“Assessing payment for ecosystem services to improve lake water quality using the InVEST model”

Payment for ecosystem services is a conservation strategy designed to offer farmers financial incentives for managing land to provide ecological benefits without disturbing livelihoods. However, the distribution of spatial financial feasibility is challenging when implementing this strategy on watershed scale. This study aimed to develop payment for ecosystem services model to improve quality in lake water catchment. The model estimated incentive values based on the costs of farmers’ losses, water yields, and pollution loads. The potential loss was calculated by determining the income of farmers in lake water catchment spent on land conversion from intensive agriculture to agroforestry. Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) modeling tool was used to calculate water yield and pollution load. The model was tested with case study approach at Lake Rawa Pening in Indonesia, consisting of nine sub-basins and 75 village administrations. The results showed that the reference compensation for farmers was 1,255.97 USD/ha/year. Considering the spatial distribution of water yields, the incentive for each village varied widely from 891.54 USD/ha/year to 1,557.06 USD/ha/year, even within the same sub-basin. Ten villages had an incentive above 1,450.00 USD/ha/ year. However, considering the water pollution load, 26 villages had an incentive above 1,450.00 USD/ha/year with a maximum of 2,024.17 USD/ha/year. Therefore, village boundary should be an analysis unit for determining spatial incentive feasibility rather than a sub-basin boundary. Moreover, the level of water pollution load can become an additional variable to justify the amount of incentives received by farmers.


INTRODUCTION
Payment for Ecosystem Services (PES) is an environmental economic instrument designed to incentivize land users for engaging in environmentally friendly behavior, particularly agricultural land management.PES schemes use the principle that users or beneficiaries should compensate those who provide ecosystem services, such as water purification or carbon sequestration.This concept requires clearly defining the environmental service supply, understanding the influence range of services, identifying stakeholders, and establishing the appropriate incentive.
PES approach has been widely studied and implemented in various contexts globally, reflecting the adaptability and potential for addressing environmental challenges through economic mechanisms.However, challenges to implementation on watershed scale still exist, including addressing perceived unfairness, overcoming financial pressures, navigating the complexity of establishing payment schemes, balancing economic growth with environmental conservation, as well as ensuring social acceptability and spatial financial feasibility.Addressing these challenges is crucial for the successful implementation and sustainability of PES initiatives.

LITERATURE REVIEW AND HYPOTHESES
Assessing ecosystem services entails evaluating the benefits offered to humans.Assessment process typically includes identifying and quantifying ecosystem services, understanding the value to human societies, and determining sustainable management.Mapping and Assessment of Ecosystem Services (MAES) provides a comprehensive approach to understanding the capacity to provide many services essential for human wellbeing and environmental sustainability (Sieber et al., 2022).Numerous ecosystem services are beneficial to humans, such as providing water and food (Ioannidou et al., 2022), controlling flood and erosion (Udawatta, 2021), regulating climate (Pandey & Ghosh, 2023), serving tourism (Rosehan et al., 2020), and presenting educational facilities (Banela et al., 2024).However, the main challenge to maintenance is balancing human development and environmental conservation, such as managing agricultural land to provide profits without damaging the environment (Kremen, 2020).
Human activities could affect sustainable ecosystem services; for example, agriculture negatively influences water quality (Anderson et al., 2021), biodiversity (Sharma et al., 2018), carbon stocks, and soil retention (Wang et al., 2022).Consequently, the growth of agricultural land, specifically intensive agriculture, is often considered an enemy of sustainable water ecosystem services due to the production of pollutants, including fertilizer and chemical inputs (Rashmi et al., 2020), as well as the tendency to reduce water supply through irrigation (Munyaradzi et al., 2022).Increasing agricultural land is urgently needed to meet the rising population growth estimated to reach 9.7 billion globally by 2050 (UN, n.d.).
Economic valuation of ecosystem services in lake water catchment (LWC) is one of the efforts to integrate water and food demands.LWC has an important role in maintaining the quantity and quality of lake water because it functions as a natural basin to collect rainfall and channel into lake.However, the system tends to be vulnerable due to intensive agriculture, for example, farming cereal and peas influences spatial and seasonal variability in lake water balance (Tigabu et al., 2019).
Additionally, the use of pesticides on agricultural land in the upper basin caused water pollution (Jayawardana et al., 2023).Moreover, changes in one ecosystem service affect others, for example, efforts to increase the supply of clean water could reduce the capacity for flood control and water purification (Wang & Xu, 2023).These results showed that the nexus of human activities and water ecosystem services at watershed scale was still challenging; hence, trade-offs between ecosystem services must be managed wisely.
Several previous studies have discussed the tradeoff between the provision of ecosystem services and the impact of human activities, where compensation was given to parties willing to spearhead improvement and maintenance, in the form of PES.The PES scheme could be an alternative environmental economic instrument for attracting people in LWC to participate in efforts to reduce pollution load.The society uses the LWC as a space for crop cultivation, negatively impacting lake water quality (Jayawardana et al., 2023).The conflicts between agriculture and the provision of water ecosystem services upstream can not be avoided, but sustainable land management might minimize these conflicts (Zhao et al., 2023).Accordingly, LWC might be used for cultivating agriculture and providing water ecosystem services when PES model recognizes agricultural activities as providing food and controlling pollution loads entering lake waters.Agriculture design should be agroforestry system due to the positive correlation with improving water quality (Ye et al., 2023).Moreover, increasing forest cover can improve water quality due to the function as pollutant filters, reduce erosion, and maintain the hydrological cycle (Qiu et al., 2023).For example, increasing forest cover by 1% of the basin can reduce water turbidity by 3%, while raising built-up land by 1% increase turbidity by 3% (Warziniack et al., 2017).
Converting intensive agriculture into agroforestry requires conversion costs and compensation due to the loss of farmers' income (Paudel et al., 2022;Wondimenh, 2023).Consequently, compensation is needed for sacrificing land to support conservation measures.Previous studies showed that people tended to reject compensation as the value did not match the losses incurred (Nuñez Godoy & Pienaar, 2023).PES schemes that do not consider social justice, such as the distribution of benefits and risks and community needs that could lead to refusal in supporting conservation programs (Lliso et al., 2021).To overcome this gap, there is an urgent need to assess the compensation scheme that considers economic losses and the risk of land conversion.For example, the land with the highest water yield and pollution load should have more compensation than other places.

METHODS
The study area was Lake Rawa Pening, located in Central Java province, Indonesia, with a total basin area of 27,307.25 hectares (ha).This area consists of 1,994.93 ha of lake water and 25,312.32ha of catchment (Figure 1).A total of nine main rivers feed lake, namely Galeh, Kedungringin, Legi, Panjang, Parat, Rengas, Ringis, Sraten, and Torong.On the other hand, only one outlet exists, namely the River Tuntang.
Rawa Pening is one of the 15 National Priority Lakes for the central government due to the important roles and functions as agricultural irrigation, the source of drinking water, hydropower plants, flood control, and tourism destinations.However, anthropogenic activities in lake basin caused water pollution.Data from the Environment and Forestry Department of Central Java Province (2023) showed that Rawa Pening was lightly polluted in 2018-2022.This pollution emanated from various sources, including agricultural, livestock, domestic, and aquaculture waste.Land cover in the basin was dominated by agricultural land with a total area of 16,028.41ha (58.76%), consisting of dryland (12.03%), mixed dryland (33.20%), and rice fields (13.53%).The land used for settlement was estimated at 6,617.91 ha (24.26%), while the forest cover (secondary and plantation forest) was relatively small, with an area of 1,561.66ha (5.72%).
There are four steps to set up the MPES model (Figure 2):  (1) calculate the average income of farmers in the LWC, represented by the value of agricultural production per hectare for one year; (2) calculate the potential water yield in the LWC; (3) assess the potential pollution load in the LWC using total phosphate (TP) parameters; and (4) estimate the PES value by considering farmer's loss, water yield, and pollution load.
This model selects statistical analysis to analyze agricultural production, while the InVEST 3.14.1 software was used to assess water yield and pollution load.The output was presented using ArchMap 10.8 software.The unit analysis of the result was village administration boundary.

Potential loss of farmers
To estimate the potential loss of farmers for implementing agroforestry, a proxy of the average economic value of agricultural production per hectare in the LWC for one year was used.Data from the Bureau of Statistics of Semarang Regency (2022a, 2022b, 2022c, 2022d, 2023) were selected as a source of information on agriculture in Lake Rawa Pening, with the products being vegetable and food crops.The total economic value of agricultural production in the LWC for a year was divided by the size of used land to obtain the average value of output for one year.The average economic value represents the revenue of farmers per hectare per year.The revenue from agriculture was selected as a reference source to obtain the potential loss of farmers.According to Asfawi et al. (2021), the cost of intensive farming was 42% of the revenue and the net income of farmers was 58%.Pissarra et al. ( 2021) considered 50% of farmers' net income to calculate the loss of farmers.Therefore, in this model, the potential loss of farmers was 0.29 of revenue (trade-off coefficient).Calculations to obtain the potential loss of farmers are as follows: where: Lc: Potential loss of farmers per hectare per year (USD/ha/year); Pc: Average revenue of farmers per hectare per year (USD/ha/year); NE: Total revenue of farmer for one year (USD/year); L: Total area of agricultural land used for agricultural cultivation (ha); Ti: Total agricultural output per year on commodity i (kg); Hi: Average price of agricultural products for commodity i (USD).

Water yields
The InVEST model for annual water yield (AWY) was used to estimate water yields in the LWC.This model is based on water balance method and assumption of Budyko (1974) for water and heat balance.The primary principle is that the higher water production, the greater water supply for lake waters.
where Y(x): Average annual water yield (mm); AET(X): Average annual evapotranspiration (mm); P(x): Average annual rainfall (mm); R(x): Budyko drying coefficient; PAWC (x): Water content in plants, which by the amount of water in the soil used by plants (mm); Z: Zhang coefficient was obtained from calculations based on the Budyko curve; W(x): Non-physical parameters; ETo: Annual evapotranspiration.
Model result was compared with observed data of water discharge downstream feeding lake for validation.Based on the number of the main rivers feeding lake, nine monitoring stations of water discharge were selected.The observed data in 2022 were collected for the dry season (July 12, 2022) and the rainy season (November 10, 2022) obtained from the Environment and Forestry Department of Central Java Province.The model result was in units of m3/year, while the observed data were in units of m3/second.For comparative consistency, observed data were converted into m3/year.The average percentage of bias between

Water pollution load
According to the Indonesia Minister of Environmental Regulation Number 28 of 2009, the total phosphate (TP) concentration in water could determine lake's trophic status.Previous studies also reported that the TP parameter was an indicator of eutrophication (Wu et al., 2021).The level of water pollution was represented by the potential TP pollutant (TP nutrient export) in the basin.In this study, the InVEST model for nutrient delivery ratio (NDR) was selected to calculate the TP pollutant.This model uses a mass balance approach that describes the amount of pollutants in the basin but does not represent the nutrient cycle in detail.It represents the amount of nutrient concentration in each land pixel for a year, influenced by land use, runoff, and pollution load coefficient in each land use (Sharp et al., 2016).The equation for calculating TP pollutant is presented in formula (10).The data needed for this model included watershed boundary, land use map, digital elevation model (DEM), precipitation (nutrient runoff proxy), flow accumulation threshold, Borsellu K, and biophysical table, as shown in Table 1.
The result was compared to observed water quality data to validate the InVEST model.A total of nine monitoring stations were used based on the number of main rivers feeding lake.The observed data were obtained in 2022 during the dry season (July 12, 2022) and the rainy season (November 10, 2022) from the Environment and Forestry Department of Central Java Province.The result was in units of kg/ year, while the observed data was in units of mg/liter.For comparative consistency, observed data were converted into kg/year.The average percentage of bias between model result and observed data was calculated based on the method of Moriasi et al. (2007) as presented in formula (8).
The level of pollution load was calculated based on the average TP pollutant in every village (TC).The value of TC was classified into four classes, namely lightly (class 1), moderately (class 2), heavily (class 3), and extremely polluted (class 4) with a TP value < 0.0100 kg/ha/year, 0.0100-0.0119kg/ha/year, 0.0120-0.0139kg/ha/year, and > 0.0139 respectively.The TC value reflects the impact (increase) of land conversion costs, contributing to pollution load.In this study, the increase amounted to 1% (class 1/no increase), 10%, 20%, and 30% for classes 2, 3, and 4 respectively.

PES value
PES value is a function of farmers' income loss, water ecosystem services, and water pollution load.The loss of farmers' income (Lc) was calculated based on the revenue of agriculture in the LWC times the tradeoff coefficient (factor X = 0.29).Water ecosystem services (ES_AWY) were calculated based on the proportion of water yield per village (AWY) to the average water yield per village in the LWC (A̅ ̅ W̅ ̅ Y̅ ̅ ).Meanwhile, water pollution load (TC) level was calculated based on the increase in land conversion costs due to TP pollutants in land use.The calculation of PES to improve lake water quality (USD/ha/ year) is presented as:

RESULTS
Based on the calculated economic value of 14 commodities cultivated in the LWC of Rawa Pening, the average revenue of farmers was 4,329.97USD/ ha/year (Table F1, Appendix F).Considering the cost of cultivation and land conversion, the tradeoff coefficient was 0.29.Therefore, the average potential loss of farmers for implementing agroforestry was 1,255.69USD/ha/year.This value served as the basic reference for calculating the PES value.
The InVEST model result was the estimated water yield (AWY) (Table G2, Appendix G) and validation was carried out by comparing the results with the observed water yield (Table G1, Appendix G).The bias percentage in the Ringis, Ringas, and Legi sub-basin was significantly large, with values of 26.22%, 23.62%, and 17.44%, respectively.This was due to the relatively small size of the sub-basins, affecting water flow variation between the rainy and dry seasons.The smallest bias percentage was found in the sub-basins of Panjang, Torong, and Galeh, with the value of 0.68%, 1.10%, and 1.32%, respectively.The estimated water yield was generally greater than the observed in all subbasins, with a bias percentage of 5.63% (Table G3, Appendix G); hence, the result could be used to establish the MPES model.
Based on the results, the amount of estimated TP (Table H2, Appendix H) is lower than the observed in all sub-basins (Table H1, Appendix H) because the input data to estimate TP pollutants were only from non-point sources, such as agriculture and urban land.This calculation did not consider point sources, such as cage animal husbandry and industry.Consequently, the bias percentage of estimated TP was significantly large, reaching 44.26% (Table H3, Appendix H).This result is very logical despite the large bias percentage; hence, the estimated TP would be used to calculate the PES value.In contrast, the lowest AWY was found within villages in the sub-basin of Galeh, Panjang, and Rengas, with a value of less than 120 mm/ha/year.AWY value in every village varied even within the same sub-basin but the average was 135.66 mm/ha/year.The highest was found in Wates village (56), with an AWY of 168.46 mm/ha/year, while the lowest was in Sidomukti village (25), with 96.08 mm/ha/year.This result shows significant differences in AWY within the same catchment.A total of 10 villages had the largest AWY with a value above 155 mm/ ha/year (Figure 3a in red).
Based on Figure 3b, the spatial distribution of TP pollutants in the LWC was relatively diverse with the average TP in each village of 0.0101 kg/ha/ year.Furthermore, villages in the same sub-basin had a variety of pollution loads.For example, in the Parat sub-basin, some villages had high TP pollutants (Figure 5b, number 56) while others had low TP pollutants (Figure 5b, number 50).The largest TP pollutant was in Ngrawan village (49), at 0.196 kg/ha/year, while the lowest was in Salatiga ( 74) and Kalicacing village (60), with a TP of 0.0032 kg/ha/year.The different TP pollutants in every village may be attributed to land use, elevation, and runoff.
The ES_AWY in LWC, reflecting village's potential to supply lake water was calculated according to formula (9).The highest value was found in Wates village (56) at 1.24, while the lowest was in Sidomukti village (25), at 0.71.The spatial distribution of ES_AWY values was quite diverse, as presented in Figure 4a.The TC reflects the impact of land conversion costs contributing to pollution load.Based on Figure 4b, the most extreme TC values were found in 10 villages illustrated in red.These villages had higher precipitation with lower evapotranspiration due to the highlands.of ES_AWY values in the LWC (Figure 4a), the PES value for each village varied from 891.54 USD/ ha/year to 1,557.06USD/ha/year (Figure 5a), despite these villages existing in the same sub-basin.For example, in the Parat and Sraten sub-basins, the PES value per village ranged from 1,200.01 to 1,557.06USD/ha/year.This diverse value was caused by the different spatial distribution of climate change parameters, such as precipitation and evapotranspiration in every topography.This result supports H1 stating that the PES value for each village in the same sub-basin is different.
A total of 10 villages had PES values above 1,450.00USD/ha/year (Figure 5a in red).However, considering the TC, 26 villages had a PES value above 1,450.00USD/ha/year with a maximum of 2,024.17USD/ha/year (Figure 5b in red).With the TC consideration, the average PES value increased to 1,389.70 USD/ha/year from 1,255.86 USD/ha/year.This increase was caused by the potential pollution generated by land use, topography, and runoff.For example, upstream villages with massive human activities (agriculture) tend to have higher pollution load than downstream.This result supports H2 stating that the level of pollution loads could increase the PES value.The distribution of PES value within each village in LWC of Rawa Pening is presented in Table J1, Appendix J.

DISCUSSION
The MPES model was applied to the LWC of Rawa Pening, consisting of nine sub-basins with 75 villages.It estimated the compensation value to attract farmers based on water ecosystem services within each village in the LWC (Figure 4a).The reference compensation value calculated according to the potential losses of farmers due to land conversion was 1,255.69USD/ha/year.Meanwhile, water ecosystem services were estimated based on AWY obtained from the InVEST model.AWY was classified according to village boundary with a volume of water yield between 96.08 mm/ha/year to 168.46 mm/ha/year (Figure 3a).The value of water yields was obtained with a value range of 0.71-1.24(Figure 4a).Based on these calculations, the compensation value for each village in LWC of Rawa Pening varied widely from 891.54 USD/ha/year to 1,557.06USD/ha/ year (Figure 5a).Within one sub-basin, the compensation value for each village varied relatively because the value of ecosystem services was estimated based on modeling results that considered land topography, land cover, and precipitation.In previous studies, the value of water ecosystem services was calculated using water production approach in each sub-basin leading to each village in the same sub-basin having a similar compensation value (Pissarra et al., 2021).Therefore, the calculation of incentives using sub-basin analysis units could be less precise and probably cause bias in the value of compensation given to farmers.
Chen et al. ( 2022) estimated the value of ecosystem services by ecosystem area units, making implementation at the administrative area level difficult, specifically for ecosystem in more than one administrative area.In addition, other studies predicted the economic value of ecosystem services using the contingent valuation method (Guo et al., 2023;Admasu et al., 2024), where the value obtained depends on the perception (Lee & Kim, 2024), and the respondent's experience (Sulistiyono et al., 2023), without considering the actual value.Incentive assessment based on the results of the InVEST model with village administrative analysis unit could be a novelty in this study.This approach can reduce bias in calculating ecosystem services with a simple method of application at the policy level.
The people use LWC for intensive agriculture, which is not sustainable and produces pollutants capable of polluting water.Meanwhile, the MPES model estimates the compensation given to farmers for converting agricultural land into agroforestry to improve lake water quality.The maximum value of compensation ranged from 1,450.00USD/ha/year to 2,024.17USD/ha/year, with this value being influenced by the level of pollution load on the land (Figure 4b).According to Pissarra et al. (2021), the estimated PES value in the Fazenda Gloria basin (Brazil) ranged from 284.35 USD/ha/year to 749.45 USD/ha/ year depending on the level of land vulnerability.In the Santee River basin (USA), the value of PES was predicted to reach USD 4.6 million to USD 6.2 million per month, depending on geographic position, type of intervention, and environmental service targets (Ureta et al., 2022).
Similar results were also observed in the Xin'an River watershed (China), where the PES value in the upstream and downstream watersheds was 22.54 USD/month and 8.63 USD/month respectively (Li et al., 2022a).Li and Zipp (2019) estimated the value of PES to control Chesapeake Bay (USA) pollution between 35 USD/ha to 390 USD/ha, depending on pollution level.These results show that the value of PES for improving ecosystem services differs in each case, depending on local conditions and objectives.However, the MPES model is a suitable alternative for estimating PES values in other regions, provided that agricultural production, water yield, tradeoff coefficients, and pollution load levels can be adjusted to the characteristics.As the sm allest administrative unit, village boundaries can be a unit of analysis for determining PES values for easy application at the policy level.
Agroforestry system should provide ecological and economic benefits for the people to ensure that converting intensive agriculture into agroforestry could improve water ecosystem services in Lake Rawa Pening and be accepted by the community.Ecologically, agroforestry system must be able to control water pollution (Zhu et

CONCLUSION
The MPES model estimated the incentive to compensate for the loss of farmers by implementing agroforestry to improve water ecosystem services.The PES value was calculated based on the costs of farmers' losses, water yields, and pollution loads.Based on the results, the potential loss of farmers was approximately 1,255.69USD/ha/year.Considering these losses, farmers would receive PES value between 891.54 USD/ha/year and 1,557.06USD/ha/year.PES values varied even among villages in the same subbasins.Therefore, calculating PES value based on village boundary is better than a sub-basin boundary.
Considering the level of pollution load, the number of villages with PES above 1,450.00USD/ha/year increased to 26 from 10 with a maximum PES of 2,024.17USD/ha/year.This result proved that the level of pollution load would increase the PES value.
Agroforestry system should provide ecological and economic benefits for the community to overcome the failure of the MPES model, underscoring the need to select appropriate tree species and crops.Therefore, the mechanism for providing PES must consider planting and maintenance costs of agroforestry trees as well as the period for providing PES.The MPES model could be a reference for policymakers to improve lake water quality without disturbing community economic activities.Implementing this model is challenging due to budget availability and community participation, underscoring the need to indulge beneficiaries of improved lake ecosystem services as a funding source.To determine the costs and benefits of implementing PES, further studies are needed to compare the costs with the economic benefits of improving lake water ecosystem services.

Figure 3 .
Figure 3. Spatial distribution of the result of the InVEST model

Figure 4 .
Figure 4. Spatial distribution of ES AWC and TC

Figure 5 .
Figure 5. Spatial distribution of PES value

Table 1 .
Input parameters for the InVEST model PAWC is a water content in the plants, represented by the amount of water used in the soil.Description of PAWC value is presented in Table B1, Appendix B AWY Root depth Raster Li et al. (2022b) Root depth is a soil characteristic that functions as water storage.In that soil, there is 90% of root biomass.The depth value is influenced by soil type and land cover AWY

Table 1
The input data for water yield are land use map, root depth, evapotranspiration, plant available water content (PAWC), precipitation, Z parameter (Zhang coefficient), watershed boundary, and biophysical table.Each data item is explained in

Table 1 (
Moriasi et al. (2007)nd use class information.Each land use class has a pollution load value in units of kg/ha/year.This value reflects the estimated amount of pollutants produced by each land use NDR cont.).Input parameters for the InVEST model model result and observed data was calculated based on the method ofMoriasi et al. (2007):

Table B1 .
Zhou et al. (2005)ter Content (PAWC) is the water content in plants, represented by the amount of water in the soil used by plants.Calculation of PAWC values using the approach developed byZhou et al. (2005)is based on the composition of the soil layer (sand%, silt%, and clay%) and the organic matter (OM) of the soil layer for each land use.The soil layer data were obtained from the Local Management Unit of Pemali Jratun (2022).Calculations for PAWC are presented in formula (B1).Description of PAWC valueIn calculating precipitation value, the data are annual rainfall for 2018-2022.Only one climatological station is available in the Rawa Pening basin, so rainfall data from other monitoring points are needed.Rainfall data were obtained from the CHRS satellite to complete the data.Therefore, six rainfall monitoring stations are carried out: one from the Koppeng climatological station and five from CHRS.The data were processed by ArcMap (v10.8) for interpolation, projection, resample, mask by extraction, and presentation of spatial distribution. 12)p://dx.doi.org/10.21511/ee.15(1).2024.12

Table C1 .
Average annual precipitation in the Rawa Pening basin

Table D1 .
Table D1 contains land use class information.In each land use class, there is a planting depth (mm), a plant coefficient (Kc), and the presence or absence of vegetation cover in a land cover class (LULC-veg).Biophysical table Source: Li et al. (2022b).

Table G1 .
Data of observed water dischargeNote: * Observed water yields are calculated based on the average water discharge multiplied by one year (31,536,000 seconds).

Table G2 .
Data of estimated water yield

Table H1 .
Data on observed water quality

Table H2 .
Estimated data of the InVEST model

Table H3 .
Comparison of estimated and observed TP in Lake Rawa Pening