3.1. Land Use Changes
Table 2 lists the land use area and proportion of the total area in each land cover type in 1992, 2003 and 2014. In 1992, agriculture and grassland comprised the largest proportion of the total area, approximately 65.31% of the total area, while the water and other land use types covered less than 5% of the study area by 1992.
Figure 3a shows that agriculture and grassland were located mainly in the areas surrounding settlements, except in the southeastern and southern parts of Denpasar. Interestingly, during 2003, the proportion of agriculture and grassland began to equalize with that of settlements, with settlement areas and agriculture and grass areas accounting for approximately 43.49% and 42.03% of the total area, respectively. During this period, the settlement land use type began to spread in all directions from the city centre and clustered in the southeast region. It is noteworthy that compared to the spatial pattern in 1992, agriculture and grassland became more fragmented and scattered between settlements in 2003 (
Figure 3b), and a large area of forest began to establish in the southern part of Denpasar. The forest in the southern part of Denpasar is dominated by mangroves, which began to be replanted in the pond area in 1995, although the project itself began in 1992 [
70]. Settlement became the most dominant land cover type in Denpasar city during 2014, accounting for 58.57% of the study area by 2014, followed by agriculture and grassland, accounting for 28.21% and 9.68% of the study area, respectively. The two other types of land use covered less than 5% of the total area. The area of settlement expanded broadly and resulted in clear visibility of the agriculture and grass land use type, especially in the southern part of the study area, which is adjacent to the centre of tourism activity (
Figure 3c). Tourism centres are not only located in the southeastern part of Denpasar but also in another district neighbouring the southwestern side of Denpasar, namely, the Kuta tourism centre.
Land use has changed in the tourism area of Denpasar from 1992 to 2014. Based on the area in 1992, settlement increased by approximately 204.75% in 2014, and the forest and agriculture and grass land cover types decreased to 12.68% and 55.29% ha in 2014, respectively (
Table 2 and
Table 3). The fastest change in land use in Denpasar was found from 1992 to 2003, especially for settlement and agriculture and grassland. Based on the LUEI value, the settlement area increased by approximately 0.69 ha day
−1, and the agriculture and grass land cover type lost an area of approximately 0.60 ha day
−1. Meanwhile, forests decreased by approximately 0.04 ha day
−1 during the first nine years. During the period 2003 to 2014, settlement still showed an expanding pattern with an expansion rate of 0.43 ha day
−1, and agriculture and grass declined with an LUEI rate of 0.39 ha day
−1. In general, the rate of change in the settlement areas in Denpasar over the past two decades increased by approximately 0.67 ha day
−1, and the forest and the agriculture and grass areas decreased by approximately 0.02 ha day
−1 and 0.60 ha day
−1, respectively.
The characteristics of the settlement land use type in Denpasar became increasingly complex from 1992 to 2014, with the geographical variations characterized mainly by the transformation from greenspaces into buildings for residence and government activity as well as into multi-functional buildings that provide general tourism services, including accommodation, shopping, restaurants, and entertainment. The changes in land use in the Denpasar region not only show expanding settlement areas but also expanding forestland due to the conversion of water and other land use types in the southern part of Denpasar and reclamation areas in Serangan Island. The clear land use change in this tourism city is consistent with the studies of Kytzia et al. [
71], Xi et al. [
72], and Mao et al. [
73]. However, the process of greenspaces changing into settlement areas in Denpasar is very fast compared to that in other cities in Indonesia and the world. In Jakarta, the capital city of Indonesia, the settlement area has grown to approximately 0.02 ha day
−1 over the past four decades (1970 to 2008) [
13], and in Beijing, China, the urban land cover from 1986 to 2001 expanded at a rate of 0.06 ha day
−1 [
74].
Denpasar has been the capital of Bali Province since 1958. Therefore, all activities of the provincial government and activities related to education and business are centred in Denpasar. Although Denpasar City is not a major tourist destination in Bali, being the centre of government and the presence of tourism destinations in the Sanur region have led to growth in tourism, which can have a direct impact on land use change. As shown in
Figure 3a,b, most of the forest in southeastern Denpasar, which is the Sanur region, has was converted to settlement from 1992 to 2003, and most of the change was from forest to multi-functional buildings for tourism activities. Bali has been a tourism destination since before Indonesia became independent [
75]. Tourism in Bali has continued to grow, especially after the construction of the Bali Beach Hotel (now known as the Inna Grand Bali Beach Hotel), the first five-star accommodation for tourism located in the Sanur region of Denpasar. This hotel has been in operation since 1966, has a height of 10 floors, which was the tallest building at the time of its construction and is still the tallest in Bali today. For example, over 11,278 foreign tourists per year visited the island in 1969, and by 2014, the number had grown to 3,766,638. From 1990 to 2000, foreign tourist visits to Bali grew an average of 17.08% per year, and from 2000 to 2014, the average increase was only 11.11% per year. The rapid growth of foreign tourists visiting the island before the new millennium may have led to rapid changes in land use in Denpasar from 1992 to 2003 by increasing the number of multi-functional buildings for tourism services and the incomes of employees in the tourism industry. In addition, the change in land use resulting from the indirect impact of the development of the tourism industry on population growth through migration in the southern part of Bali, including Denpasar, has increased the need for residential buildings. Denpasar City is the most populous area in Bali Province, and in 1992, the population density reached 2704 people per km
2 and increased to 6622 people per km
2 in 2014. Bali has experienced the most rapid economic development of all the islands of Indonesia [
76]. One of the main reasons for this growth is the many income potential created directly by tourism and indirectly by industries that produce goods for the tourism industry. Consequently, tourism activities can ensure income increases and improvements to the standard of living. In total, 51 percent of people’s income and 38 percent of Bali’s employment opportunities are directly linked with expenditures by tourists and tourism investment [
77]. On average, the tourism sector has accounted for 30% of the gross domestic product (GDP) of Bali Province over the past two decades, where economic growth is largely contributed by the southern regions as centres of tourism activity, including Denpasar city [
14]. The growth in economic income, living standards, and population are common causes of land use change in urban areas [
78,
79]. The relationship between population growth and regional GDP and land use change was also described by Wu et al. [
80] in the Yangtze River Basin, China. Moreover, an increase in the number of tourists can affect the land needs associated with their activities [
72], and similar results were presented by Allen et al. [
81] in South Carolina.
As previously described, the Inna Grand Bali Beach Hotel is the tallest building in Bali today with a height of 10 floors. In Bali, the Decree of the Governor of Bali Province No. 13/Perbang.1614/II/a/1971, which was made in 1971, sets the maximum building height to be 15 metres, typically only four floors. This regulation stems from local wisdom of the Hindu religion and Balinese culture on the urban and tourism development processes. This rule ensures that the construction of buildings such as residences, government offices, businesses, multi-functional buildings for tourism, and others do not spread skyward. Consequently, sideways expansion is a logical choice for obtaining large buildings with the largest volumes but directly and indirectly sacrifices greenspaces for settlement areas. This regulation may be one of the causes of the rapid change in land use in Denpasar, although government regulation can control and block the acceleration of land use change [
82,
83,
84]. As mentioned earlier, Jakarta and Beijing did not experience any rapid land use change compared with Denpasar. Although both cities are capital cities, they allow the existence of taller buildings. Therefore, further studies related to the impacts of this regulation need to be undertaken to identify their linkage with declining greenspaces and the expansion of settlement.
3.2. Annual Changes in GPP
The VPM was used to estimate GPP in 1995, 2003 and 2014 in the tourism city of Denpasar.
Table 4 reports the annual total GPP estimates for each of the land cover types and their percentage of the total. Meanwhile, the spatial distributions of annual GPP in Denpasar as analysed from the Landsat satellite data by the VPM method for 1995, 2003 and 2014 are presented in
Figure 4. Because the current study only studied changes in GPP, the study only shows GPP values for land use types related to greenspaces and settlement and does not include GPP values for non-greenspaces and settlement land types. The annual GPP for the entire region in the years 1995, 2003 and 2014 was estimated to be 61,609.15, 55,621.81 and 53,815.19 tC year
−1, respectively. In 1995, the estimated GPP in agriculture and grassland accounted for the largest portion, approximately 75.37% of the total estimated GPP, followed by forest with a proportion of 13.10%, and settlement accounted for 11.53%. It can be clearly seen in
Figure 4a that the GPP was estimated to be higher than 550 gC m
−2 year
−1 in the agriculture and grass land use type. In 2003, the allocation of agriculture and grassland still accounted for the largest area, although it is lower than that in 1995, 55.33% of the total area, while the forest and settlement land use types exhibited increasing allocations of estimated GPP with proportions of 16.61% and 28.06% of the study area by the year 2003, respectively. There is an interesting spatial pattern of estimated GPP in 2003. Inside the area around the settlement, there is still a fairly high GPP value (>550 gC m
−2 year
−1), and the south part of Denpasar has been dominated by high GPP values as a result of the replanting of mangroves. However, the high estimated GPP regions in the southeastern part of Denpasar in 1995 became more fragmented and scattered with low values of GPP in 2003 (
Figure 4b), while the same location in 2003 was dominated by multi-functional buildings that serve tourism activities. Interestingly, the GPP estimation for settled areas has overtaken that of agriculture and grass areas, with a value of 41.39% of the total GPP estimation in 2014. The GPP estimation from forest increased by 19.08%, while that of agriculture and grass areas also declined, although the proportion was still quite large with a share of 39.53% of the total area in 2014. It is noteworthy that compared to their spatial pattern in 2003, the medium to high value GPP values were more fragmented and scattered around the regions with low GPP estimates in 2014. High estimated GPP was clearly observed in the southern part of Denpasar and in small areas of the agriculture and grass land use type (
Figure 4c).
Figure 5 shows the mean estimates of annual GPP as analysed from Landsat satellite data for each land use type for 1995, 2003 and 2014. The mean GPP estimation for settlements decreased slightly (339.23 to 304.99 gC m
−2 year
−1) during the 20-year period and increased marginally for the agriculture and grass land use type (593.53 to 605.65 gC m
−2 year
−1). Meanwhile, the average GPP estimation in forests over the past two decades increased dramatically compared to that for the other two types of land use from 1995 to 2014 (544.97 to 846.40 gC m
−2 year
−1). However, in general, the mean GPP estimated for the settlement and in agriculture and grassland land cover types was almost similar, and that for the forest increased over the past two decades. This study found that the estimated GPP for the settlement region is half of that for agriculture and grassland and that a change of 1 ha of agriculture and grassland into settlement will cause Denpasar city to lose 3 tC ha
−1 year
−1 of carbon uptake carbon by green vegetation (GPP).
The region-wide net decrease in GPP was approximately 7793.96 tC year
−1 during the past two decades, a decrease of 12.65% from the GPP value in 1995 (
Table 5). The estimated total GPP decreased during the 20-year period with an average decline of 0.63% per year. However, the fastest decline occurred in the first nine years compared to 2003 to 2014 (
Table 6), which is similar to the decrease in the agriculture and grassland use type. Interestingly, based on land use data, GPP estimates increased for settlements and forests, although the forest area decreased, while a decline in the value of estimated GPP occurs only in the agriculture and grassland use type. The GPP estimate for settlements increased at a rate of 13.31% per year, while during 2003 to 2014, the increase in the GPP estimate was less than that for the first nine years, and over the past 20 years, the increase has been large, reaching 10.68% per year. The estimated GPP in forests over the past two decades has continued to grow, even though the forest area has declined, with the fastest growth occurring from 1995 to 2003. Decreasing agricultural and grass land use has also led to a decline in the value of GPP estimates, and the decline in the past 20 years reached more than 50%. The fastest change, a value of 3.75% per year, was found from 1995 to 2003, and from 2003 to 2014, the value of estimated GPP loss was approximately 2.57% per year.
The current study found that the total value of estimated GPP for the settlement land use type increased dramatically. From the lowest proportion in 1995, this land use type achieved the highest ability to uptake total carbon through vegetation of all land use types in 2014. The average value of estimated GPP is similar in the three observation periods, although the average value is lower than the result obtained by Zhao et al. [
19] (390.24 gC m
−2 year
−1), which may be due to the different methods used in GPP calculation. The increase in the GPP estimation value is very high but is still less than the increase in the settlement area, where in general, there is a linear relationship between the increase in area and the total GPP in the settlement area (see
Table 3 and
Table 5). The linear relationship between changes in settlement area and primary production is consistent with the study of Lu et al. [
5]. A linear relationship was also found in the agriculture and grass land use type, with decreases in area of approximately 55.29%, followed by a loss of ability to uptake total carbon of 54.19%. Although only approximately half of the agriculture and grass land use type has been converted, the area reached up to 4350.51 ha. The settlement areas increased to 4906.44 ha during the study period, which indicates that agriculture and grass have been converted to settlement areas. Consequently, the ability of cities to uptake carbon from the atmosphere is reduced, as displayed in
Table 5. Settlement areas can uptake carbon due to the presence of vegetation. As-syakur et al. [
85] have shown that not all settlement areas in Denpasar are covered by structures, but have various built-up land uses that include vegetated and non-vegetated areas such as backyards and front yards. However, this city is unique in that the majority of the population is Hindu, and the city is a “flat city” due to the absence of tall buildings. These characteristics are not found in other regions in Indonesia or in other regions in Southeast Asia. The city of Denpasar contains a “holy area” for Hindu religious activities at each residence that is characterized by the presence of vegetation with varying levels of cover and heights. However, the increase in settlement area has not impacted the average GPP estimates, unlike the results obtained by Zhao et al. [
19], which show an increase in the mean annual GPP at the time of an increase in the built-up area in the eastern United States. Given that the development of settlement areas in the city is difficult to control due to high demand for tourism activities, an increased settlement area accompanied by good management of “holy areas”, backyards, and front yards should be able to increase the ability of settlement areas to uptake atmosphere carbon through vegetation.
Interestingly, although the forest areas are degraded, forest restoration has been helpful in preventing a drastic decline in GPP in Denpasar city over the past two decades. A forest area of approximately 175.41 ha was lost during the study period, but the annual estimated GPP increased by approximately 27.23% from 1995 to 2014. This increase is due to the restoration of mangrove forests in the southern part of Denpasar from 1995 to the present. Reforestation is conducted by replanting mangroves in water (pond) land use areas and maintaining these areas by making them protection areas. The mangrove forest in southern Denpasar is part of a protected area established by the Indonesian government, the Ngurah Rai Grand Forest Park, through decision No. 544/Kpts-II/1993 in 1993.
The current study indicated the importance of forest (mangrove) restoration to sustain the carbon cycle in urban areas that have experienced rapid land use change due to increased tourism activity and urbanization. Ecological restoration is important to increasing the total carbon uptake by vegetation. Successful ecological restoration associated with primary production has been reported previously by Yang et al. [
86], but this result was an increase in the amount of carbon sequestration due to an increased terrestrial vegetation area, not under the decline of forest area. However, Briber et al. [
87] found an increase in tree productivity during forest conversion to urban areas. Forest productivity is a function of many factors such as species, age, resource availability, growing season length, and competition [
88,
89]. Mangroves are among the most carbon-rich forests in the tropics [
90]. Duarte et al. [
91] also suggest that mangrove carbon production is more rapid than other estuarine and marine primary producers. On the other hand, Landsat data are some of the primary sources for identifying areas of forests in space and time with good accuracy [
92], including in urban areas [
93] and mangroves [
94].
However, this study still has some uncertainties. The land use data used in this study do not match the year, especially for GPP analysis in 1995, which may cause a mismatch in extracting the estimated GPP. This mismatch will lead to an underestimation of GPP values in settlements and an overestimation in agriculture and grass areas because most of the settlement areas come from agriculture and grass areas. From 1992 to 1995, Denpasar had an average population growth rate of 2.85% per year, and foreign tourist visits to Bali grew an average of 13.47% per year in the same period. One consequence is the increasing number of multi-functional buildings for tourism services and the number of residences needed to accommodate the population growth. However, the four-year difference may not have a significant impact on a study that had a long span period. In addition, the use of 30-m Landsat data can cause errors in detecting vegetation in heterogeneous settlement areas. As shown by As-syakur et al. [
46], different spatial resolutions in the satellite data used for analyses of GPP in urban areas will lead to different averages for settlement areas, and more detailed spatial resolutions will give larger average values. Increased pixel sizes (or decreased spatial resolution) result in the loss of image detail [
95]. Satellite data with high spatial resolutions may be able to reduce the difficulties with remotely sensed data from coarse-resolution satellites [
96]. Finally, this is the first study that used the VPM method with medium-spatial-resolution satellite imagery to calculate GPP in urban areas; therefore, the results are questionable. Previously, Ciu et al. [
97] tried to compare GPP-based VPM estimates (using MODIS data) and MODIS GPP data products (MOD17A2) with solar-induced chlorophyll fluorescence (SIF) over the most populous megacity area with better results compared to those produced with the MOD17A2 product. On the other hand, the total annual GPP estimated using the VPM in the current study was not very different from the GPP estimate produced by other LUE models or from the MODIS GPP product, as presented by As-syakur et al. [
46] in the same location. This study can be used as an initial source of information related to land use changes and their impact on terrestrial carbon uptake by vegetation. However, urban areas are highly heterogeneous landscapes that have different or even opposing effects on overall urban vegetation productivity [
98]. Urban areas also have complex climate systems that are affected by complex socio-ecological systems [
99] and the urban–rural proportion as related to urban heat islands [
100], which directly and indirectly affect the calculation of light-use efficiency. Therefore, a validation of the estimated GPP results of the VPM method with eddy flux towers needs to be performed to advance our quantitative understanding of the capability of this method to contribute to analyses of vegetation carbon uptake in urban areas.