the southeast portion of the Iberian Peninsula (Fig. 1).

Fig. 1.

a) Location of the Segura River Watershed; b) Location of the Campo de Cartagena (CC) within the Segura River Watershed; c) the CC.

This region has one of the highest structural water deficits in Europe (Senent-Aparicio et al., 2016), further aggravated by economic activities. The CC covers 1244 km², most of which is agricultural land (approximately 75 %). This area is considered one of the main national and international producers of agricultural products in Europe (Castejón-Porcel et al., 2018). In addition, tourism has great relevance in the closest surrounding areas to the lagoon. Intensive irrigated agriculture in this region is mainly maintained by a combination of groundwater resources, reused wastewater, seawater desalination and the Tagus-Segura water transfer scheme (Álvarez-Rogel et al., 2020). The Tagus-Segura water transfer scheme, established in 1979, is one of the major water providers of the CC, with an average of 122 10⁶ m³ year⁻¹. However, recent studies (Senent-Aparicio et al., 2021b) have demonstrated that water availability from this source could decrease in upcoming years due to the impact of climate change on water resources at the headwaters of the Tagus River. Additionally, the water scarcity issues in this region have led to a high level of agricultural technification, with drip irrigation applied to approximately 90 % of crops (Alcon et al., 2011).

The semi-arid climate of the study area, corresponding to cold semi-arid climate (BSk) according to Köppen-Geiger climate classification, is characterised by an average annual temperature of approximately 17 °C (range 13°–24 °C) and average annual precipitation of 300 mm with high inter-annual variability. Ephemeral streams are the main drainage system of the CC, flowing only during intense rainfall events (Alcolea et al., 2019; Senent-Aparicio et al., 2015). Among all ephemeral streams, the eight most relevant in the CC are shown in Fig. 1: Mirador, Peraleja, La Maraña, Albujon, Miranda, Miedo, Beal and La Carrasquilla. Extreme rainfall events are common in Mediterranean watersheds and can produce severe floods (Garcia-Ayllon and Radke, 2021). During these extreme rainfall episodes, large amounts of water and NPS pollution loads are moved through ephemeral streams into the Mar Menor coastal lagoon (Velasco et al., 2006).

The CC's topography is characterised by gentle slope gradients (approximately 40 % of the slopes are <2 %) and altitude ranges from 0 to 1063 m above sea level. The dominant soil type is Calcaric Cambisols (FAO–ISRIC, 1990), a loam textural class mainly composed of silt and sand. Regarding land use, intensive irrigated agriculture is the dominant land use (45 %), followed by rain-fed agriculture (30 %) and forest and shrub lands (15 %). Intensive agriculture mainly consists of horticultural crops, citrus trees and greenhouses. In the horticulture of the CC, common agricultural practices are an annual three-crop rotation (broccoli, cantaloupe and lettuce) and the application of fertilisers (nitrogen and phosphorus) by fertigation.

2.2. Soil and water assessment tool

The SWAT model is one of the world's most extensively applied hydrological models (Gassman and Wang, 2015), and the one selected for this study. SWAT is a physically-based, semi-distributed, continuous-time hydrologic model developed by the United States Department of Agriculture (USDA) and Texas A&M University (Arnold et al., 1998). The SWAT model allows users to simulate water quality and quantity, nutrient and sediment loss and management practices at the watershed scale (Neitsch et al., 2011). Furthermore, SWAT can predict the effects of anthropogenic changes in complex watersheds (Lee et al., 2020) and is considered an effective tool for evaluating the impacts of NPS pollution on water quality variables such as sediment and nutrients (Liu et al., 2013). SWAT has also been widely used to evaluate water policies (Ricci et al., 2022; Čerkasova et al., 2021; Gassman et al., 2014). The SWAT model delineates streams and sub-basins based on a digital elevation model (DEM); each sub-basin is further divided into hydrologic response units (HRU) based on unique intersections of land uses, soil types and slopes. A water balance is derived for each HRU (Eq. (1)), and their outputs are aggregated through the channel network to obtain the hydrologic cycle of the watershed:

$${\mathit{SW}}_{ti}={\mathit{SW}}_{oi}+\sum \left(R_{\mathit{day}i}-Q_{\mathit{surf}i}-{\mathit{ET}}_i-W_{\mathit{seep}i}-Q_{\mathit{gw}i}\right)$$ (1)

where SW_t is the final soil water content (mm); SW_o is the initial soil water content on day i (mm); R_day is the precipitation on day i (mm); Q_surf is the surface runoff on day i (mm); ET is the evapotranspiration on day i (mm); W_seep is the percolation on day i (mm) and Q_gw is the groundwater return flow on day i (mm).

In SWAT, surface runoff is simulated by the Soil Conservation Service (SCS) Curve Number (CN) method (USDA-SCS, 1972), soil erosion is estimated by the Modified Universal Soil Loss Equation (MUSLE; Williams, 1975) and water quality variables are calculated by equations from the QUAL2E model (Brown and Barnwell, 1987). In this study, the Penman-Monteith method was selected to estimate potential evapotranspiration (PET). Detailed documentation of the SWAT model is found in Neitsch et al. (2011).

2.3. Model inputs and setup

SWAT is a hydrological model requiring a large amount of input data, such as meteorological variables, topography, land uses, soil maps and management practices (Neitsch et al., 2011). Meteorological variables in SWAT include daily precipitation, maximum and minimum temperature, solar radiation, relative humidity, and wind speed. In this study, the climatic information was collected from six weather stations of the Murcian Institute of Agrarian and Food Research and Development located in and around the study area (Fig. 1). These weather stations provided observations of all meteorological variables from 2000 to 2021. Detailed information about the input maps used to set up the SWAT model of the CC is listed in Table 1.

Table 1.

SWAT input maps of the CC.

Map	Spatial resolution	Source
DEM	25 m × 25 m	National Geographic Institute of Spain
Land uses	200 m × 200 m	Spanish Ministry of Agriculture, Fisheries and Food (2000−2010)
Soil properties	1 km × 1 km	Harmonized World Soil Database

This study used the open-source geographic information system (GIS) interface for SWAT (QSWAT; Dile et al., 2016) to prepare and execute the SWAT model. Based on the SWAT input maps, after applying the SWAT2lake tool (Molina-Navarro et al., 2018), which is a plugin to assist in the delineation of the entire watershed that flows into a reservoir or lagoon, the CC was discretised by 152 sub-basins. Subsequently, slope classes of <2 %, 2 %–8 % and >8 % combined with a minimum area threshold of 100 ha (any HRU below this threshold was disregarded) were selected, resulting in a total of 520 HRUs (Senent-Aparicio et al., 2021a).

In irrigated agricultural areas, a realistic representation of agricultural management practices is an important requirement to achieve good performance when simulating water quantity and quality (Samimi et al., 2020). Therefore, the main agricultural management practices of the dominant CC land use (irrigated agriculture) were included in the SWAT model. The intensive irrigated land use in the study area was mainly composed of horticultural crops, citrus trees and greenhouses. For horticultural crops, an annual three-crop rotation (broccoli, cantaloupe and lettuce) was the standard crop schedule implemented (Table 2). Fertilisation rates for each crop were obtained from official government documentation (BORM. Boletín Oficial de la Región de Murcia, 2012, BORM. Boletín Oficial de la Región de Murcia, 2012, BORM. Boletín Oficial de la Región de Murcia, 2012; BOE, 2004). Additionally, irrigation volumes of the intensive agricultural area were extracted from the Hydrological Plan of the Segura River Watershed.

Table 2.

Standard crop schedule of the CC.

Year	Date		Operation	Application Rate	Crop
	Month	Day
1	January	1	Sowing		Broccoli
1	January	1	Irrigation	∼28 mm month−1	Broccoli
1	January	1	Fertilisationa	245 KgN ha−1 year−1	Broccoli
				100 KgP ha−1 year−1
1	April	30	Harvest and kill		Broccoli
1	May	1	Sowing		Cantaloupe
1	May	1	Irrigation	∼48 mm month−1	Cantaloupe
1	May	1	Fertilisationa	225 KgN ha−1 year−1	Cantaloupe
				105 KgP ha−1 year−1
1	August	31	Harvest and kill		Cantaloupe
1	September	1	Sowing		Lettuce
1	September	1	Irrigation	∼25 mm month−1	Lettuce
1	September	1	Fertilisationa	100 KgN ha−1 year−1	Lettuce
				58 KgP ha−1 year−1
1	December	31	Harvest and kill		Lettuce

^aTotal amount applied throughout the crop schedule.

2.4. Model calibration and validation

Simulating the hydrologic response of semi-arid agricultural areas is challenging due to complexities related to anthropogenic alterations and, hence, model calibration in these regions (Samimi et al., 2020). Furthermore, the calibration process is hampered by the lack of reliable gauging data, as in the CC study case. Therefore, a calibration and validation approach based on satellite-based actual evapotranspiration (AET) data obtained from Global Land Evaporation Amsterdam Model version 3b (GLEAM v3b; Miralles et al., 2011) was carried out in this study, and for nutrient exports, we compared our baseline estimates with those estimated in other studies.

In agricultural areas, evapotranspiration is a key variable in the hydrologic cycle of the watershed (Odusanya et al., 2019). GLEAM v3b is an AET dataset generated by a combination of remote sensing observations from several satellites, highly validated with eddy-covariance towers and in-situ sensors (Martens et al., 2017), covering a data period from 2003 to 2015 at a spatial resolution of 0.25° regular grid. In recent years, several studies have satisfactorily applied and validated the satellite-based AET calibration and validation process with GLEAM (Bennour et al., 2022; Odusanya et al., 2021; Puertes et al., 2021; López-Ballesteros et al., 2019).

The SWAT simulation period for the CC included a warm-up period of 3 years (2000−2002), a calibration period of 7 years (2003–2009) and a validation period of 6 years (2010–2015). Manual calibration was first conducted by selecting a SWAT parameter set relevant to AET (CN2, ESCO, EPCO, SOL_AWC and RCHRG_DP), comparing AET values from the SWAT with satellite-based AET values. Furthermore, the values of the calibrated AET parameters were preserved for consistency with the previous study (Senent-Aparicio et al., 2021a). Finally, the SWAT model performance of the CC was evaluated graphically and statistically. Originally, three well-established statistical evaluation indices were used: the coefficient of determination (R²), the percent bias (PBIAS) and the Nash-Sutcliffe efficiency (NSE; Nash and Sutcliffe, 1970). Moreover, the Kling-Gupta efficiency (KGE; Gupta et al., 2009) was included to comprehensively assess the SWAT model performance. R² ranges from 0 to 1, with a value closer to 1 producing the most accurate model. The PBIAS indicates model overestimation or underestimation, with an optimum value of 0. NSE and KGE ranges from −∞ to 1, where 1 is the optimal value. For classifying the model performance, the rating proposed by Moriasi et al. (2015) and Kouchi et al. (2017) for a monthly time step were applied, where satisfactory thresholds were R² > 0.6, PBIAS ≤ ±25 % and NSE and KGE ≥ 0.5. The calibrated and validated SWAT model of the CC was considered the baseline scenario for the next sections of this study.

2.5. Modelling individual BMP scenarios

The selection of BMPs was mainly based on the official legislation developed by the Spanish and regional governments (BOE, 2020; BORM. Boletín Oficial de la Región de Murcia, 2019, BORM. Boletín Oficial de la Región de Murcia, 2018) to counteract the degradation of Mar Menor. BMPs are expected to play a key role in the recovery process of Mar Menor (Álvarez-Rogel et al., 2020). In this study, vegetative filter strips, shoreline buffers, contour farming, removal of illegal agriculture, crop rotation management, waterway vegetation restoration, fertiliser management and greenhouse rainwater harvesting were the BMPs selected and implemented based on the baseline scenario. Although eight BMP types were assessed in this study, 13 individual BMP scenarios were simulated with SWAT as some BMPs were evaluated under several settings such as filter strips, shoreline buffers and contour farming. Therefore, 13 different individual BMP scenarios were designed and simulated in SWAT to estimate the reduction of NPS pollution loads flowing from the CC into the Mar Menor coastal lagoon.

2.5.1. Vegetative filter strips (VFS)

Vegetative filter strips (VFS) consist of installing a vegetated area along the edge of agricultural land to slow surface runoff, trap sediments and absorb nutrients. The SWAT model simulated the filter efficiency (trap_ef) of the VFS using Eq. (2) (Neitsch et al., 2011),

$${\mathit{trap}}_{\mathit{ef}}=0.367·{\mathit{FILTERW}}^{0.2967}$$ (2)

where FILTERW is the width of the VFS in metres.

According to BOE (2020), all agricultural land uses in the CC (approximately 75 % of the watershed) are forced to implement VFS. A filter strip width of 2–3 m is required for intensive irrigated agriculture, whereas for rain-fed agriculture, a filter strip width of 1 m is required in fields with average slopes <2 % and a filter strip width of 2 m in remaining slope classes. Therefore, two VFS scenarios were simulated in the SWAT. The first scenario applied the least restrictive filter strip width (2 m) in intensive irrigated land, while the second applied the 3 m filter strip width. In both scenarios, the VFS width of rain-fed land was applied as described above.

2.5.2. Shoreline buffers (SB)

Shoreline buffers (SB) can reduce NPS contamination by changing land-use patterns (Wang et al., 2013). SB consists of a band around the shore of the lagoon where land uses are modified according to the limitations imposed by law. Three SB scenarios were simulated with the SWAT model of the CC based on SB areas proposed by government regulations: (1) a buffer of 500 m from the Mar Menor shoreline (SB500; BORM, 2019); (2) a buffer of 1500 m (SB1500; BOE, 2020); and (3) a buffer including the special protection SB area established by BOE (2020) known as Zone 1 (SBzone1). These scenarios were implemented in SWAT by changing the intensive irrigated agriculture land uses inside the SB area to forest and shrub land use. The percentages of land-use change at the watershed scale are listed in Table 3.

Table 3.

Land-use changes of shoreline buffers (SB) scenarios at the watershed scale.

Scenario	Intensive irrigated land use		Forest and shrub land use
	Area (km2)	Percentagea (%)	Area (km2)	Percentagea (%)
Baseline	562.26	45.19	193.43	15.54
SB500	558.40	44.88	197.29	15.85
SB1500	538.07	43.25	217.62	17.49
SBzone1	484.08	38.91	271.61	21.83

^aPercentage of total study area.

2.5.3. Contour farming (CF)

Contour farming (CF) is an agricultural practice entailing planting, tilling and harvesting following the terrain contour lines. CF increases soil infiltration and decreases surface runoff, reducing soil erosion and NPS pollution loads that flow into streams, especially during intense rainfalls (Liu et al., 2013). CF was represented by changing the parameters CONT_CN and CONT_P to 65 and 0.8, respectively, in the SWAT operations module (.ops; Arnold et al., 2012). In this study, three CF scenarios were simulated to evaluate their effect on NPS pollution loads reduction for three different slope ranges (<2 %, 2 %–8 % and >8 %). Following the BOE (2020) requirements, CF was only applied for agricultural non-woody land uses of the CC.

2.5.4. Removal of illegal agriculture (RIA)

According to official data from SIGPAC (2016), farmers have officially declared approximately 474 km² of the CC as irrigated land with irrigation rights. However, according to the land-use map, the irrigated area in the CC is approximately 90 km² larger than the official data. Other reports have also identified these irregularities (Mar Menor Scientific Advisory Group, 2017), which are being prosecuted by law (BOE, 2020). Therefore, a scenario called the removal of illegal agriculture (RIA) has been simulated with SWAT to evaluate the impact of removing these illegal land uses on NPS pollution loads to the Mar Menor watershed. Thus, a new land-use map was created to simulate this RIA scenario, where approximately 90 km² of intensive irrigated land use was randomly removed and changed to forest and shrub lands. After introducing the modified land-use map, the calibrated SWAT model was re-established, and the RIA scenario was simulated.

2.5.5. Crop rotation management (CRM)

Crop rotation management (CRM) consists of modifying the annual crop schedule from three-crop to two-crop rotation, as the BOE (2020) established. This CRM limitation mainly affects horticultural crops, where the standard crop schedule is as follows: broccoli, cantaloupe and lettuce. In this study, the CRM scenario was implemented in the SWAT model of the CC by removing the cantaloupe crop due to its market price instability (Puertes et al., 2021). Additionally, a vegetation cover (representing a catch crop) was established during the non-crop period to prevent soil erosion.

2.5.6. Waterway vegetation restoration (WVR)

Waterway vegetation restoration (WVR) consists of recovering the autochthonous vegetation of streams to reduce flow velocity and channel erosion, enhancing NPS pollution loads retention. This WVR measure has been contemplated in the government regulations (BOE, 2020) and is expected to be carried out in the main ephemeral streams of the study area in the next years (BOE, 2021). In this study, the WVR scenario was simulated by re-vegetating the last kilometre of the main ephemeral streams of the CC (Fig. 1). This BMP was implemented in SWAT by using its operations module (.ops), where grassed waterway parameters were established, such as Manning's n value for vegetated surfaces (GWATN = 0.1) and length of waterways with vegetation (GWATL = 1 km).

2.5.7. Fertiliser management (FM)

Water quality can be improved by reducing applied fertiliser from agricultural watersheds (Risal and Parajuli, 2022). This study represented the fertiliser management (FM) scenario by reducing fertiliser application rates in intensive agriculture areas of the CC. Hence, the applied nitrogen and phosphorus amounts were reduced by 20 %, as suggested in the guidelines provided by the Code of Good Agricultural Practices of Murcia (BORM, 2018) and the European Commission (EC, 2020).

2.5.8. Greenhouse rainwater harvesting (GRH)

The BOE (2020) established that all greenhouses inside the CC must implement rainwater harvesting systems with a capacity of at least 100 l m⁻². Rainwater harvesting consists of the interception and storage of rainwater for irrigation and stormwater reduction (Waidler et al., 2009). Thus, a greenhouse rainwater harvesting (GRH) scenario was simulated in this study to evaluate the effect of GRH in reducing NPS pollution loads at the watershed scale. The GRH scenario was simulated using the pond module (.pnd; Arnold et al., 2012) in the SWAT model of the CC. First, a supervised classification of the CC orthophotos obtained from digital aerial orthophotographs of the Spanish National Orthophoto Program (PNOA) was carried out in QGIS with the maximum likelihood algorithm (Basukala et al., 2017) to identify the location and estimate the area of each greenhouse per sub-basin (Fig. 2). A total greenhouse area of 23.74 km² was quantified in the CC. Finally, pond parameters were adjusted using the previous GIS information, such as the fraction of sub-basin covered by greenhouses (PND_FR), the volume of water to fill the deposit (PND_PVOL) and the water surface area of the filled deposit (PND_PSA).

Fig. 2.

Location of greenhouses and sub-basins in the CC.

2.6. Modelling combined BMP scenarios

Considering synergistic effects, combining BMPs may provide better results than individual BMPs at the watershed scale (Liu et al., 2019; López-Ballesteros et al., 2019; Mtibaa et al., 2018). In practice, several BMPs are often implemented simultaneously in agricultural areas to control NPS contamination (Uniyal et al., 2020). BMPs are usually classified as agricultural or structural. Agricultural BMPs (AgriBMPs) are practices carried out at the field scale by farmers, while structural BMPs (StruBMPs) are artificial or natural practices implemented outside the field borders. As can be observed in Table 4, this study simulated three different combined BMP scenarios, two of them based on the previous BMP classification (AgriBMPs and StruBMPs) and the last one combining both scenarios (AllBMPs).

Table 4.

Combined BMP scenarios simulated.

Combined BMP scenario	Classification	Applied BMPs
AgriBMPs	Agricultural	VFS2m
		CF2–8
		CRM
		FM
StruBMPs	Structural	RIA
		WVR
AllBMPs	Agricultural + structural	VFS2m
		CF2–8
		CRM
		FM
		RIA
		WVR

For the AgriBMPs scenario, the selected BMPs were: VFS of 2-metre width (VFS2m), CF for the slopes between 2 % and 8 % (CF2–8), CRM and FM. For the StruBMPs scenario, the selected BMPs were: RIA and WVR. This selection was carried out considering the BMP implementation responsibility, which for the AgriBMPs lied with the farmers, while the responsibility for the StruBMPs lied with the government. Finally, for the last scenario, AllBMPs, a combination of all the above-mentioned BMPs was simulated.

2.7. Assessing BMP scenarios and cost-effectiveness

Annual NPS pollution loads, such as sediment (S), total nitrogen (TN) and total phosphorus (TP), discharged into the Mar Menor were compared with those of the baseline scenario to evaluate the impact of implemented BMP scenarios at the watershed scale. The total amount of NPS pollution loads flowing into the Mar Menor through the ephemeral streams network was obtained by aggregating the outputs of all surrounding Mar Menor sub-basins. Following the approach proposed by López-Ballesteros et al. (2019), the effectiveness (%) of the BMPs scenarios was computed using Eq. (3) for 2003–2021.

$$\mathit{Effectiveness}=\frac{\left(Y_{\mathit{baseline}}-Y_{\mathit{BMP}}\right)}{Y_{\mathit{baseline}}}·100$$ (3)

where Y_BMP and Y_baseline are the annual NPS pollution loads (t year⁻¹) produced by the selected BMP scenario and the baseline scenario, respectively. All final average effectiveness values were discussed with experts in the field.

Implementing BMPs also impacts farmers and society economically (Ricci et al., 2020). Therefore, a cost-effectiveness assessment was conducted to select the most effective BMPs at a reasonable economic cost. The cost-effectiveness (CE) ratio was calculated using the following expression (Eq. (4)):

$$\mathit{CE}\mathit{ratio}=\frac{\mathit{Total\; Cost}}{\mathit{Effectiveness}}$$ (4)

where Total Cost is the implementation cost in euros (€) of the simulated BMP scenario (Table 7) and Effectiveness is the percentage of change (%). The CE ratio represents the cost per percentage unit of change in NPS pollution loads. Therefore, a lower CE ratio means a more cost-effective scenario. Table 5 shows the cost per hectare estimated from the literature of each assessed BMP.

Table 7.

Assessment of BMP cost-effectiveness for controlling NPS contamination.

Scenario ID	Area (ha)	Total cost (€)	CE ratio
			S	TN	TP
VFS2m	93,446	1,734,928	266,912	64,257	47,017
VFS3m	93,446	2,297,188	319,054	75,815	55,622
SB500	386	1,544,000	–	–	–
SB1500	2418.5	9,674,000	3,869,600	4,606,667	4,030,833
SBzone1	7817.8	31,271,200	1,979,190	2,282,569	2,171,611
CF < 2	21,800.7	218,007	99,094	31,595	26,586
CF2–8	27,036.6	270,366	19,175	9259	6258
CF > 8	11,973.6	119,736	1,197,360	18,142	10,596
RIA	8974.3	3,500,000	2,692,308	198,864	184,211
CRM	44,852	2,466,860	601,673	172,508	97,891
WVR	2000	10,000,000	490,196	5000,000	4,347,826
FM	56,226	5,622,600	–	730,208	453,435
GRH	2374	1,543,100	5,143,667	15,431,000	15,431,000
AgriBMPs	93,446	10,094,754	566,388	169,381	145,188
StruBMPs	10,974.3	13,500,000	617,480	597,708	583,432
AllBMPs	95,446	23,594,754	620,770	351,143	313,401

Table 5.

Cost per hectare of each assessed BMP.

BMP	Cost per hectare (€/ha)	Source
Vegetative filter strips	10a	(López-Ballesteros et al., 2019)
Shoreline buffers	4000	(MITECO, 2019)
Contour farming	10	(López-Ballesteros et al., 2019)
Illegal agriculture removal	390	(MITERD, 2021)
Crop rotation management	10–100	(Amin et al., 2020)
Waterway vegetation restoration	5000	(MITECO, 2019)
Fertiliser management	100	(López-Ballesteros et al., 2019)
Greenhouse rainwater harvesting	6500b	(Sanches-Fernandes et al., 2015)

^aCost per installed metre width.

^bCost per rainwater harvesting system of 100 m³.

3. Results

3.1. SWAT model performance

The SWAT model of the CC achieved a good performance (Fig. 3) both in the calibration (R² = 0.73, PBIAS = −9.11 %, NS = 0.67 and KGE = 0.81) and the validation period (R² = 0.74, PBIAS = −5.22 %, NS = 0.71 and KGE = 0.82) for monthly AET according to the statistics criteria proposed by Moriasi et al. (2015) and Kouchi et al. (2017). A slight improvement in PBIAS was achieved compared to the previous SWAT model due to the more accurate information about agricultural management practices. More details about the AET manual calibration and selected SWAT parameters appear in Senent-Aparicio et al. (2021a).

Fig. 3.

Graphical comparison of AET for the calibration and validation periods.

In the SWAT model of the CC, the annual average values of the hydrologic cycle components for 2003–2021 were precipitation = 301 mm, PET = 1296 mm, AET = 408.5 mm, surface runoff = 38.1 mm, lateral flow = 7.25 mm and groundwater discharge = 4.7 mm. Similar water balance values were validated in a previous study by Senent-Aparicio et al. (2021a). AET was greater than precipitation, aligned with the distinctive agricultural nature of the CC.

Due to the lack of reliable gauging data to assess the SWAT model performance in simulating NPS pollutants, a soft validation (Arnold et al., 2015) was conducted for sediment and nutrient loads by comparing our annual estimates with those in other studies. Regarding sediment loads, the SWAT model of the CC estimated an average sediment yield of 2.52 t ha⁻¹ year⁻¹, which is within the range of 0–5 t ha⁻¹ year⁻¹ quantified by the National Soil Erosion Inventory 2002–2012 (MIMAM, 2012) for the study area and in line with the average estimation of approximately 2 t ha⁻¹ year⁻¹ of Romero-Díaz et al. (2011) for this region. Regarding nutrient loads, SWAT estimated an average TN inflow to the Mar Menor coastal lagoon of 482.4 t year⁻¹ for 2003–2021, which is in agreement with the range (515 ± 176 t year⁻¹) of nitrogen inputs to Mar Menor obtained by García-Pintado et al. (2007). The average TP inflow simulated by SWAT was 242.2 t year⁻¹, similar to the TP value of approximately 240 t year⁻¹ estimated in other studies of the Mar Menor coastal lagoon (Mar Menor Scientific Advisory Group, 2017).

3.2. Efficiencies of individual BMP scenarios in reducing NPS pollution loads

The impact of each BMP on NPS pollution loads that flow into the Mar Menor is listed in Table 6 based on its effectiveness.

Table 6.

Summary of the effectiveness of individual BMP scenarios.

Individual BMP	Scenario details	Scenario ID	Effectiveness (%)
			Sediment (S)	Total nitrogen (TN)	Total phosphorus (TP)
Vegetative filter strips	2 m width in irrigated agriculture	VFS2m	6.5	27.0	36.9
	3 m width in irrigated agriculture	VFS3m	7.2	30.3	41.3
Shoreline buffers	500 m buffer from the shoreline	SB500	–	–	–
	1500 m buffer from the shoreline	SB1500	2.5	2.1	2.4
	Zone 1 buffer from the shoreline	SBzone1	15.8	13.7	14.4
Contour farming	Slopes < 2 %	CF < 2	2.2	6.9	8.2
	Slopes between 2 % and 8 %	CF2–8	14.1	29.2	43.2
	Slopes > 8 %	CF > 8	0.1	6.6	11.3
Illegal agriculture removal		RIA	1.3	17.6	19.0
Crop rotation management		CRM	4.1	14.3	25.2
Waterway vegetation restoration		WVR	20.4	2.0	2.3
Fertiliser management		FM	–	7.7	12.4
Greenhouse rainwater harvesting		GRH	0.3	0.1	0.1

In the CC, VFS3m was the most effective individual BMP scenario in reducing TN and TP, with effectiveness values of 30.3 % and 41.3 %, respectively. For reducing S loads into the Mar Menor, WVR was found more effective than all other individual BMP scenarios, with a reduction of 20.4 %. In terms of land-use management practices, restoration of illegal agriculture to forest and shrub lands was the most effective scenario in abating TN and TP, with a reduction of 17.6 % and 19 %, respectively. Regarding CF scenarios, the results of this study showed that CF was most effective in the slope range of 2 %–8 %. Similar results about CF were found in other studies (Wu et al., 2022; Liu et al., 2013). Two agricultural BMP scenarios farmers can easily implement, CRM and FM, showed a low effect on S loads but reduced TN and TP discharge into the Mar Menor by 14.3 % and 25.2 %, and 7.7 % and 12.4 %, respectively. These results were consistent with the BMP trends of other studies in small parts of the study area (Puertes et al., 2021; López-Ballesteros et al., 2019).

3.3. Efficiencies of combined BMP scenarios in reducing NPS pollution loads

Fig. 4 shows the effectiveness of the combined BMP scenarios for the assessed NPS pollution loads in this study.

Fig. 4.

Boxplot of the effectiveness of the combined BMP scenarios for the 19-year period (2003−2021).

As can be observed in Fig. 4, all combined BMP scenarios simulated a reduction in NPS pollution in the Mar Menor. For sediment loads, 17.8 %, 21.9 % and 38 % effectiveness were estimated for the AgriBMPs, StruBMPs and AllBMPs scenarios, respectively. TN effectiveness was 59.6 % for the AgriBMPs scenario, 22.6 % for the StruBMPs scenario and 67.2 % for the AllBMPs scenario. TP effectiveness showed the highest reduction values, with 69.5 %, 23.1 % and 75.3 % for the AgriBMPs, StruBMPs and AllBMPs scenarios, respectively.

3.4. BMP cost-effectiveness

Table 7 shows the cost-effective assessment carried out in this study for all simulated BMP scenarios. CE ratios were obtained from the effectiveness results in both individual and combined BMP scenarios and BMP implementation costs shown in Table 7. The total cost considers the investment to implement each BMP at the watershed scale.

Among the individual BMP scenarios, CF2–8 was the most cost-effective with CE ratios of 19,175 for S, 9259 for TN and 6258 for TP. Regarding land-use changes, the buffer area of Zone 1 (SBzone1) showed a higher cost-effectiveness than the buffer area of 1500 m (SB1500) with CE values of 1,979,190, 2,282,569 and 2,171,611 for S, TN and TP, respectively. However, RIA was still the best BMP scenario for land-use changes in reducing nutrient inputs with CE ratios of 198,864 for TN and 184,211 for TP. VFS was the second most cost-effective BMP, with CE ratios of 266,912 for S, 64,257 for TN and 47,017 for TP in the VFS2m scenario and approximately 15 % higher values in the VFS3m scenario. The WVR scenario, although being one the most expensive with a total cost of 10M €, had a remarkable performance in the CE ratio of S with a value of 490,196.

Regarding combined BMP scenarios, the initial investment to apply structural or agricultural BMP scenarios was similar with a total cost of >10M €. However, the CE ratio in reducing nutrients was better in the AgriBMPs scenario with values of 169,381 and 145,188 for TN and TP, respectively, than in the StruBMPs scenario with values of 597,708 for TN and 583,432 for TP. Despite the high implementation cost of the AllBMPs scenario (approximately 24M €), its CE ratio was satisfactory from a cost-effective point of view with values of 620,770, 351,143 and 313,401 for S, TN and TP, respectively.

4. Discussion

4.1. Efficiencies of individual and combined BMP scenarios in reducing NPS pollution loads

All the assessed BMP scenarios showed a reduction in S, TP and TN at a watershed scale except for FM, which only affected nutrients, and SB500, which had no effects on sediment and nutrients, likely since the coast of Mar Menor is mainly urban land use. VFS3m was one of the most effective individual BMP scenarios for reducing TN and TP. This could be attributed to its high level of implementation in the watershed (approximately 75 % of the CC). Similar efficiencies in reducing TN and TP were found for the CF2–8 scenario, because 2 %–8 % is the slope range where CF is usually most effective (Wu et al., 2022). Furthermore, WVR was found more effective than all other individual BMP scenarios for reducing S loads. This result is likely due to this kind of structural BMP usually works by promoting sedimentation and avoiding in-stream erosion (Nepal and Parajuli, 2022). The impact of GRH on NPS pollution loads was almost negligible at the watershed scale, possibly due to the greenhouse areas inside the CC only covering approximately 2 % of the watershed. It should be noted that the effectiveness of the BMPs could vary within the CC, as their effects are dependent on the slope, soil and land use types within CC. For example, most of the steeper slopes are located in the north-western part of CC, which means that this part could react somewhat differently to a specific BMPs compared to the flatter areas in the south-eastern part of CC near the coastal lagoon.

An optimal combination of BMPs can potentially result in a greater impact on the control of NPS contamination (López-Ballesteros et al., 2019) and an adequate number of BMPs must therefore be selected to achieve an efficient solution (Uniyal et al., 2020). As expected, combined BMPs also showed a higher effectivity in reducing NPS pollution loads than individual BMPs. However, the effectiveness of the combined BMP scenarios differed greatly. The AgriBMPs scenario was found to be more effective in reducing nutrient inputs than the StruBMPs scenario. This higher effectiveness of the AgriBMPs scenario in reducing TN and TP can be attributed to the fact that it operates at the source of the nutrient production. On the other hand, the StruBMPs scenario had the best performance in reducing sediment inputs into Mar Menor, which is in accordance with the results of the individual BMPs. Finally, the AllBMPs scenario showed the synergistic effects of implementing all the BMPs at the same time. As can be observed in Fig. 4, it had the highest effectiveness in controlling the NPS contamination. The results of this study suggest that implementing a combination of BMPs can be a better solution to reduce the NPS pollution loads inputs into the Mar Menor coastal lagoon at a watershed scale than individual BMPs. Overall, the results of this study can provide information for a better management of nutrients, which is relevant to 16 of the 17 Sustainable Development Goals (SDG) proposed by the United Nations (UN) (Kanter et al., 2016), especially the SDGs that focus on tackling environmental problems (SDG 3, 6, 11–15). An effective nutrient management could also help reach a balance in the trade-offs between environmental sustainability and economic development.

In this study, all the assessed BMPs have been extracted from the government regulation (BOE, 2020) and thereby their implementation is mandatory. However, most of them have not been applied yet due to the lack of consensus between the stakeholders and the absence of coordination in the public administration (Guaita-García et al., 2022). One reason for the lack of consensus could be the assessment of the effectiveness of these BMPs, which has not been evaluated until now. Therefore, the effectiveness of BMPs should be scientifically demonstrated to gain the trust of stakeholders and, in that way, to reach a consensus to prioritize the better measures to protect the Mar Menor coastal lagoon.

4.2. BMP cost-effectiveness

BMPs are critical for reducing NPS contamination, but a balance between their implementation cost and their effectiveness should be achieved. The CF2-8 scenario showed the best CE ratio because the implementation of CF is inexpensive, only involving changes in how the land is managed. With regards to the land use change BMPs, RIA was the best scenario in terms of CE, since its implementation cost was negligible compared to the other land-use management practices. The differences of CE values between the VFS2m and the VFS3m scenario showed that 2 m filter strip width was more profitable than 3 m width. Restoration of vegetation in the main ephemeral streams of the CC was an expensive measure. However, this scenario (WVR) showed an acceptable CE ratio in reducing sediments inputs to Mar Menor. A similar action was proposed by the Spanish government with an investment of 70M € (MITERD, 2021), although that proposal included a larger scope than the one simulated in this study. The assessment of cost-effectiveness carried out in this study is expected to help decision-makers when selecting the most appropriate BMPs, it is important to note that the priority order of these measures could change if subsidies were provided, which would likely lead to better acceptance of BMP implementation by farmers (Ricci et al., 2022).

The differences in the CE ratios between StruBMPs and AgriBMPs can be attributed to the fact that the AgriBMP scenario had a higher effectiveness in reducing TN and TP loads that flow into the coastal lagoon. The AllBMPs scenario was considered economically sustainable and its benefits can have a major impact on improving the environmental situation of the Mar Menor coastal lagoon. Although in general terms, the implementation cost of the assessed BMP scenarios may seem expensive, it is negligible in comparison with the economic value of the Mar Menor under good status, which amounts to 45M € per year according to Perni et al. (2011), is a negligible cost. Therefore, private and public investments are necessary to get the implementation of BMPs at the watershed scale.

4.3. Limitations and uncertainties

Model results are affected by several uncertainties, including errors in input data, simplification of complex processes and non-unique model parameter values (Evenson et al., 2021). In this study, input data uncertainty could be significant due to the lack of reliable gauging observations. Although weather records are very complete, gauging data of streamflow, sediment and nutrient loads are missing. This puts a major constraint on the calibration and validation of the model. This limitation was overcome to some extent by using satellite-based AET data, although the spatial resolution may result in an additional error. In areas characterised by a semi-arid climate and an agricultural nature such as CC, AET is the main driver of the water balance (Odusanya et al., 2019). Remote sensing data allow to deal with the limited availability of hydrological data (Bennour et al., 2022). In this study, it was assumed that water balance components and nutrient and sediment loads were well estimated and a soft-validation was carried out to validate these estimates. However, some uncertainties remain and the results should be cautiously interpreted. All BMP scenarios were designed based on literature, thus their effectiveness were not validated with field observations. However, a study by Arabi et al. (2007) demonstrated that a relative intercomparison between different BMP scenarios is affected by lower uncertainty, providing consistent results when ranking their effectiveness.

5. Conclusions

Despite the scarcity of reliable data and the high complexity of the study area, an improved SWAT model of the CC for agricultural practices was achieved in this study. This more accurate modelling of the Mar Menor watershed provided a useful tool to simulate BMP scenarios and estimate their impact at the watershed scale. Moreover, this model could assist with a better global understanding of the Mar Menor environmental issues. As its main objective, this study simulated several BMP scenarios at the watershed scale to evaluate their impact on controlling NPS pollution in the CC and test possible solutions to improve the Mar Menor ecological status.

Regarding individual BMPs, VFS and CF were the most effective in reducing nutrients exports at the watershed scale. In terms of land-use management practices, RIA showed the highest reduction of nutrients and also the best CE ratio for this type of BMPs. Therefore, prosecuting illegal agriculture can be a relevant action to improve the Mar Menor ecological status. The combination of several BMPs showed a synergistic effect and was the best solution for reducing NPS pollution loads inputs into the Mar Menor. The scenario with the most implemented BMPs of this study (AllBMPs) gave the highest reduction percentages. Thus, the results obtained in this study can have important implications in selecting the most appropriate BMPs to effectively reduce the NPS pollution loads that flow into the Mar Menor coastal lagoon. Mar Menor's law enforcement may positively impact the ecological status of Mar Menor, although the most effective and cost-effective measures should be prioritised over others. Future research could consider identifying critical areas of intervention for a more effective implementation of BMPs and simulating other potential scenarios. The modelling approach, simulated BMP scenarios and outcomes of this study can be applied to similar coastal lagoon areas also affected by highly anthropogenic watersheds, guiding decision-makers in controlling NPS contamination issues.

CRediT authorship contribution statement

Adrián López-Ballesteros: Conceptualization, Investigation, Methodology, Writing – original draft. Dennis Trolle: Supervision, Writing – review & editing. Raghavan Srinivasan: Supervision, Writing – review & editing. Javier Senent-Aparicio: Conceptualization, Project administration, Supervision, Writing – review & editing.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Editor: Jurgen Mahlknecht

Data availability

Data will be made available on request.