We selected two villages with contrasting agroecological characteristics in the upper Citarum basin: Wangisagara and Sukapura. Both villages are on the Citarum River that runs approximately 350 km from its source on Mount Wayang north to the Java Sea (north of Jakarta). The river is a major water resource in the region and three large reservoirs have been constructed along it.

Wangisagara village is located on gentle slopes that range in altitude from 800 to 850 m above sea level (asl), whereas Sukapura village is on steeper land ranging in altitude between 950 and 1250 m asl. The difference in altitude results in a considerable difference in climate between the two villages. The mean annual rainfall in Wangisagara ranged from 1500 to 2500 mm with minimum daily temperature in December and January between 20˚C - 25˚C and maximum between 25˚C - 30˚C. In Sukapura, the mean annual rainfall was considerably greater at 2000 to 3000 mm, and the daily minimum temperature in December and January was lower at 15˚C - 20˚C with the maximum ranging from 22˚C to 28˚C (data obtained from the Regional Office of Meteorology and Geophysics). Other general characteristics of the study villages are summarized in Table 1.

Prior to the selection of respondents for the interview survey, a full census of bamboo-tree garden owners in both villages undertaken at the rukun tetangga (RT) level was analyzed, since there were no formal records on bamboo-tree garden ownership. An RT headed by an elected chief is the smallest administrative unit within village administration. Each RT usually consists of 30 to 40 households, depending on the total number of households and the area of the village. The census revealed that only a small percentage of the households owned bamboo-tree gardens. Only 68 households owned bamboo-tree gardens in Wangisagara (3.30%) and 52 households in Sukapura (2.46%), the bamboo-tree owners in the two villages were all men.

The required number of respondents was determined by the following formula [18] :

where n = number of respondents, N = number of households in each study village, Z = the value of the normal distribution (1.96) for a reliability level of 0.95, p = the highest possible proportion (0.5), and d = the sampling error (0.1).

Using the above formula, 36 households in Wangisagara and 30 households in Sukapura were randomly selected. In addition to the present owners, interviews were also carried out with former owners if they were identified. Unfortunately, there was no formal record of ownership change, so only nine previous owners were interviewed in Sukapura and five in Wangisagara.

The data were collected based on structured interviews with randomly selected bamboo-tree garden owners in the two villages. To determine the owners’ perceptions of their gardens, we asked them 56 questions using the semantic differential technique based on arguments as follows:

・ Similar socio-cultural background of owners in the two villages;

・ Same (local) knowledge with regard to bamboo-tree garden cultivation practices;

・ Individual preference with regard to decision making process in bamboo-tree garden management.

For each question, the answer was scored from 1 to 7, ranging from very negative (1) to very positive perception (7). The score for every answer inevitably contained an element of subjectivity on the part of interviewers. Therefore, a probing questioning technique was used during the interviews. Where doubtful scoring occurred, the interviewer had to return to the respondent to clarify ambiguous answers. A single interview could last more than 2.5 hours. Information on agricultural land ownership and social parameters such as education level and main occupation were also collected.

The questionnaire was divided into several categories related to the owner’s perceptions of their bamboo-tree garden or, in some questions, bamboo tree-gardens in the village (Table 2). These categories were (A) garden ownership; (B) the socioeconomic, cultural, and ecological functions of the garden; (C) management of the garden; (D) recognition of any decline in bamboo-tree gardens in the village and causes for that decline; and (E) the outlook for preservation of the gardens. Some of these categories consisted of several subcategories (Table 2). For example, the first category, about the general attitude to owning a garden, consisted of three subcategories pertaining to socioeconomic status, cultural meaning, and economic benefit. The 56 questions were classified into each subcategory, and 33 major questions were chosen to reflect the major aspects of each category or subcategory. The codes used for the 33 major questions and a brief explanation of each major question are shown in Table 2. The remaining 23 questions were to provide supporting information related to the major questions and/or additional perceptions of the respondents.

Semi-structured interviews were also carried out with former owners and other relevant informants, such as the village leader, local farmers, and fuelwood gatherers, to cross check the answers given to particular questions by bamboo-tree garden owners. Secondary data were collected from various sources including previous

studies conducted at the present study sites. Data on the general characteristics of the villages were obtained from the village office.

We performed principle components analysis (PCA) of the perception scores. Eight respondents from Wangisagara did not answer all of the questions, so the answers of 58 respondents (28 in Wangisagara and 30 in Sukapura) were available for analysis. The respondents’ scores on each PCA axis were correlated with external factors such as the area of agricultural land owned and education level obtained to determine causal relationships between perceptual rankings and external factors. Kendalltau rank correlations, a nonparametric measure of the degree of correspondence between rankings, were also calculated.

The trends in respondents’ perceptions obtained from the PCA differed between Wangisagara and Sukapura. We tested differences in perceptions of the bamboo-tree garden owners between the two villages using the Mann-Whitney U-test on the responses to the 33 major questions as well as to some of the supporting questions. All statistical analyses were performed using SPSS 10.0 (SPSS Inc., Chicago, Illinois, USA).

There was no significant difference in the size of bamboo-tree gardens between Wangisagara and Sukapura (Ta- ble 3). The number of owners and size of upland fields of vegetable cash-crop fields and kebun palawija, which is a mixed cropping system of inter-planting a number of annual crops in the same field at the same time, and usually consisted of maize, cassava, sweet potato, and peanut, were higher in Sukapura (P < 0.01), whereas Wangisagara has more owners and greater area of paddy fields (P < 0.01).

The principle components analysis (PCA) on the data set with 33 variables and 58 samples yielded five principle components or axes that had eigenvalues > 2. The eigenvalues for the first to the fifth axis, in order, were 5.25, 4.31, 2.90, 2.41, and 2.23 (Table 4). Together, these five axes accounted for more than half the total variance (51.8%). Factor loadings for each variable or question in the five axes calculated by PCA are shown in Table 4.

Correlations between the five principle component scores and agricultural landownership are shown in Table 5. The size of bamboo-tree gardens was significantly positively correlated with the first axis and negatively with the third axis. The size of paddy fields was significantly positively correlated with the first axis and negatively

^**P < 0.01; N.S. indicates no significant difference; - indicates no test performed.

^aSee Table 2 for the abbreviations of variables.

^*P < 0.05; ^**P < 0.01.

with the second axis. The area of paddy fields was significantly positively correlated with the area of bamboo-tree gardens and negatively correlated with the area of upland fields. The latter correlation is understandable because respondents who owned paddy fields were predominantly in Wangisagara where there were relatively few upland fields (Table 3).

Figure 1(a) shows the factor loadings for the 33 questions and the principle component scores of the 58 respondents for the first and second components although the cumulative explanation of variance was about 30% for these two components. The scatter patterns of the respondents were different for the two villages; respondents from Wangisagara had larger scores on the first axis and smaller scores on the second axis than those of Sukapura (Figure 1(b)).

Although the difference in the scores of the two villages for the first axis was not significant (P = 0.054), the scores of Wangisagara tended to be higher, and there was a significant difference between the villages for the second axis (P < 0.01). There were no significant differences for the third and fourth axes (P = 0.99 and 0.84, respectively), but the fifth axis was correlated with the size of upland fields, where Sukapura had higher scores (P < 0.01).

Regarding the relationship between social parameters (such as main occupation and educational level) and principle component scores for each respondent, there was no significant relationship for the three educational levels: no education, completed primary school only, and completed higher school (U-test with Bonferroni corrections).

The PCA scores of respondents whose main occupation was farmer were significantly higher on the second PCA axis than those whose main occupation was non-farmer (P < 0.01). Since all seven non-farmers belonged to the village of Wangisagara, we compared the relationship between the scores of the second axis with occupations of respondents of that village only; there was no difference. This means that the perceptions of respondents were not influenced by either educational level or main occupation.

Along the first axis of correlation with the area of bamboo-tree gardens, perceptual categories such as economic and ecological functions of the gardens (B1 and B3), management knowledge to improve productivity of the gardens (C2), and willingness to preserve the gardens, especially with economic improvements (E0 and E1), had higher values (Figure 1(a)). The axis reflected generally positive perceptions on the existence of the gardens and indicated an economic dependency on the gardens.

Regarding the second axis, perceptual categories having higher values were those that referred to management frequency, especially of the more intensive activities such as pesticide input, weeding, and tree management (C1-3, C1-4, and C1-5). Other categories with higher scores included the amount of time spent on maintenance (C3-2), social functions (B2), and awareness of the decline in the garden through conversion to cash crops (D1-1). Scores with lower values were for socioeconomic status of the garden’s owner (A1), maintenance cost

(C3-1), need for marketing support to increase produce values (E1-2), and need for governmental support programs (E2) (Figure 1(a)). The second axis would therefore indicate intensification of the management of gardens. Respondents who had higher scores along the axis presumably managed their own gardens intensively, while those who had lower scores were likely to depend on sources of income other than their bamboo-tree garden.

For the third axis, which was significantly negatively correlated with the area of bamboo-tree garden, the questions with the five highest factor loadings were E2-1, B1-2, D3, E1-2, and D0, whereas the five lowest were A1-1, A1-2, A3, B1-3, and E1-1 (Table 4). Since there were positive perceptions on garden ownership in a negative direction, the axis explained the degree of negative perception for social and economic categories. For the fourth axis, the five highest categories were C3-1, C1-5, C1-3, C1-1, and E2-2, whereas the five lowest were B3-2, B3-3, B2-2, B2-1, and D0 (Table 4).

The association of perceived garden decline with positive perceptions on frequent management activities or with positive perceptions on social and ecological functions could indicate a perceptual contrast between whether the garden is seen as a part of nature or a part of the agricultural system. Finally for the fifth axis, which was significantly positively correlated with the area of upland fields, the five highest perceptual categories were D2, B3-1, D1-1, B3-4, and D3, whereas the five lowest were D0, A2-1, A2-2, B2-2, and B2-1 (Table 4).

Considering that the respondents of Sukapura had significantly higher scores on this axis than Wangisagara, it would appear to explain a perceptual difference. The farmers in Sukapura appear to recognize and accept a general garden decline for economic reasons, whereas those in Wangisagara appear to be alarmed by the decline because of the gardens’ longstanding role for subsistence and particular social functions.

Overall, bamboo-tree garden owners in the two study villages have positive perceptions of the socioeconomic, cultural, and ecological roles of the bamboo-tree garden as an element of the landscape. Differences in the intensity of perceptions are influenced by the frame of reference for the perception, such as factual knowledge and emotional experiences [8] . The more coherent the factual knowledge and emotional experiences, the more likely the influence on the perception of people. In general, the bamboo-tree garden owners in Wangisagara have more positive perceptions of the gardens than those of Sukapura. This was seen very clearly in perceptions of the economic function of bamboo-tree gardens; the owners in Sukapura exhibited less positive perceptions. This is because most of the owners in Sukapura village have other, more productive sources of income, for example, vegetable crops. With capital-intensive agricultural practices, they are able to obtain higher incomes than they can from their bamboo-tree garden. Therefore, they consider bamboo-tree gardens not as their primary source of incomeand consequently have a lower intensity of desire to maintain their garden. Bamboo-tree gardens are the least managed type of agricultural land in this landscape (see Wiersum) [15] .

The respondents in Sukapura perceived their bamboo-tree gardens as an asset that indicated socioeconomic status but not as a primary source of income. The perceived value of the bamboo-tree garden derives from the land itself or its potential to be converted to a more productive vegetable garden.

The difference in perception intensities between the two villages with regard to the social function of bamboo-tree gardens is presumably related to the local dynamics in livelihoods experienced by the people in the two villages. Although the people in Sukapura are likely to be more market oriented, since capital-intensive agricultural activity is more predominant there than in Wangisagara, they still allow other villagers to freely collect branches, medicinal plants, or raw vegetables for their own use. It is therefore not necessarily the case that engagement in a market-oriented livelihood overrides perceptions of the social function an owner’s bamboo-tree garden. Our finding was concomitant to other studies which were carried out by Okubo et al. [16] and Muhamad et al. [19] in other villages in West Java.

The respondents in Sukapura who practice capital-intensive agriculture still retain a sense of community sharing. This finding contrasts with the condition in capital-intensive agricultural practices in developed countries, e.g. Petrzelka [4] mentioned that among conventional farmers in Iowa USA, nature is viewed as a resource to be used and controlled, and farmers usually do not hold a sense of community but see farming primarily as a business. Even though the owners in Wangisagara rejected their garden’s social function in terms of providing fuelwood freely for others, but this does not mean that the bamboo-tree gardens in Wangisagara do not have a social function. They limited access for other villagers to collect fuelwood, but they still shared certain products like fruits from their bamboo-tree garden to neighbors as an important part of maintaining social relationships, meaning that the social role of the gardens was to supply excess fruit to villagers.

Respondents in the two study villages exhibited a more or less similar positive perception to the ecological functions of their bamboo-tree garden. Nevertheless, there was some inconsistency between what they perceived to be important and what was actually happening. Despite their perception about the ecological importance of the bamboo-tree garden for the village, such as preventing soil erosion, storing water, and providing habitat for wildlife, the conversion of this land-use type to less environmentally desirable systems was still occurring.

Although no quantitative data estimates have been made on the rate of conversion bamboo-tree gardens, our interviews and visual observations clearly indicate that such conversions continue unabated up to the present time. This suggests that positive perceptions do not always correlate with positive behavior (see also Jackson- Smith and Buttel) [20] , and sometimes the connection between farmer perception and practice is weak. This is similar to the results of a number of previous studies [4] [6] [7] [21] .

In many cases, farmers’ practices diverge from their perceptions because the way individual farmers act is not only influenced by their own perceptions but also by cultural and socioeconomic forces and the suitability of the bio-physical environment (see also Mekoya et al.; Muhamad et al.) [7] [19] . The development of a strong commitment leading to positive behavior of rural people towards the sustainability of bamboo-tree gardens is essential for their continued existence.

The perceptions of owners of the ecological functions of their bamboo-tree gardens were somewhat in concordance with a study of bird communities in fragmented bamboo-tree gardens reported by Erawan et al. [17] and Parikesit et al. [11] . The diversity of bird species found in bamboo-tree garden patches is presumably related to the structural pattern of vegetation layers and food abundance.

Therefore, ecological concern exhibited by bamboo-tree garden owners is an encouraging sign for conservation efforts. Unfortunately, no other studies concerning the ecological functions of bamboo-tree gardens in the upper Citarum basin have been carried out. It would be important to investigate the role of bamboo-tree gardens as a filter or sink for soil nutrients exported from neighboring cultivated land in the changing agricultural landscape of the Citarum basin. The role of this land-use type in relation to pest outbreaks and predator-prey interactions in vegetable crop gardens are other challenging topics that need to be addressed.

Compared with other types of agricultural land (rice fields, vegetable gardens, kebun palawija), bamboo-tree gardens receive the lowest levels of management inputs for production purposes [15] . The number of non- cultivated plants, which were found to grow naturally in bamboo-tree gardens including forest species such as Ficus spp., Sterculia oblongata, Arctostaphylos glauca, and Neonauclea lanceolata, was higher than that of domesticated species in both villages (data not shown). Therefore, it is not surprising that the economic productivity of these gardens is lower than that of other agricultural lands.

Management of bamboo-tree gardens without high external energy inputs as practiced by the owners in the two study villages is an example of traditional subsistence agriculture that was presumably predominant in the upper Citarum basin prior to the introduction of capital-intensive agricultural practices in the 1970s.

There has been an obvious decline in this element in the landscape over the last few decades due to rapid development of modern agricultural practices, particularly in the upper part of the basin (see Parikesit et al.) [11] . Owners displayed a wide awareness that the roots of bamboo clumps are very effective in binding the soil, so severe soil erosion under dense bamboo-tree gardens is unlikely to occur. However, some owners had converted their bamboo-tree gardens to cash crop (vegetables) gardens even though they were located on steep land. This is most conspicuous in Sukapura, where bamboo-tree gardens are usually found in small patches scattered within a matrix of other agroecosystem components such as rice fields and upland cash-crop gardens no larger than a few hectares.

The underlying cause of the decline in bamboo-tree gardens in the upper Citarum basin is clearly economics. Land conversion, population pressure, population growth, and lack of serious effort from the government to conserve bamboo-tree gardens appear to be some by products of these underlying factors. Economic forces are increasingly determining the structure and function of bamboo-tree gardens. When asked to explain certain practices, most respondents focused on pragmatic considerations and concerns about profitability, which were often the overriding factor in decisions regarding land use.

Thus, as pointed out by Corselius et al. [22] , many farmers’ decisions become more profit-oriented and focused on short-term production. Most of the owners who had already converted their bamboo-tree garden system to cash-crop gardens in Sukapura justified their decisions based on short-term economic viability. It should be noted, however, that farmers do not always, or necessarily ever, make decisions with the sole goal of profit [23] [24] . In concordance with the findings of Lynnne and Rola [25] , some farmers in our study argued that income alone is not a significant predictor of conservation behavior; a positive attitude to the environment is also required. However, several other studies suggest that profit motives override environmental motives, even when an awareness of environmental problems exists [19] [26] [27] .

Farmers responding to market demand and obtaining short-term benefits through such behavior has not come without environmental cost within the agricultural landscape of the upper Citarum basin. The decline in bamboo-tree gardens due to land conversion to capital-intensive agriculture has changed the landscape structure [11] .

The conversion of bamboo-tree gardens to vegetable gardens has undoubtedly caused a loss of habitat for various wild animals that use bamboo-tree gardens but not vegetable gardens. The conversion will ultimately cause a homogenization of the landscape structure [28] and loss of diversity of the agroecosystem of the upper Citarum basin. The use of pesticides in vegetable gardens may further contribute to environmental pollution [11] [29] [30] .

The perception of owners that continuing economic pressure threatens the future existence of bamboo-tree garden is entirely rational. The results of our study show that the economic pressures are clearly leading to conversion of bamboo-tree gardens to other land uses, whereas we are not aware of any concerted efforts to preserve this traditional agricultural system. Various levels of government emphasize on production rather than searching for a balance between production and conservation of traditional agroecosystems [28] . Bamboo-tree gardens may soon be confined to small patches occupying steep, non-arable parts of the Citarum basin [11] .

The finding that there is not always a direct correlation between human perception and action is in line with that of Mekoya et al. [7] . But gaining an understanding of human perceptions could still be an important step in revitalizing traditional bamboo-tree gardens because only through understanding of human perceptions can the interaction between the social system and the agroecosystem be shaped [3] .

The tradition of maintaining bamboo-tree gardens as an important family asset offers opportunities with respect to the objectives of conserving biodiversity and maintaining the socioeconomic role of this land-use type in the local production system. The owners certainly exhibit some desire to maintain the existing bamboo-tree garden patches in the study area, but this alone does not assure the future existence of bamboo-tree gardens in the upper Citarum basin. Support from government and other agencies through incentive-oriented programs that can act against pure market economic forces are needed. An example of such incentives might be to make payments to the owners of bamboo-tree gardens for the environmental services provided by maintaining this land use [31] [32] .

Ideally, efforts to conserve bamboo-tree gardens should be undertaken throughout the entire upper Citarum basin. However, the present study indicates that conservation efforts are likely to be much more challenging in the uppermost part of the basin in villages such as Sukapura. In this area, the local people, bamboo-tree garden owners in particular, have long been subject to a market-oriented livelihood (i.e. capital-intensive agricultural practices). Therefore, conservation efforts need to be directed toward improving the economic value of bamboo- tree garden products and at the same time improving the ecological function of this landuse type.