Water Scarcity, Mountain Deforestation and the Economic Value of Water in a Small-Scale Irrigation System: A Case Study in East Java, Indonesia

This study was conducted at Curahtakir village, in the Sanenrejo sub-watershed. It is located in the eastern region of East Java, in the sub-district of Tempurejo in the district of Jember. Figure 1 shows the location of the study area relative to Indonesia. Like a typical Indonesian village, Curahtakir is dominated by agriculture. Agriculture in the study area strongly depends on rainfall in the first season and water supplied by the irrigation in the second and third seasons. The average annual rainfall in the study area is 2,404 mm. The seasonal distribution of rainfall in the study area is 1,378 mm in the rainy season (November 2015–February 2016), 529 mm in the first dry season (March–June 2016) and 497 mm in the second dry season (July–October 2016). Agricultural irrigation was provided and managed by farmers utilising water coming from a spring.

Irrigation in the study area can be defined as gravity-flow hill irrigation; for detailed discussion on the technical and management aspects of this irrigation see Hill (2013, 2017).

Fig. 1

The altitude of the studied irrigation system is 200 metres above sea level. The studied irrigation system consists of three sections: an upper, middle and lower area. There are three cropping seasons annually in this area with relatively similar cropping patterns in each section: paddy in the first and second seasons and mostly maize in the last season.

Actually, the cropping season is divided based on the two main seasons in Indonesia, the rainy and dry seasons. The first cropping season is called the rainy season, while the second and the third season are called dry season 1 and dry season 2 respectively.

A full illustration of the studied irrigation area is shown in Figure 2. Water coming from a spring in the forest is stored in four dams: the Agung and Punco dams in the upper area and the Bloksadeng and Asmoya dams in the middle and lower areas. There are two other water sources in the middle and lower areas: two dug wells and eleven water pumps. Both middle and lower areas have one dug well, but only three water pumps are located in the middle area while the rest are located in the lower area. Many water pumps are required in the lower area since the water scarcity is high in the second and third seasons.

Fig. 2

Figure 3 illustrates the actual condition of each II. Figure 3A shows the Agung dam located in the upper area, which stores water from the spring. Figure 3B shows one of the water pumps (water pump 7) that pumps water from the drainage canal from the Asmoya and Bloksadeng dams (Figure 3C). Figure 3D shows one of the wells owned by a farmer (digging well 2). Figure 3E shows spring water from the forest in the mountain. Figure 3F shows substantial deforestation in the mountain forest, which causes a decrease in spring water quantity that leads to water scarcity.

Fig. 3

We collected data that represent both physical and economic water scarcities. The physical water scarcity is represented by the rainfall, streamflow and forest cover change data. These data are crucial because the primary water source in the studied SSI is rainfall and forest spring water. We used daily rainfall data from 2010 to 2015 from a rainfall station in Sanenrejo (the nearest station to Curahtakir). Additionally, we collected daily streamflow data from 2010 to 2015. Recently, the quantity of spring water started decreasing due to deforestation. The spatial data of land-use change in the Sanenrejo sub-watershed between 2011 and 2019 represent the most current land use in the studied area. The perceived economic value of water represents the economic water scarcity. We used the contingency value method (CVM) to calculate the economic value of water. Interviews with a sample set of farmers were used to collect information on the economic value of water and socio-economic characteristics of the farmer.

The research sample was determined using simple random sampling (SRS). SRS was used because the population of the study was identified. In the initial sampling stage, the farmer population was determined with the help of ulu-ulu.

Responsible for irrigation operation, ulu-ulu knows exactly how many farmers he is responsible for and where their plots are located.

A structured questionnaire was used to gather data by interviewing sample farmers from Curahtakir village. The main characteristic of Curahtakir agriculture is that it is irrigated by water from a spring utilising non-technical II. However, spring water can only serve the upper area in the dry season, forcing the middle and lower areas to use an irrigation water pump. Thus, the sample farmers used in this study were dispersed across all three areas. The questionnaire consisted of three sections. The first section focused on the social and economic characteristics of the farmer. This section was divided into four parts. The first part dealt with farmer identity. In this part, farmer age (X₂), education (X₃) and income (X₄) were identified. The remainder of the first section was structured as follows: the second part dealt with general farming conditions, the third part dealt with cultivation land status and size (X₁) and the final part dealt with the annual cropping calendar.

The second section focused on irrigation management in each cropping season. This section was divided into irrigation management in the rainy season, the first dry season and the second dry season. The focus was on the technical aspect of irrigation, such as water source, distance of farmed plots to a water source (X₅), information on who provides irrigation services and the service fees, the irrigation schedule and irrigation management in the dry season (D). In the dry season, irrigation management was a dummy variable used to categorise farmers who use the water pump and those who do not irrigate their land in the dry season. The last section of the questionnaire focused on eliciting farmers’ WTP for irrigation. This section consisted of 12 questions to explore how a farmer perceives the current irrigation price and the highest price. In this section, the farmer was also asked about what they would do if the irrigation price was beyond their ability to pay. Finally, this section also inquired about the role of ulu-ulu as a current irrigation provider.

The first step of the analysis was to determine the farmer's WTP. In determining WTP, we used the CVM. There are two ways to measure WTP: revealed preference and stated preference methods. The revealed preference method utilises the actual payment data to measure WTP, while the stated preference method measures WTP using direct surveys and is utilised when there is no adequate information about historical payments (Breidert 2006). CVM is the most commonly used method to measure WTP for non-marketed goods. This method has two main advantages: (1) its ability to evaluate proposed goods or services and (2) its usefulness in addressing values that cannot be dealt with any other way (Young 2005; Tang et al. 2013).

The second step of the analysis was to determine factors affecting farmers’ WTP. To do so, we used multiple linear regression with WTP as the dependent variable and the variables are shown in Table 1.

The multiple linear regression method estimates the nominal value of the farmer's WTP from the identified social, economic and technical characteristics. Our predicted variable is the WTP value obtained from the contingent valuation method, while X₁–X₅ and D are the explanatory variables. The multiple linear regression method was based on the least squares estimation and estimation performed with SPSS software (version 25; SPSS Inc., Chicago, IL, USA). The empirical model is shown in Eq. (2).

The empirical model contains a dummy variable representing two groups: 1 for those who use the water pump in the dry season and 0 for those who do not use. The final model consists of two Eqs (3) and (4). (3) WTP1=b0+∑i=15bixi+b6+e WT{P_1} = {b_0} + \sum\limits_{i = 1}^5 {{b_i}{x_i} + {b_6} + e} (4) WTP0=b0+∑i=15bixi+e WT{P_0} = {b_0} + \sum\limits_{i = 1}^5 {{b_i}{x_i} + e} where WTP₁ is the WTP of a farmer who uses a water pump to irrigate their crop during the dry season while WTP₀ is the WTP of a farmer who does not use the water pump in the dry season, b₀ is the regression intercept, and the other b(s) are the regression coefficients of the variables.

For eight years, the land-use pattern was relatively constant (Table 2). The most extensive land type in the Sanenrejo sub-watershed was forest (close to 50%) followed by plantations (>35%), and then Tegal, rice fields, open land and settlements covered an area <20%. In the same period, the spatial analysis results show that certain types of land use decreased there, and conversely, several types of land use increased. Forest land use was reduced by 1,550 ha, followed by plantations and paddy fields (by 194 ha and 50 ha respectively); on the other hand, there was an increase in land use in Tegal, open land and settlements (1,521 ha, 170 ha, and 102 ha respectively).

The forest ecosystem contains various types of plants with a multilevel canopy that functions as the main rainwater catchment area in the watershed area. Thus, reduced forest land cover can impact the amount of rainwater that can be stored in the soil. Table 2 shows a decrease in average discharge during the rainy and dry seasons, even though the rainfall in the same period showed an upward trend of 184 mm/year and 64 mm/year in the rainy and dry seasons (Table 3).

The volume of stored water shows the amount of rainwater that can be stored in the soil. Table 3 shows that in the wet season, water is stored at a rate of 576 mm/year. In dry season 1, the volume of stored water showed a negative trend. This trend has implications for the lack of water availability in dry season 1. On the other hand, Table 4 shows a decreasing trend in rainfall (natural water input) of −1.082 in the 2011–2019 period compared with the 1993–2010 period. Therefore, in several micro watersheds in the Sanenrejo sub-watershed, farmers have used a water pump to meet plant water needs.

Figure 4 shows that land use changed in the study area. Forest cover has long been a significant factor for water resources. Forests act as ‘sponges’ that retain water through soil infiltration. Deforestation reduces soil infiltration, increases overland water flow and reduces groundwater recharge and baseflow (Peña-Arancibia et al. 2019). The increasing overland water flow is reflected by the increasing streamflow during a period with heavy rainfall. The declining vegetation cover is also responsible for the declining streamflow and shares an equal role with climate change in transforming a catchment area into a water-limited area (Cecílio et al. 2019; Liu et al. 2019). Thus, the data indicate that the studied SSI faced physical water scarcity due to deforestation in the mountain area, which reduced soil infiltration and increased overland water flow.

Fig. 4

WTP for irrigation varies by location and season and is mostly attributed to water scarcity and difficulties in providing water to farmer plots. As water scarcity increased during the second and third seasons, the WTP increased. In the first season, when water was sufficiently available, the average WTP was the lowest. The average WTP for the upper, middle and lower areas were US$ 19.62/ha, US$ 27.44/ha and US$ 23.57/ha respectively. In contrast, the average WTP for the second season was US$ 31.03/ha, US$ 158.68/ha and US$ 108.16/ha for the upper, middle and lower areas respectively. This value is significantly higher than the irrigation water fees charged in an area with technical irrigation, as Syaukat et al. (2014) identified that the water fees paid by a farmer in Bogor and Kudus were US$ 5.29/ha and US$ 10.59/ha annually

Bogor and Kudus are two districts located in the Province of West and Central Java.

These results demonstrate that water is as an ‘economic good’ during water scarcity. Although this study focused on irrigation from spring water, the same condition was identified for almost all kinds of water sources, such as groundwater (Knapp et al. 2018), surface water (Chandrasekaran et al. 2009) and even for the demand of recycled water for irrigation (Bakopoulou et al. 2010). Moreover, the cost of irrigation for farmers accounts for 20% of their revenue

Total WTP for the sampled farmers was US$ 5,897.65 and total revenue was US$ 31,692.47.

In total, the average irrigation cost for an entire plot (144 ha) was US$ 31,080.95. Furthermore, there was a significant WTP increase of 578% and 458% in the middle and lower areas, while WTP in the upper area was only 192%. This number further increases in the third season with 260%, 799% and 684% rises for the upper, middle and lower areas respectively. This figure indicates that the economic value of water for the farmer is high. Thus, they are likely to invest in building II, which lessens the irrigation cost and increases its efficiency. Although water scarcity grows in land altitudes, the middle area had the highest WTP in all seasons, showing that the value of water is not affected by water availability alone. Hence, it is crucial to investigate the determinants of WTP further.

The determinants of farmers’ WTP were identified by multiple linear regression method. The F-test for the overall fit of the model is shown in Table 5. It tests the null hypothesis that all coefficients in the model are 0. Since the F-test p-value is p < 0.05; p = 0.00, the null hypothesis that all variables coefficients are 0 is rejected. Thus, it can be concluded that the model is better at estimating farmers’ WTP for irrigation water. The explained variance of the dependent variable can be measured with the R² value. The R² value of 0.83 indicates that the model can explain 83% of the WTP variation. This percentage is satisfactory since the model does not violate the normality, multicollinearity, homoscedasticity and linearity assumptions.

The t-test in Table 5 measures the significance of each independent variable. If the p-value is <0.05, then the variable has a significant contribution to the model. The t-test shows that farm size, income, distance to a small dam and management in the dry season are significant. Each of these variables has a p-value <0.05. The effects of each independent variable are reflected by the value of the coefficients. A negative coefficient indicates that the corresponding variable reduces farmers’ WTP. Conversely, a positive coefficient indicates that the corresponding variable increases farmers’ WTP. The value of coefficients reflects a multitude of effects. The estimation results show that social variables—age and education—do not significantly affect farmers’ WTP. The farmer's age and education had a t-test value of −3.012 and −0.027 with p-value >0.05: p = 0.98 and p = 0.11 respectively. This result shows that social characteristics do not affect the value of farmers’ WTP. Identical results were found by Jaghdani and Brümmer (2016), who identified WTP determinants for Iran's groundwater. Their study found that social characteristics do not significantly affect farmers’ WTP.

Both economic and technical variables affect WTP significantly. Farm size has a negative effect, while income has a positive one. The results show that larger farms tend to be less willing to pay higher irrigation prices. Conversely, farmers with higher incomes tend to be more willing to pay higher irrigation price. Larger farms require a larger volume of water, thus increasing the overall cost of farming. Hence, farmers with larger farms tend to be unwilling to pay higher irrigation prices. A similar result was found by Jaghdani and Brümmer (2016). Larger land endowment reduces farmers’ WTP although farmers with higher incomes can pay higher irrigation prices.

Both economic and technical variables had p-value <0.05. Distance to the small dam had a negative coefficient, while management in the dry season had a positive coefficient. This means that farmers with plots located far away from a water source (a small dam) have a higher WTP since it is increasingly costly to provide water to their farm. This factor explains why the WTP in the middle area is higher than in other areas, especially in the dry season. Out of the eleven water pumps in the studied region, only two are located in the middle area. Consequently, the cost to provide water to the middle area is higher. The last variable—management in the dry season—had a positive coefficient and showed that farmers who use water pumps are willing to pay higher irrigation prices.