Assessment of the vulnerability of coastal agriculture to seawater intrusion using remote sensing, GIS, and Multi-Criteria Decision Analysis

ABSTRACT

The Philippines heavily relies on crop production, with a significant portion cultivated in coastal agricultural areas. However, these farmlands are impacted by salinity intrusion, which adversely affects agricultural productivity. This study assessed the coastal agricultural barangays in Kalibo (Aklan, Philippines) using the exposure, sensitivity, and adaptive capacity indicators, and following the Intergovernmental Panel on Climate Change (IPCC) framework. It aimed to enhance local and national stakeholders’ understanding of vulnerability to seawater intrusion. Physical, agroecological, and socioeconomic parameters were used as indicators, with weights determined through the Analytic Hierarchy Process (AHP). Various aggregation methods were used, and datasets from participatory methods and remote sensing technologies, including LiDAR, were integrated to evaluate vulnerability on a 500 × 500-meter grid. The assessment focused on coastal barangays selected by local officials and agricultural experts. Caano was the most vulnerable, while Bachaw Norte was the least. The exposure map highlighted Nalook, Estancia, and Caano as areas with the highest levels of exposure. The results were compared with the knowledge of city agricultural experts. A sensitivity analysis was conducted to assess the model for exposure indicators. This research contributes to planning and management approaches in Kalibo and advances knowledge of critical indicators in addressing the current challenges.

KEYWORDS:

• Vulnerability assessment
• coastal agriculture
• seawater intrusion
• analytic hierarchy process
• participatory methods
• sensitivity analysis

1. Introduction

The coastal aquifer is an important resource for irrigation in agricultural areas. However, both aquifers and coastal agriculture/farmlands are vulnerable to seawater intrusion, which can result in substantial adverse consequences for the aquifer and crop production. Seawater intrusion is the landward movement of saline water into freshwater aquifers arising due to natural conditions, climate variability, and/or human activities (Ivkovic et al. 2012). Movement of saline water intensifies during surface water consumption and contamination, excessive groundwater pumping, and withdrawals from coastal wells which influences the coastal agricultural areas and low-salt tolerant crops negatively, including changes in soil salinity, reduction in water availability, and decrease in crop yield (Duan 2016). Intrusion is also driven by an increase in sea level, rainfall shortages, and occurrence of storm surges leading to coastal inundation, and an increase of salt concentration in brackish water which can augment the effects of drought. Tully et al. discussed the five main drivers of seawater intrusion: (1) sea-level position relative to land and water table; (2) frequency and magnitude of storms and tides; (3) frequency and duration of drought; (4) water use patterns; and (5) hydrologic connectivity. Each of the drivers varies spatially, temporally, and in terms of frequency and duration (Figure 1). Consequently, salinization can occur gradually or rapidly, depending on the unique combination of drivers at a specific location (2019, 6:368–378).

Figure 1. Drivers of seawater intrusion and their impacts (Weissman, Tully, and McClure 2020; adapted from Tully et al. 2019, 6:368–378).

Irrigated agriculture accounts for 70% of worldwide freshwater use and withdrawal (Dar 2017; Duan 2016). Agriculture serves as the main livelihood and driving force of the Philippines’ economy, with most coastal communities depending on crop production (Aklan Provincial Government 2019). Regarding water usage, a high percentage is allocated to agricultural purposes. As of 2009, freshwater has become the primary source of water withdrawal, accounting for the highest proportion of global consumption (Frenken 2012). Furthermore, the rates and demands for freshwater continue to rise (Dar 2017; Duan 2016). In addition to freshwater demands, various climatic factors influence water salinity. These demands and climatic factors render coastal agriculture vulnerable to soil and water salinity, leading to direct and indirect impacts on agricultural production (Alam et al. 2017). Different crops have varying salinity thresholds or tolerances, and if crops with low salt tolerance are exposed to saline water, agricultural production will be highly affected. The constant intrusion of seawater also leads to the accumulation of salts in the soil, causing damage to land productivity and resulting in economic losses (Burger and Čelková 2003, 7:314–320; Dar 2017; Duan 2016; Nguyen 2016). Conducting a vulnerability assessment is significant to demonstrate the current and potential future vulnerability of coastal agriculture to seawater intrusion.

The Municipality of Kalibo (Province of Aklan, Philippines) has been selected as the study area (Figure 2). Kalibo is located on the northeastern border of Aklan and is bounded on the northeast by the Sibuyan Sea. Kalibo’s terrain is generally flat, with slope ranging from 0% to 3%, and favors both agricultural and urbanization growth. Land is mostly devoted to agriculture (62.2% of the total land area) and is planted with root crops, fiber crops, rice, coconuts, and other high value crops. Kalibo is situated in a Type III climate zone, which is characterized by a relatively dry period from November to April and a wet season for the rest of the year (FAO 2019). Agriculture, with rice as the primary crop, is the main source of livelihood in Aklan and Kalibo. In addition to rice, other high-value crops such as coconut, banana, mango, rambutan, and lanzones are also grown, contributing to the local economy (Aklan Provincial Government 2019). The land cover map indicates that rice is the most widely cultivated crop in Kalibo. The rice harvesting season was considered when selecting images, as it lasts for approximately 30 days from mid-September to October, within a 190-day cycle (Arnold 2004).

Figure 2. Study area in Kalibo with political boundaries and LiDAR-derived agricultural land cover map.

Kalibo has a total of 16 barangays, 7 of which are coastal barangays/farmlands or those barangays with agricultural lands near the shore or located in the coastal areas. While all barangays are considered, the assessment primarily focuses on the coastal barangays, namely, Bachaw Norte, Buswang Old, Caano, and Mabilo. These coastal agricultural barangays were selected by local officials and agricultural experts with firsthand experience on seawater intrusion. These areas are typically low-lying, making them susceptible to seawater intrusion.

The excessive pumping of groundwater from wells can lead to a decline and depletion of the water table resulting in the seawater to move inland and upward (The Groundwater Foundation 2022; USGS 2018). JICA (2012) highlighted the negative impacts of seawater intrusion on agricultural production, particularly on rice. To address these issues, different measures for mitigation including the development of irrigation systems, promotion of salt-tolerant crop species, and proper management of water resources have been recommended. However, these measures can be implemented effectively if the vulnerability of coastal agriculture to seawater intrusion is properly addressed, assessed, and understood.

Following the Intergovernmental Panel on Climate Change (IPCC) framework, this study assessed vulnerability in terms of the system’s degree of exposure to hazards, sensitivity to climate variability, and its adaptive capacity to environmental changes relevant to seawater intrusion (2017). Vulnerability is driven by the levels of exposure, sensitivity, and adaptive capacity. Systems with high exposure and/or sensitivity and lower adaptive capacity are more vulnerable. On the contrary, a system is less vulnerable when exposure and/or sensitivity is low and adaptive capacity is high. Therefore, to mitigate vulnerability, it is necessary to reduce the impact and/or increase adaptive capacity (Koh 2011, 411–427).

The assessment and developed models aimed at enhancing knowledge and increasing awareness of decision-makers, local and national stakeholders to seawater intrusion vulnerability for better planning and management. They will also serve as a guide in enhancing the community’s understanding of the indicators of seawater intrusion. The results are expected to provide information to farmers and local government units regarding the impacts of seawater intrusion on their farmlands.

2. Materials and methods

The vulnerability assessment of coastal agriculture to seawater intrusion considered the combined environmental, social, physical, and ecological contexts encompassed by climatic variations, and how they interact with each other (Bogardi, Birkmann, and De Léon 2005). The extent of assessment and/or unit of analysis is based on the 500 × 500-meter grid following the IPCC framework of vulnerability assessment (IPCC 2014, 2017).

The analysis of vulnerability factors considered the availability of relevant datasets. Exposure indicators included the agricultural land use/cover map, distance to shoreline, rate of sea-level rise, elevation of land area, vegetation and salinity indices, extraction rate and levels, and hydraulic property. Sensitivity and adaptive capacity indicators were administered to farmer household surveys. Weights for each indicator were determined using the Analytic Hierarchy Process (AHP) and were combined using different aggregation methods. Results were integrated to collectively assess the vulnerability. Maps were validated through local’s knowledge of hazard existence and the assessment of the model was tested through the sensitivity analysis following the one-at-a-time (OAT) method (Figure 3). Reference materials and relevant indicator datasets were utilized and served as inputs for the vulnerability component indices.

Figure 3. Methodological framework for assessing vulnerability of coastal agriculture to seawater intrusion.

2.1. Vulnerability indicators

Indicator-based assessment was conducted in the coastal barangays in Kalibo. Indicators of exposure, sensitivity, and adaptive capacity relevant to seawater intrusion have been selected and validated through participatory methods such as focus group discussions (FGD) and key informant interviews (KII). Indicators included the physical, agroecological, and socioeconomic parameters. The selection considered the following: each indicator must be agriculture-specific or related; indicator has predictive value to the assessment; availability and accessibility of the data; and the type of data, preferably, ordinal, interval, or ratio types.

The exposure index assesses the extent or degree to which coastal agriculture is vulnerable to seawater intrusion. The physical and agroecological indicators were considered in the exposure component and were estimated from models developed in this study (Table 1). Various sources of information, including data gathered from national and local offices, reports, expert opinions, field surveys, or a combination of these sources, were utilized.

Table 1. List of exposure indicators, concept, and importance in assessing vulnerability.

Exposure Indicators	Concept and Importance in the Assessment
Land use/cover map	Reveals Kalibo’s land use and cover types, including agricultural land location and extent. Useful for assessing vulnerability to seawater intrusion and identifying high-potential agricultural areas (Masum 2019; Nong et al. 2021).
Distance to shoreline	Measures proximity of agricultural land to the coastline. Critical for evaluating seawater intrusion risk and severity, particularly soil salinization (Hounsinou 2020; Kennedy 2012; Santha Sophiya and Syed 2013, 70:1197–1209; Tajul Baharuddin et al. 2013, 6:3909–3928).
Rate of sea-level rise	Tracks sea-level rise due to climate change. Increased rates heighten flood frequency and severity, escalating seawater intrusion risk in coastal agricultural zones (FAO 2015; Griggs and Reguero 2021; Mimura 2013, 89:281–301; Roy et al. 2022).
Elevation of land area	Indicates land height above sea level. Essential for identifying flood and saline water intrusion-prone areas, impacting drainage patterns and groundwater flow direction (Griggs and Reguero 2021; Hossain, Atibudhi, and Mishra 2023, 1:141–165; Neumann et al. 2000; Roy et al. 2022).
Vegetation index	Measures vegetation cover quantity and health in Kalibo, indicative of soil moisture, nutrient availability, and plant stress. Useful for assessing coastal agriculture vulnerability to seawater intrusion (Fang and Liang 2008, 2139–2148; Kennedy 2012; Nguyen et al. 2020; Steyer, Couvillion, and Barras 2013, 63:118–130).
Soil salinity	Measures soil salt content, affecting plant nutrient and water absorption. Useful for assessing crop productivity and seawater intrusion impacts (Corwin 2021; FAO 2015; Khamidov et al. 2022; Mazumder and Kabir 2022, 464–479).
Groundwater extraction rate	Quantifies water withdrawal from aquifers. Excessive extraction can lead to seawater intrusion, contaminating freshwater resources and impacting coastal agriculture (Alfarrah and Walraevens 2018; Bosserelle, Morgan, and Hughes 2022; Hounsinou 2020; Prusty and Farooq 2020, 3:61–74; Santha Sophiya and Syed 2013, 70:1197–1209; Tajul Baharuddin et al. 2013, 6:3909–3928).
Groundwater levels	Measures the distance between land surface and freshwater table in an aquifer. Declining levels increase vulnerability to seawater intrusion, affecting agricultural productivity (Bosserelle, Morgan, and Hughes 2022; Hounsinou 2020; Melloul and Goldenberg 1997, 51:73–86; Santha Sophiya and Syed 2013, 70:1197–1209; Tajul Baharuddin et al. 2013, 6:3909–3928).
Hydraulic property	Assesses soil and subsurface layers’ physical characteristics affecting water movement. Crucial for identifying areas at higher risk of seawater intrusion and planning adaptation measures (Alfarrah and Walraevens 2018; Bosserelle, Morgan, and Hughes 2022; Jasechko et al. 2020; Santha Sophiya and Syed 2013, 70:1197–1209).

When assessing the vulnerability of coastal agriculture to seawater intrusion, it is important to consider both surface and subsurface factors. Surface analysis includes data on land cover, elevation, and groundwater extraction rates. Subsurface analysis includes indicators such as hydraulic conductivity and groundwater levels. However, some of the subsurface data were either outdated or unavailable and may have limited utility in assessing the vulnerability of groundwater and aquifers to seawater intrusion. Thus, the study primarily focused on assessing soil and crop vulnerability rather than subsurface factors.

Sensitivity captures characteristics of farmer households and crops that are influenced by the vulnerability associated with seawater intrusion. The assessment considers various sensitivity indicators (Table 2). On the other hand, adaptive capacity refers to the ability of agricultural lands to adapt with seawater intrusion. It reflects the resilience and adaptability of farmer households in responding to, coping with, and recovering from the impacts of hazards (Weis et al. 2016, 136:615–629). Indicators of adaptive capacity vary and include factors such as income from non-agricultural activities, savings, access to technology and information, economic capability, institutional capability, availability of disaster and hazard policies, and crop selection (Table 3). The evaluation of socioeconomic indicators was conducted through household surveys, which included the assessments of crop production and environmental awareness.

Table 2. List of sensitivity indicators, concept, and importance in assessing vulnerability.

Sensitivity Indicators	Concept and Importance in the Assessment
Population density	Represents the number of individuals per square kilometer. Indicates human pressure on coastal agricultural lands, affecting water resource use and land use change (Carnegie et al. 2022; Edwards 2011; National Geographic Society 2022; Tuong et al. 2003, 1:65–73).
Dependence ratio	Measures the ratio of dependent or unemployed household members to working members. A higher ratio may indicate greater vulnerability to changes in the agricultural system (Edwards 2011; FAO 2015; Hayes 2003; Hoffmann, Šedová, and Vinke 2021; Zhao, Xiang, and Zhao 2023, 166:239–267).
Membership in farmer organization	Refers to farmer household participation in farmer organizations. Indicates the local ability to cope with and adapt to the impacts of seawater intrusion (Agriculture Improvement Act of 2018; Castillo and Simnitt 2023; FAO 2017; Goedde, Ooko-Ombaka, and Pais 2019; Hurst, Termine, and Kar 2007; OECD 2001).
Hunger incidence	Acts as a proxy for food security, based on whether the household has experienced hunger or food shortage in the past three months (FAO 2022, 2023; Mbow et al. 2019; USDA 2022).
Percent agricultural income	The percentage of total household income derived from agriculture. A higher percentage can signify greater vulnerability to seawater intrusion (Danso-Abbeam, Dagunga, and Ehiakpor 2020; FAO 2017; Hurst, Termine, and Kar 2007; Jayne, Yeboah, and Henry 2017; Rapsomanikis 2015; Senadza 2012, 24:233–244).
Percent debt/loan	The percentage of a farmer's income used to repay debts or loans. High levels may indicate difficulty in investing in adaptation measures (Castillo and Simnitt 2023; FAO 2022; Rapsomanikis 2015).
Tenurial status	Indicates the household's agricultural land management, such as ownership, lease, or tenancy. Impacts farmers’ resilience, adaptive capacity, and resource access (FAO 2004; Kime 2016).
Access to post-harvest facilities	Refers to the household's access and use of post-harvest facilities. Influences farmers’ resilience to hazards by affecting their capacity to store and transport produce (FAO 2018).
Access to roads and bridges	Pertains to infrastructure and transportation networks access for product and equipment transport. Affects farmers’ resource and market access, impacting their adaptability (FAO 2004; OECD 2001; Starkey and Hine 2014; White and Kaplan 2017).
Level of education	Represents the highest educational attainment of household members. Influential in assessing vulnerability as it affects access to information, technology adoption, and decision-making (Pauw et al. 2015, 11:15693–15717; UNESCO 2017).

Table 3. List of adaptive capacity indicators, concept, and importance in assessing vulnerability.

Adaptive Capacity Indicators	Concept and Importance in the Assessment
Percent non-agricultural income	The proportion of a household's income from non-agricultural sources. Indicates income diversification and capacity to cope with agricultural losses (Danso-Abbeam, Dagunga, and Ehiakpor 2020; FAO 2017; Hurst, Termine, and Kar 2007; Jayne, Yeboah, and Henry 2017; Rapsomanikis 2015; Senadza 2012, 24:233–244).
Household assets/savings	The total value of a household's assets and savings. Reflects financial resources available for adaptation measures against seawater intrusion impacts (Rapsomanikis 2015).
Access to training courses	Availability and participation in programs and training courses related to climate, farming, and agriculture. Indicates readiness to cope with seawater intrusion impacts (Coman et al. 2020; Hurst, Termine, and Kar 2007; OECD 2001).
Access to financial aid	Availability of financial resources from institutions, government, and non-government agencies. Helps evaluate capacity to obtain alternative livelihood options or invest in adaptation measures (BSP 2020).
Means of transportation	Modes of transportation available for moving crops, equipment, and inputs. Impact’s ability to access markets, purchase inputs, and transport crops (Kaiser and Barstow 2022).
Access to alternative water sources	Availability and accessibility of water sources other than groundwater or surface water. Determines vulnerability of coastal agriculture to seawater intrusion impacts (Sasakova et al. 2018; White and Kaplan 2017).
Access to water treatment	Availability and capacity of facilities or services that treat water contaminated by seawater intrusion. Critical for determining coastal agriculture's vulnerability to seawater intrusion (Sasakova et al. 2018).
Farming experience	Number of years a farmer has practiced agriculture. Evaluates the farmer’s ability to adapt to changes caused by hazards (Coman et al. 2020; Jayne, Yeboah, and Henry 2017; OECD 2001).
Crop selection based on hazard events	Ability to choose alternative crops less vulnerable to saltwater damage. Enhances resilience of coastal agriculture to seawater intrusion impacts (FAO 2015, 2021).
Agricultural products are covered by insurance	Entitlement to insurance for crops or agricultural products damaged by seawater intrusion. Reduces financial risks and increases resilience to this hazard (USC 2023).

2.2. Data collection and analysis

2.2.1. Collection of socioeconomic data from farmer household surveys

The collection of socioeconomic data from household surveys was carried out using a random sampling strategy. Socioeconomic data was analyzed and used as inputs in determining the coastal agriculture’s sensitivity and adaptive capacity to the impacts of seawater intrusion. Assessments for these components were limited on the surveyed farmer household data and had greater reliance on expert judgement in comparison with the exposure assessment.

The respondents were households engaged in agricultural activities, excluding those involved in poultry and livestock. A 99% confidence interval and a 10% margin of error were used to determine the household sampling size, with the distribution of the target households based on an 80% response rate. However, the initial household survey did not yield enough respondents due to low awareness of seawater intrusion and some respondents declining to participate as their farmlands were not yet affected. As a result, the survey was revised to focus on barangays that have experienced or are currently experiencing seawater intrusion in their farmlands. These are the coastal barangays that have been verified by the local chairpersons or barangay captains with seawater intrusion experience. Out of the 136 targeted households, only 75 (55%) participated and provided socioeconomic data that is necessary for assessing sensitivity and adaptive capacity indicators.

2.2.2. Agricultural land use and analysis of the physical and geographical indicators

The information on the relevant land use and cover types in Kalibo was derived from the agricultural land cover map generated from the classification of Light Detection and Ranging (LiDAR) data using object-based image analysis (OBIA). Information on the extent and location of agriculture land use in the coastal zone is important in identifying agricultural lands closer to the coast as these areas may be more vulnerable to the effects of seawater intrusion, such as soil salinization which can have an impact on crop production (USDA Climate Hubs 2019; Weissman, Tully, and McClure 2020). A geographical parameter, the distance to shoreline indicator, was used in assessing the distance of farmlands from the coast, or the proximity of agricultural land to coastline. In this study, distance was estimated using coastal vignettes, which refers to the cartographic representations of the water-land boundary providing a visual representation of the distance between the coast and the agricultural land (Gopalakrishnan et al. 2019; Miller 2012).

The land cover map was updated to reflect land use, with non-agricultural classes excluded and the remaining classes were reclassified into three categories: trees, grain crops, and oil crops. The extent of agricultural land use within each grid and barangay boundary was then calculated to determine the total land use value within each grid (measuring 500 × 500 meters). The extent of each grid with agricultural area was also computed, and the results of the two calculations were used to determine the percent grid coverage of land use value to land use area.

Higher rates of sea-level rise leading to frequent and severe flooding can increase the risk of seawater intrusion into coastal farmlands. To assess the impact of rising sea-levels on farmlands, two raster datasets were created: one representing seawater, and the other showing the potential area of inundation. The seawater raster was generated by identifying areas that are at or below sea level, while the potential area of inundation was produced by identifying raster values that are at or below 2 meters, which is the highest recorded tide in 2019 for Kalibo based on tide times and tide charts (Tideschart 2019). To estimate the 2-meter inundation scenario, the sea-level raster was subtracted from the raster with elevations below or at 2 meters.

Assessing the elevation of agricultural areas is critical in determining their vulnerability to seawater intrusion and flooding caused by sea-level rise. It is also essential to consider the impact on its drainage patterns and the direction of groundwater flow. An area of low elevation will have a water table that is closer to the surface, making it more susceptible to seawater intrusion (Bhattachan et al. 2018; Hajji et al. 2022; USDA Climate Hubs 2019; Weissman, Tully, and McClure 2020). To quantify the sensitivity of an area based on its elevation, the elevation data (above mean sea level) was reclassified into five sensitivity classes, each represented by a numerical value: ‘5’ for very highly sensitive class (elevation <0.5 meters), ‘4’ for highly sensitive class (elevation between 0.5 and 1.5 meters), ‘3’ for medium sensitive class (elevation between 1.5 and meters), ‘2’ for low sensitive class (elevation between 2.5 and meters), and ‘1’ for very low sensitive class (elevation >3.5 meters). The value of the elevation indicator was calculated by multiplying the reclassified value (representing the sensitivity class) by the elevation grid area. To produce this indicator, the data was processed by excluding non-agricultural areas and reclassifying the remaining elevation values into the corresponding sensitivity classes.

2.2.3. Processing of vegetation and soil salinity indicators

Vegetation was used as an indirect indicator to predict and map soil salinity and with high consideration of vegetation salinity. Sentinel-2A images were used in the estimation. Salinity index highlights the surface reflectance of salt-affected regions and vegetation index highlights vegetated areas. Salinity and vegetation indices can help track salinity changes in vegetated areas (Azabdaftari and Sunar 2016; Elhag 2016; Taghadosi, Hasanlou, and Eftekhari 2019, 52:138–154). The Normalized Difference Vegetation Index (NDVI) value can be used to characterize the growth of rain-fed rice (Suwannachatkul et al. 2019). The downloaded and converted Sentinel-2A images were transformed into NDVI using Equation (1)(1) $\mathrm{NDVI}={\displaystyle \frac{\mathrm{NIR}-R}{\mathrm{NIR}+R},}$(1) : (1) $\mathrm{NDVI}={\displaystyle \frac{\mathrm{NIR}-R}{\mathrm{NIR}+R},}$(1) where $\mathrm{NIR}$ stands for the near-infrared band and $R$ for the red band. NDVI maps from August 2017 to December 2018 were analyzed to determine which image date best represents the pre-harvest season of rice in Kalibo. The expected NDVI values for the target state and percent cloud coverage were taken into consideration during this analysis. During the pre-harvest season, rice growth becomes dry, resulting in a drop in the NDVI value. It was also observed that the harvest season typically occurs in mid-September to October.

The analysis of image data revealed that September 2017 and February 2018 had the highest majority of NDVI values (0.75). However, the maximum value (0.99487) was observed in November 2017. Non-agricultural areas were masked out using a LiDAR-derived map and pixel values from 0.02 to 0.20 to ensure accurate NDVI values. The September images, which best represented the target state, were used in the vegetation index. This was further validated by focusing solely on the rice land cover (Figure 4).

Figure 4. The NDVI map for September 2018, focusing on rice land cover to emphasize the pre-harvest season. This map is utilized as an input for the vegetation index indicator. The index maps are generated from NDVI subsets, LiDAR-derived agricultural land cover, and a false color composite image.

In addition to the vegetation index, several salinity indices derived from Sentinel-2A were considered to detect and predict salinity in areas with high vegetation cover. These indices are shown in Table 4.

Table 4. Indices derived from Sentinel-2A that were utilized in the estimation of vegetation salinity.

Indices	Equation	References
Salinity Index	$\mathrm{SI}=\sqrt{B\times R}$	(Khan et al. 2005, 77:96–109)
Salinity Index	$\mathrm{SI}=\sqrt{R\times \mathrm{NIR}}$	(Dehni and Lounis 2012, 33:188–198)
Salinity Ratio	$\mathrm{SR}={\displaystyle \frac{R-\mathrm{NIR}}{G+\mathrm{NIR}}}$	(Dehni and Lounis 2012, 33:188–198)
Vegetation Soil Salinity Index	$\mathrm{VSSI}=2\left(G-5\left(R+\mathrm{NIR}\right)\right)$	(Dehni and Lounis 2012, 33:188–198)
Canopy Response Salinity Index	$\mathrm{CRSI}=\sqrt{{\displaystyle \frac{\left(\mathrm{NIR}\times R\right)-\left(G\times B\right)}{\left(\mathrm{NIR}\times R\right)+\left(G\times B\right)}}}$	(Scudiero et al. 2017, 71:231–238)
Normalized Difference Water Index	$\mathrm{NDWI}={\displaystyle \frac{\mathrm{NIR}-\mathrm{SWIR}}{\mathrm{NIR}+\mathrm{SWIR}}}$	(Gao 1996, 58:257–266)

B = blue, G = green, R = red, NIR = near-infrared, SWIR = short-wave infrared.

The Salinity Index (SI) is a measure of salinity that combines the blue and red bands and is sensitive to the surface reflectance of salt-affected areas with sparse vegetation cover (Douaoui, Nicolas, and Walter 2006, 134:217–230; Elhag 2016). Another combination for SI is the red and near-infrared bands, which is affected by salt crust, soil color and moisture content. Saline soils with a smooth and light salty crust surface exhibit high reflectance, while those with a coarse, dark, puffy surface show low reflectance (Asfaw, Suryabhagavan, and Argaw 2018, 17:250–258). The Salinity Ratio (SR) is computed using the ratio between the red and near-infrared bands and the green and near-infrared bands. The use of near-infrared bands can be highly correlated with field electrical conductivity, resulting in accurate estimations for satellite-based salinity detection. The blue and green bands are also used in this index, which can show an increase in the reflectance when salinity is present (Morshed, Islam, and Jamil 2016). However, cross-checking with the field data was not performed in this study.

Vegetation Soil Salinity Index (VSSI) is an index used to distinguish between soil and vegetation stress. Canopy Response Salinity Index (CRSI) is used to determine the salinity of the root zone. While CRSI is not a vegetation index specifically designed for salinity, it can be compared with other vegetation indices to provide better estimates of salinity levels. CRSI is used to estimate salinity based on canopy reflectance, where reflectance increases in the blue, green, and red ranges of the electromagnetic spectrum and decreases in the near-infrared range when the crop is stressed by root zone salinity, as noted by Scudiero et al. (2017, 71:231–238).

The Normalized Difference Wet Index (NDWI) is a useful indicator for plant water stress that is highly correlated with plant water content. It is derived from the near-infrared and short-wave infrared channels and is effective in enhancing the reflectance spectra of wet soil (JRC-European Commission 2011; Gao 1996, 58:257–266). High NDWI values indicate a high vegetation water content and a high fraction of vegetation canopy, while low NDWI values indicate a low vegetation water content and a low fraction of vegetation canopy. During periods of water stress, NDWI typically decreases due to a decrease in the amount of water in vegetation canopies, which is related to the soil water content (JRC-European Commission 2011). Unlike the NDVI, NDWI is less sensitive to atmospheric scattering effects (Noeda 2012), however, it does have a limitation: NDWI tends to overestimate water bodies due to its sensitivity to built-up structures (Sentinel Hub 2017).

The Combination Salinity Index (CSI) was generated by combining multiple salinity indices, including SI, SR, VSSI, CRSI and NDWI. Notably, NDVI was excluded from the index because CRSI, already integrated, prove more effective in detecting salinity than NDVI (Dehni and Lounis 2012, 33:188–198). The CSI is mapped based on the extent of the LiDAR-derived agricultural land cover map. This approach also incorporates information about built-up and water classes, addressing the overestimation of water bodies from NDWI.

To further complement the analysis, a crop height map was generated and combined with the NDVI map to distinguish trees based on their height levels and differentiate agricultural crops from other types of vegetation and other surface features in the study area of Kalibo (Figure 5). Additionally, the trend of salinity with respect to distance from the coast was examined. For this purpose, 21 zones were created with a 300-meter distance interval.

Figure 5. Approximation of agricultural crop height.

2.2.4. Estimation of groundwater and hydraulic property indicators

The groundwater extraction rate refers to the amount of groundwater withdrawn from an aquifer or well for human and agricultural use (Bierkens and Wada 2019; USGS 2018). To measure the quantity of water extracted in a particular area, the average monthly discharge diverted from the Aklan river irrigation system was used as an indicator. This is an important indicator to consider, as high rates of groundwater extraction, such as excessive pumping of wells, can lead to a decline in the water table and increase the likelihood of seawater seeping into the groundwater aquifers (Reilly et al. 2008; USGS 2019).

The groundwater levels indicator, which refers to the depth at which groundwater is found below the land surface, was estimated from JICA’s report in 2012, which recorded groundwater data levels in Kalibo. As seawater intrusion occurs, saltwater moves into the aquifer, displacing the freshwater and causing the groundwater levels to decline (Alfarrah and Walraevens 2018; Barlow 2003). The averages of the rainy season and dry season groundwater levels were interpolated to grid values, along with the locations of wells.

The hydraulic conductivity (K-value), a measure of a material’s ability to transmit water, is estimated using a model that considers soil type and permeability, with data sourced from the Department of Environment and Natural Resources – Mines and Geosciences Bureau (DENR-MGB) and the Philippine Rice Research Institute (PhilRice). The model, based on Arras (2014), estimates K-value using factors such as topography, flow path distance, recharge, and aquifer thickness. The study assumes a steady-state dynamic equilibrium, an unconfined aquifer, and groundwater flow following Darcy’s Law. The K-value is then calculated using Equation 2(2) $K={\displaystyle \frac{4W^{2}R}{(H^2-(H-d{)}^2)},}$(2) , where, K represents the hydraulic conductivity (length/time), W is the length of effective groundwater drainage (length), R is the recharge rate (length/time), d is the valley depth from channel water level (length), and H is the aquifer thickness (length). (2) $K={\displaystyle \frac{4W^{2}R}{(H^2-(H-d{)}^2)},}$(2) The results are summarized in Table 5. Note that this study focuses on the impact of salinity on crops and soil, not groundwater.

Table 5. Summary of estimated K-values based on various approaches.

Soil Type	K (m/s)
	Based on a Developed Model	Based on Soil Type	Based on Soil Permeability	Average of the Three K-Values
Bantog clay	1.57E-04	1.87E-06	2.50E-07	5.30E-05
Beach sand	1.52E-03	1.76E-04	1.00E-01	3.39E-02
Hydrosol	8.39E-04	8.86E-05	5.00E-06	3.11E-04
San Manuel clay loam	3.93E-04	2.31E-06	1.00E-01	3.35E-02
Sta. Rita clay	1.57E-04	1.28E-06	2.50E-07	5.28E-05
Umingan sandy loan	1.31E-03	3.45E-05	5.00E-06	4.50E-04

2.3. Quantification and normalization of indicators

The indicator parameters were measured both quantitatively and qualitatively. To enable comparison of indicators across barangays on the same scale, linear normalization was applied to the quantified indicator values (OECD 2008). Equation (3)(3) $Z_{\mathrm{ij}}={\displaystyle \frac{X_{\mathrm{ij}}-X_i^{\min }}{X_i^{\max }-X_i^{\min }},}$(3) was utilized for normalization of indicator values. (3) $Z_{\mathrm{ij}}={\displaystyle \frac{X_{\mathrm{ij}}-X_i^{\min }}{X_i^{\max }-X_i^{\min }},}$(3) where $Z_{\mathrm{ij}}$ stands for the normalized value of indicator $i$ of the barangay/municipality $j$, $X_{\mathrm{ij}}$ refers to the original value of the indicator $i$ of the barangay/municipality $j$, $X_i^{\max }$ is the highest value among all the barangays/municipalities, and $X_i^{\min }$ is the lowest value among all the barangays/municipalities.

The collected household surveys for sensitivity and adaptive capacity were quantified and their values were also normalized, including the exposure indicators.

2.4. Weight estimation using AHP

The AHP weighting method, proposed by Saaty (1980), was utilized to determine appropriate indicator weights while considering the 20% threshold consistency rate. The consistency check was calculated using Equation (4)(4) $\mathrm{Consistency}=\mathrm{MA\lambda }=1.55398\left(n-0.87026\right),$(4) : (4) $\mathrm{Consistency}=\mathrm{MA\lambda }=1.55398\left(n-0.87026\right),$(4) where $\mathrm{A\lambda max}$ refers to the maximum allowable $\mathrm{\lambda max}$ at the 20% threshold, and $n$ represents the number of indicators. Experts were requested to rate the level of importance of the indicators using Saaty’s rating scale, which includes a graduated system of numbers describing the intensity of importance (Table 6).

Table 6. The Saaty rating scale.

Intensity of Importance	Definition	Explanation
1	Equal importance	Two factors contribute equally
3	Somewhat more important	Experience and judgement slightly favor one over the other
5	Much more important	Experience and judgement strongly favor one over the other
7	Very much more important	Experience and judgement very strongly favor one over the other
9	Absolutely more important	The evidence favoring one over the other is of the highest possible validity
2, 4, 6, 8	Intermediate values	When compromise is needed

The relative importance for each indicator, including the rating of land cover classes, was determined by consulting 17 experts from national agencies, organization, academia, and local offices in Aklan, Kalibo and respective barangays using survey questionnaires. Only responses with consistency ratios within the 20% threshold were considered in the calculation of AHP, with a total of 4 responses for exposure, 4 for sensitivity, 6 for adaptive capacity, and 10 for land use/cover classes (Table 7). The consulted experts for AHP weights were categorized into five groups: (1) barangay experts, including heads from barangay offices (i.e. chairman, agriculture committee, agriculture technician, secretary); (2) city experts, including officers from offices in agriculture, planning and development, and disaster risk reduction management; (3) provincial experts from offices in agriculture and environment and natural resource; (4) experts from national agencies; and (5) experts from academia.

Table 7. Number of responses from experts and stakeholders.

Component	Number of Responses within the 20% Tolerance	Total Number of Responses
Exposure	4	12
Sensitivity	4	12
Adaptive capacity	6	12
Land use/cover	10	17

Weights were used to calculate the values for the exposure, sensitivity, and adaptive capacity indices. Different aggregation methods such as arithmetic mean, geometric mean, median, minimum, and maximum were employed to combine the individually calculated weights. These methods were used to identify any potential gaps in the complex individual information and to evaluate whether the combination of weights using different approaches would affect the output maps.

2.5. Composite vulnerability index assessment

Models were developed to assess the level of salinity in coastal farmlands. The assessment considered the analysis of multidimensional complex systems to easily comprehend the results as a range of values and rankings to aid planners, decision-makers, and stakeholders (USAID 2014). A composite index was created by integrating the indicators of exposure, sensitivity, and adaptive capacity to calculate vulnerability using Equation (5)(5) $\mathrm{Vulnerability}=\left[\left(\mathrm{Exposure}+\mathrm{Sensitivity}\right)\text{–}\mathrm{Adaptive}\mathrm{Capacity}\right],$(5) : (5) $\mathrm{Vulnerability}=\left[\left(\mathrm{Exposure}+\mathrm{Sensitivity}\right)\text{–}\mathrm{Adaptive}\mathrm{Capacity}\right],$(5) The vulnerability assessment of coastal farmlands in Kalibo included the generation of maps, using weights determined through AHP and aggregated using various methods. Maps and values were produced for each vulnerability indicator, and vulnerability estimates were derived based on the vulnerability values and the rating scheme outlined in Table 8.

Table 8. Vulnerability values and ratings.

The K-indicator, which introduced some uncertainty, generated multiple outputs, resulting in different versions of vulnerability assessments. K-values were estimated using: (1) a developed model, (2) soil type, (3) soil permeability, and (4) the average of the three K-values.

A total of 20 exposure maps, 5 sensitivity maps, 5 adaptive capacity maps, and 20 vulnerability maps were generated. The assessment of agricultural vulnerability to seawater intrusion covered all barangays, with a particular focus on Bachaw Norte, Buswang Old, Caano, and Mabilo. It is important to note that the exposure maps were able to assess both agricultural and non-agricultural coastal barangays.

2.6. Validation of maps and models

The vulnerability maps generated were compared with geotagged areas of seawater intrusion available on the LiDAR Portal for Archiving and Distribution (LIPAD) website. These geotagged areas were identified based on the expertise of city agricultural professionals who have knowledge of seawater intrusion occurrences over the past 5–10 years.

To assess the uncertainties and reliability of the models, sensitivity analysis was conducted using the OAT method. This approach evaluates how changes in individual input variables or indicators impact the model’s output. It assesses the robustness and extent of output variation when parameters are changed (Al-Mashreki et al. 2011, 9:1102–1111). Specifically, it helps identify which input variables or indicators have the most significant influence on the model’s results and assesses the model’s reliability (Kaplan Financial Limited 2020).

The behavior of the model and the impact of parameter changes were explored by varying weights assigned to exposure indicators. Eighteen different weighting schemes for each exposure indicator were used, resulting in 162 iterations with +/ – 10% variations. Throughout this process, the weights of other indicators were kept unchanged.

The sensitivity analysis primarily focused on exposure indicator values. The values were derived from the average of three K-values obtained using the arithmetic mean aggregation method. By applying ±10% weight variations to indicators such as land use/cover map, distance to shoreline, rate of sea-level rise, elevation of land area, vegetation index, soil salinity, groundwater extraction rate, groundwater levels, and hydraulic property, the study aimed to assess the model’s sensitivity.

3. Results and discussion

3.1. Characteristics of surveyed farmer households

Household characteristics show that most household participants are not members of any farmer or agriculture-related organization. Most farmer households own at least one tube well. 65.3% of household respondents indicated that their overall income comes from both agriculture and non-agriculture sources, but only 16% have attended trainings related to agriculture. Also, 64% of the households have their own tube wells and 54.7% have domestic water treatment (Table 9).

Table 9. Characteristics of participating household respondents from Bachaw Norte, Buswang Old, Caano and Mabilo.

Household Characteristics		Percentage	Number
Household size	1–5 members	89.3%	67
	6–10 members	10.7%	8
Highest education level	College/Post-graduate	65.4%	49
	High School	29.3%	22
	Elementary	5.3%	4
	No education	0.0%	0
Member of at least one farmer or agriculture-related organization? Farming experience	Yes	10.7%	7
	No	89.3%	68
	5 years or less	14.7%	11
	6–10 years	13.3%	10
	11–20 years	32.0%	24
	21–30 years	21.3%	16
	More than 30 years	18.7%	14
Household income	Non-agricultural sources	16.0%	12
	Agricultural resources	18.7%	14
	Combination of agricultural and non-agricultural resources	65.3%	49
Attended at least one agriculture-related training course?	Yes	16.0%	12
	No	84.0%	63
Household own at least one well (tube well)?	Yes	64.0%	48
	No	36.0%	27
With domestic water treatment?	Yes	54.7%	41
	No	45.3%	34

3.2. Relative importance of vulnerability indicators

Agricultural land use and vulnerability indicator weights are presented in Tables 10 and 11, respectively. In the case of land use, the arithmetic mean aggregation method was employed to combine the weights. Grain crops, such as rice and corn, received higher weights compared to oil and tree crops, which were assigned lesser weights. For vulnerability indicators, the weights were aggregated using various methods including arithmetic mean, geometric mean, median, minimum, and maximum. The resulting values indicate minimal differences, as the rankings provided by experts during the consultation process remained consistent.

Table 10. Weights of agricultural land use calculated from AHP and arithmetic mean aggregation method.

Agricultural land use classes	Weights (Arithmetic Mean)
Grain crops (rice, corn)	0.464
Oil crops (coconut, palm trees, etc.)	0.240
Tree crops (banana, citrus)	0.295

Table 11. Vulnerability indicators and corresponding weights determined using AHP and aggregation methods.

Indicators	AM^1	GM^2	Med^3	Min^4	Max^5
Exposure
Land use/cover map	0.057	0.043	0.034	0.054	0.082
Distance to shoreline	0.121	0.082	0.168	0.024	0.099
Rate of sea-level rise	0.185	0.120	0.125	0.075	0.276
Elevation of land area	0.098	0.092	0.106	0.110	0.087
Vegetation index	0.078	0.069	0.071	0.089	0.083
Soil salinity	0.130	0.109	0.127	0.101	0.142
Groundwater extraction rate	0.114	0.113	0.117	0.219	0.083
Groundwater levels	0.105	0.105	0.116	0.191	0.071
Hydraulic property	0.111	0.106	0.135	0.137	0.078
Sensitivity
Population density	0.078	0.071	0.07	0.083	0.086
Dependence ratio	0.084	0.075	0.085	0.074	0.085
Membership in farmer organization	0.082	0.075	0.081	0.096	0.079
Hunger incidence	0.119	0.107	0.119	0.135	0.113
Percent agricultural income	0.133	0.128	0.137	0.177	0.115
Percent debt/loan	0.104	0.084	0.077	0.086	0.14
Tenurial status	0.104	0.088	0.106	0.069	0.112
Access to post-harvest facilities	0.089	0.079	0.109	0.064	0.074
Access to roads and bridges	0.089	0.077	0.09	0.064	0.096
Level of education	0.119	0.111	0.126	0.151	0.101
Adaptive Capacity
Percent non-agricultural income	0.114	0.071	0.061	0.057	0.133
Household assets/savings	0.099	0.066	0.094	0.037	0.127
Access to training courses	0.084	0.079	0.116	0.112	0.058
Access to financial aid	0.069	0.063	0.087	0.09	0.05
Means of transportation	0.059	0.036	0.037	0.027	0.107
Access to alternative water sources	0.143	0.136	0.172	0.222	0.11
Access to water treatment	0.112	0.097	0.106	0.111	0.099
Farming experience	0.098	0.093	0.11	0.163	0.083
Crop selection based on hazard events	0.111	0.089	0.094	0.09	0.129
Agricultural products are covered by insurance	0.110	0.089	0.122	0.092	0.104

¹ Arithmetic mean; ²Geometric mean; ³Median; ⁴Minimum; ⁵Maximum.

3.3. Satellite-based estimation of salinity

Grain crop areas exhibit varying levels of salinity, ranging from medium to high (Figure 6). The extent and severity of salinity depend on both the crop type and its height. Inland zones, located 3600–5700 meters away from the coast, tend to have higher salinity values. This is likely due to the factors such as soil composition (Shahid, Zaman, and Heng 2018, 43–53), irrigation practices (Nick et al. 2024, 38:17), and historical salt accumulation (Shahid, Zaman, and Heng 2018, 43–53). In contrast, coastal areas experience a gradually decrease in salinity values. The proximity to the coast allows for leaching – the process by which excess salts are washed away by rainfall or irrigation. Coastal soils benefit from this natural leaching mechanism, resulting in lower salinity levels (Andros 2021).

Figure 6. Combination salinity index excluding non-agricultural areas shown as green areas in this map.

Furthermore, there exists a decreasing trend in salinity as crop height increases. The reduced salinity can be attributed to several factors such as root behavior (both root depth and soil surface layer) and increased transpiration in taller crops (Shelden and Munns 2023). However, beyond a certain height (specifically 18 meters), salinity abruptly increases – this may be influenced by factors such as root architecture and nutrient uptake.

3.4. Vulnerability assessment of coastal barangays

The vulnerability assessment of the coastal barangays was conducted using different aggregation methods (arithmetic mean, geometric mean, median, minimum, and maximum) and approaches to estimate K-values (based on a developed model, soil type, soil permeability, and an average of the three K-values). The computed vulnerability values for these selected coastal barangays are as follows:

1. Bachaw Norte: The vulnerability ranges from low to moderately low (≥ 0 to ≤ 0.4) across all approaches and aggregation methods.

2. Buswang Old: The vulnerability is moderately low (> 0.2 to ≤ 0.4) for all approaches in median and minimum aggregation methods, and ranges from low to moderately low for all approaches in arithmetic mean, geometric mean, and maximum.

3. Caano: The vulnerability ranges from moderately low to moderately high (> 0.2 to ≤ 0.8) for all aggregation methods, except in geometric mean, which has a rating of moderately low to moderate (> 0.2 to ≤ 0.6).

4. Mabilo: The vulnerability ranges from low to moderately low for all approaches used in the arithmetic mean, geometric mean, median, and minimum aggregation methods, and from low to moderate (≥ 0 to ≤ 0.6) for the maximum in all approaches, except for the results derived from the developed model.

The assessment indicated that Caano has a relatively higher vulnerability compared to other selected coastal barangays. The computed vulnerability values of coastal agriculture/farmlands, in terms of aggregation methods and different approaches used in estimating K-values, are illustrated in Figure 7.

Figure 7. Vulnerabilities for coastal agriculture/farmlands in terms of aggregation methods ((a) arithmetic mean (b), geometric mean (c), median (d), minimum (e), maximum) and different approaches used in estimating K-values. Note that the vulnerability levels are similar across the different K-value estimation approaches.

Each of the selected barangays exhibited unique characteristics contributing to their respective vulnerability levels (Figures 8 and 9). Bachaw Norte, despite its proximity to a river, exhibited the least vulnerability among the barangays, with levels ranging from low to moderately low. The vulnerability assessment focused on the distance of crops from the sea, not the river. Despite the area’s low adaptive capacity, its reduced vulnerability can be attributed to its low exposure and sensitivity. Various indicators, from land use to groundwater exposure, were considered in the assessment. The map predominantly shows low values, with the hydraulic property indicator being an exception.

Figure 8. Maps of coastal agriculture vulnerability in the barangays of Kalibo, Aklan, Philippines using various aggregation methods and K-value estimation techniques, including a developed model, soil type, and soil permeability: (a) arithmetic mean (b), geometric mean (c), median (d), minimum (e), maximum.

Figure 9. Components of vulnerability to seawater intrusion using various aggregation methods and K-value estimation techniques, including a developed model, soil type, and soil permeability: (a) arithmetic mean (b), geometric mean (c), median (d), minimum (e), maximum.

Buswang Old, located near a river and comprising significant fishpond areas, exhibited vulnerability ratings from low to moderately low. These ratings vary based on the aggregation method used, with notable differences between the minimum and maximum methods. The area’s land cover is predominantly oil crops. Despite lower impacts and adaptive capacity, there is potential for high vulnerability values under certain aggregation methods.

Caano, with its coastal fishponds and agricultural lands, was identified as the most vulnerable area. Vulnerability ratings range from moderately low to moderately high, except when using the geometric mean method, which shows moderate vulnerabilities. The exposure map reveals an increasing trend inland, with vulnerability ratings from low to moderate. The sensitivity map shows moderate ratings, while the adaptive capacity map indicates moderately low ratings. High exposure indicator values are observed for factors like sea-level rise, vegetation and salinity indices, hydraulic property, shoreline distance, and rice areas.

Mabilo's vulnerability ratings ranged from low to moderate, with variations based on the aggregation method used. The land cover includes coastal residential areas, rice crops, and large fishponds, with the latter two showing higher vulnerability. The hydraulic property indicator shows a moderate rating, while sensitivity and adaptive capacity are moderately low. Differences in vulnerability maps are observed between the minimum and maximum aggregation methods, especially in the exposure indicators.

3.5. Relationships between the potential impact and adaptive capacity

The computed vulnerability values and scatter plots of exposure, sensitivity, and adaptive capacity values of coastal barangays, based on the average of the three K-values, illustrate the relationship between the potential impact, determined by exposure and sensitivity components caused by seawater intrusion, and the adaptive capacity (Figures 10–14). Vulnerability is a function of exposure, sensitivity, and adaptive capacity, and a low adaptive capacity relative to exposure and sensitivity contributes to high vulnerability. On the contrary, higher adaptive capacity helps reduce the effects of exposure and sensitivity, resulting in low vulnerability (Thomas et al. 2019). The plots of exposure versus sensitivity are expected to have a steep inverse slope, while exposure versus adaptive capacity should have a steep slope to counteract the effect of exposure and achieve low vulnerability ratings, i.e. high exposure and low sensitivity, and low exposure and high adaptive capacity.

Figure 10. Vulnerability assessment of agricultural barangays in Kalibo, Aklan to seawater intrusion: (a) computed vulnerability values using the results from the arithmetic mean aggregation method based on the average of the three K-values, (b) scatter plot showing the relationship between exposure and sensitivity values, and (c) scatter plot illustrating the relationship between exposure and adaptive capacity values.

Figure 11. Vulnerability assessment of agricultural barangays in Kalibo, Aklan to seawater intrusion: (a) computed vulnerability values using the results from the geometric mean aggregation method based on the average of the three K-values, (b) scatter plot showing the relationship between exposure and sensitivity values, and (c) scatter plot illustrating the relationship between exposure and adaptive capacity values.

Figure 12. Vulnerability assessment of agricultural barangays in Kalibo, Aklan to seawater intrusion: (a) computed vulnerability values using the results from the median aggregation method based on the average of the three K-values, (b) scatter plot showing the relationship between exposure and sensitivity values, and (c) scatter plot illustrating the relationship between exposure and adaptive capacity values.

Figure 13. Vulnerability assessment of agricultural barangays in Kalibo, Aklan to seawater intrusion: (a) computed vulnerability values using the results from the minimum aggregation method based on the average of the three K-values, (b) scatter plot showing the relationship between exposure and sensitivity values, and (c) scatter plot illustrating the relationship between exposure and adaptive capacity values.

Figure 14. Vulnerability assessment of agricultural barangays in Kalibo, Aklan to seawater intrusion: (a) computed vulnerability values using the results from the maximum aggregation method based on the average of the three K-values, (b) scatter plot showing the relationship between exposure and sensitivity values, and (c) scatter plot illustrating the relationship between exposure and adaptive capacity values.

The results from the arithmetic mean based on the average of K-values show Caano as the most vulnerable with a rating of 0.3811 followed by Buswang Old with 0.2104, Mabilo with 0.1598, and Bachaw Norte with 0.1194 (Figure 10(a)). Caano was observed to be the most exposed, with the highest sensitivity and low adaptive capacity. On the contrary, Bachaw Norte exhibited the highest adaptive capacity and is tagged as the least vulnerable, least exposed and has low sensitivity compared to the other coastal barangays (Figure 10(b,c)).

Analyzing the results from other aggregation methods based on the average of K-values revealed that Caano was the most vulnerable, with ratings of 0.3571, 0.3976, 0.4357, and 0.3615 in geometric mean, median, minimum, and maximum respectively (Figures 11–14(a)). Relationships among the three components (exposure, sensitivity, and adaptive capacity) also demonstrated high vulnerability, as high exposure, high sensitivity, and low adaptive capacity were observed (Figure 10).

Based on the results shown in Figures 10–14, the relationships among the three components indicated that the coastal barangays in Kalibo are vulnerable to seawater intrusion. Coastal agricultural barangays with high exposure also exhibited high sensitivity and low adaptive capacity. The vulnerability assessment results are likely influenced by various factors such as access to resources, type of governance, culture, and knowledge of local and international communities/agencies (Thomas et al. 2019).

3.6. Assessment of maps through local’s knowledge

Tagged areas with seawater intrusion included the following barangays: Bachaw Norte, Bachaw Sur, Buswang Old, Buswang New, Mabilo, Briones, and Nalook (Figure 15). These areas are primarily located along the coast or at the mouth of the river. However, the assessment results show differences from the tagged areas, indicating that the assessment considered additional factors beyond just the proximity of farmlands to the ocean or the rising of water.

Figure 15. Areas identified by locals as experiencing seawater intrusion.

The selection of target barangays considered the knowledge and input of farmers and barangay officials. These stakeholders mentioned that seawater intrusion is affecting the productivity of farmlands in the barangays of Bachaw Norte, Buswang Old, Caano, and Mabilo. The intrusion is primarily caused by inundation resulting from rising water levels in both the ocean and river. The observed differences in the results can be attributed to the comprehensive nature of the assessment, which considered physical, agroecological, and socioeconomic indicators, rather than solely relying on distance or water level measurements. Furthermore, variations in estimations and observations among local farmers, officials, and agricultural experts were observed.

3.7. Assessment of model through sensitivity analysis

Sensitivity analysis results indicated the impact of modifying weights for different indicators on exposure values in the assessed barangays. When adjusting the weights for the land use/cover map indicator, there were decreased exposure values, with no significant differences observed in the coastal farmlands of Kalibo. Sensitivities were observed in all barangays when varying the weights for the distance to shoreline indicator, resulting in increased exposure in several barangays. Similar findings were observed for the rate of sea-level rise indicator, where increasing weights led to increased exposure in specific barangays.

The elevation of land area indicator showed decreasing exposure values in all barangays except for Poblacion. The vegetation index indicator resulted in decreased exposure values across all barangays, with slight changes observed in specific coastal farmlands. For the soil salinity indicator, increased exposure was observed in certain barangays, while the groundwater extraction rate indicator showed increased exposure in the +10% weight variations and decreased exposure in the -10% variations. The groundwater levels indicator exhibited decreasing exposure in all barangays, except for Linabuan Norte. Modifications in the weights for the hydraulic property indicator resulted in decreased exposure in all barangays, with specific variations observed in coastal farmlands. Overall, the sensitivity analysis provides insights into how variations in weighting schemes influence exposure values, highlighting the importance of each indicator in assessing coastal agriculture/barangays.

Figure 16 shows a summarized view of the exposure values obtained from the sensitivity analysis. It displays the changes caused by the application of ±10% weight variations to the exposure indicators. This figure provides a concise representation of the sensitivity analysis results, facilitating a quick assessment of the model's stability and robustness.

Figure 16. Summary of exposure values from sensitivity analysis using ±10% weight variations applied to land use/cover map (LULC), distance to shoreline (DTS), rate of sea-level rise (SLR), elevation of land area (ELEV), vegetation index (VI), soil salinity (SS), groundwater extraction rate (GER), groundwater levels (GL), and hydraulic property based on the average of the three K-values (K).

The sensitivity analysis of the models revealed that most indicators are sensitive to changing weight values. Through the OAT method, weights for each indicator were systematically modified, evaluating the resultant changes in exposure values. This analysis enables assessment of the sensitivity of the model to different weightings and highlights the relative importance of each indicator in determining the overall exposure assessment. Although the observed changes are minimal, they can still impact the resulting vulnerability assessment. It should also be noted that the implementation of the OAT method is relatively simple and does not account for assessing interactions between variables or the uncertainty of indicators, including potential model errors. It assumes that all other indicators remain constant, which is not realistic. Therefore, other sensitivity analysis methods, such as variance-based or regression-based methods, can be considered to account for interactions between variables (Business Case Preparation 2019; Kaplan Financial Limited 2020).

The assignment of weights applied in the vulnerability assessment is based on expert knowledge and awareness of the area, thus the generated maps and developed models can still be considered reliable. The results of the sensitivity analysis provide insights into how the assessment can be improved. Indicators with uncertainty and/or highly sensitivity should be weighted accordingly.

4. Conclusions and recommendations

This research assessed the vulnerability of coastal farmlands in Kalibo, Aklan, Philippines to seawater intrusion by considering exposure, sensitivity of resources, and adaptive capacity. The study utilized models and data from various sources, including agricultural maps, satellite imagery, and participatory methods such as interviews and surveys. The analysis included the development of salinity indices, estimation of vulnerability indicators, and generation of exposure, sensitivity, adaptive capacity, and vulnerability maps for selected barangays.

The assessment identified Caano as the most vulnerable barangay, while Bachaw Norte was identified as the least vulnerable. The exposure map highlighted Nalook, Estancia, and Caano as areas with the highest exposure levels to seawater intrusion. Sensitivity analysis provided insights for improving the assessment by testing different weighting schemes for indicators.

To facilitate the utilization of research outputs and relevant information, vulnerability maps were made publicly available through the PARMap Web Portal (https://parmap.dream.upd.edu.ph/) and the LiDAR Portal (https://lipad.dream.upd.edu.ph). This enabled local government units, stakeholders, and decision-makers to access the information and incorporate it into their decision-making process for developing programs and policies to reduce vulnerability and improve adaptability in coastal agriculture.

While the research utilized various indicators and available data, it acknowledged limitations due to the availability of surveyed farmer household data, records, and knowledge of locals and experts. To address these limitations, the study recommended expanding the selection of target areas by including farmlands near the coast and/or river. Additionally, socioeconomic data collection should consider recording the location of households to avoid generalizations in the assessment. In-depth and improved validation methods are also necessary, including intensive collection of groundwater data from wells and pumping stations, conducting pumping tests and georesistivity surveys in the field. It is crucial to consider geologic characteristics when estimating hydraulic conductivity, and detailed geologic maps should be acquired or requested from the relevant agency to enable a comprehensive assessment of subsurface, aquifer, and groundwater vulnerability, rather than focusing solely on soil and crop vulnerability.

Salinity intrusion to coastal agriculture necessitates effective mitigation strategies; a sustainable response combining nature-based solutions (NBS) with bioengineering. Coastal areas benefit from NBS, which include restoring mangroves, wetlands, and natural buffers – these ecosystems act as barriers against seawater intrusion, maintaining soil quality and supporting agriculture. Inland areas can also adopt NBS through sustainable land use planning, afforestation, and managed aquifer recharge – these approaches help maintain freshwater lenses and reduce soil salinity. Bioengineering offers promise in selecting and creating crops adapted to saline conditions. Salt-tolerant crop varieties thrive in high-salinity soils, ensuring food security even in challenging environments. Research and development efforts should prioritize strengthening the resilience of staple crops (such as rice) against salinity stress. However, challenges and limitations must be considered: (1) implementing NBS requires collaboration among stakeholders, policy adjustments, and financial investments; (2) bioengineered crops must undergo rigorous testing to ensure safety, yield, and nutritional quality; and (3) balancing NBS with conventional approaches remains essential, especially in large-scale agricultural systems (Tarolli et al. 2024).

Tarolli et al. (2024) suggest a synergistic approach: strategically planting salt-tolerant crops in NBS-implemented areas. This enhances soil health, increases yields, and improves livelihoods. This combined approach can support vulnerable regions worldwide, addressing hunger and climate change impacts. A conceptual framework (Figure 17) summarizes the mitigation solutions discussed. NBS and salt-tolerant crops offer hope for sustainable agriculture, safeguarding both food production and ecosystems.

Figure 17. A conceptual framework on sustainable solutions in mitigating soil salinization in agriculture (Source: Tarolli et al. 2024)

Overall, this research contributes to improving knowledge and awareness of seawater intrusion vulnerability in coastal agriculture and supports better planning and management approaches for decision-makers, stakeholders, and local communities. By addressing the limitations and incorporating further research recommendations, future assessments can enhance the accuracy and effectiveness of vulnerability assessments in coastal farmland.

Statements and declarations

Compliance with ethical standards

This paper adheres to all ethical standards and guidelines set forth by the relevant governing bodies.

Acknowledgements

The research presented in this manuscript was conducted as part of the R&D (Research and Development) initiatives at the University of the Philippines - Training Center for Applied Geodesy and Photogrammetry (UP-TCAGP). The project, PARMap (Agricultural Resources Extraction from LiDAR Surveys) under the Nationwide Detailed Resources Assessment Using LiDAR (Phil-LiDAR 2) Program, was funded by the Philippines’ Department of Science and Technology (DOST). The algorithms and workflows were developed utilizing the technical facilities and support provided by UP-TCAGP through the Phil-LiDAR 2 Program and PARMap Project. This support included the use of a detailed map of agricultural resources, including various crop types, derived from LiDAR and other remotely sensed data. The authors also acknowledge the Disaster Risk and Exposure Assessment for Mitigation (DREAM) Program and the Hazard Mapping of the Philippines using LiDAR (Phil-LiDAR 1) Program for the support in our research by allowing us to use the LIPAD website, which was instrumental in geotagging areas with seawater intrusion. The authors also recognized the Engineering Research and Development for Technology (ERDT) through the Faculty Research Dissemination Grant for sponsoring our publication charge. Reference materials and relevant indicator datasets were also gathered from various national agencies and government offices, including the: municipal boundary from the Philippine Statistics Authority (PSA), formerly known as the National Statistics Office (NSO); bathymetric LiDAR digital elevation model (DEM) and Interferometric Synthetic Aperture Radar (IFSAR) DEM from DREAM/Phil-LiDAR 1 Program; Sentinel-2 multispectral data; river discharge record of Aklan river irrigation system from 1977 to 2017 acquired from National Irrigation Administration (NIA) Region VI; groundwater data and well information from JICA and Local Water Utilities Administration (LWUA); and 2003 soil map from the Bureau of Soils and Water Management (BSWM). This manuscript acknowledges the researchers of the Phil-LiDAR 2 PARMap, consulted national and local experts, and the respondents for the farmer household surveys, all of whom have contributed to the attainment of the manuscript's objectives.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability

The maps and assessment results related to this study can be requested from the corresponding author. The data that supports the findings of this study are available on the PARMap Web Portal (https://parmap.dream.upd.edu.ph/). This portal serves as a repository for the knowledge products generated by the Phil-LiDAR 2 PARMap. Additionally, the LiDAR Portal (https://lipad.dream.upd.edu.ph/) provides primary access and distribution for the flood hazard and resource maps derived using LiDAR technology. For further inquiries, please direct your requests to the corresponding author.

Additional information

Funding

This research was made possible through the funding of the Philippines’ Department of Science and Technology (DOST) and monitoring by the Philippine Council for Industry, Energy and Emerging Technology Research and Development (PCIEERD), which had no influence on the study design, data collection, analyses, interpretation, or the decision to publish the results. Additional support was awarded by the Engineering Research and Development for Technology (ERDT) through the Faculty Research Dissemination Grant.