2.1 Study area
This study was conducted in Songkhon District, Lao PDR (Fig. 1), located near Lahanam, which is situated along the Bang Hiang River, a branch of the Mekong River. The climate is considered tropical monsoon, where the wet season lasts from April to October. The average seasonal temperature is about 27 °C with an average rainfall of roughly 1,390 mm during the wet season. The dry season starts from November and ends in March with an average seasonal temperature of 23°C and an average precipitation of roughly 73 mm (Savannakhet Provincial Government, unpublished). The village has a total population of 1,091 registered residents, with 628 between 18 and 65 years old as of March 1, 2010 (Village Chief notes). In addition, about 30% of registered adults in the village migrated to Thailand or Vientiane (capital of Lao PDR) for work.
A typical family in Lahanam earns a living by rice cultivating, livestock grazing, fishing, and cloth weaving. Generally, livestock grazing is the work of male adults and weaving is the work of female adults (Author’s field note in 2010, unpublished). However, both female and male adults engage in fishing year-round (Author’s field note in 2010, unpublished). Consistent with other regions of Lao PDR, rice cultivation is the major subsistence of local residents (Author’s field note in 2010, unpublished).
Our study area has roughly four seasons: wet farming, wet off-farming, dry farming, and dry off-farming. Two types of rice cultivation are carried out in the wet and dry seasons. The wet season rice cultivation, which is rain-fed, starts from May and is harvested in October [39]. The dry season cultivation is carried out in sections of the rice paddy with irrigation from December to April. The drastic differences in precipitation between wet and dry seasons have a direct impact on individual activities. To that end, our survey is predicated on the four seasons based around Lahanam’s farming activities.
2.2 Data collection
To capture the seasonal activity changes of local residents, multiple waves of surveys were carried out in Lahanam from March 2010 to March 2011. In each wave, two devices were used to collect data for each participant for seven consecutive days. The first, a portable GPS device (M241, Holux Technology, Inc.), recorded activity and farmland location of participants. The second, an accelerometer instrument (Lifecorder EX, Suzuken Ltd.), measured the participant's physical activity level and steps. Recording was made every twelve seconds for the GPS and two minutes for the accelerometer. Concurrently, a self-reported log by each participant detailed the time spent on each activity from 6 am to 6 pm. The body weight and height of each participant was measured by the researchers. GPS data were validated by the accelerometer and, if no record existed in the accelerometer, we excluded the data.
The first wave of data collection was conducted in March 2010 to test the GPS devices battery and whether the device shell was waterproof in the study area. Wave one consisted of 20 participants, 10 males, and 10 females, with each participant aged between 18 and 65 years. From Wave 2 to Wave 5, about 30 participants aged between 18 and 65 were recruited, respectively. Waves 2 to 5 covered four different farming seasons, namely wet farming season (Wave 2: June 2010), wet off-farming season (Wave 3: September 2010), dry farming season (Wave 4: December 2010) and dry off-farming season (Wave 5: March 2011). Twenty thousand Lao kips (USD $2.50) was given to each participant who completed one wave, with a total of 100,000 Lao kips (USD $12.5) given to participants who completed all five survey waves. Only Wave 2 to Wave 5 were included in the further analyses.
Neighborhood built environment attributes were derived from google maps and QuickBird satellite image which was taken on January 9, 2008. Euclidean distances from participant’s home to different types of land uses such as paddy, the health center, the primary school, the middle school, the local market, the temple, and the mini shop in the village were measured in order to identify different destinations where residents frequently visit.
2.3 Creating activity space
The most common method to operationalize activity space is the standard deviation ellipse (SDE) that measures the directional distribution of a series of GPS points or the “densest” areas where most of the individual mobility occurs [40]. The SDE is usually centered at an individual’s home [41] and extended two standard deviations to cover 95% of the observed activity locations [42]. This study uses an SDE to represent the participant’s activity space [42]. Because the SDE approach helps to assess the direction and general shape of a person’s travel area without introducing potential errors using geographically distant points or a road network that is not reliable.
For each participant in each wave, activity space was delineated as an ellipse centered at the home and extended to two standard deviations of the observed activity locations that were recorded by the GPS device worn by each participant (see Fig. 2 for a participant’s one-day activity). The area of each activity space was calculated using a directional distribution tool in ArcGIS 10.5. The tool summarized spatial characteristics and created an SDE polygon as the output activity space. A 1-, 2-, and 3 - SDE polygons were computed to cover 68%, 95%, and 99% of the input features, respectively. Considering that a 1-SDE covers 68% of activity points, several participants’ activity space may not be constructed accurately using the 1-SDE approach. Conversely, using a 3-SDE may include outliers that distort the shape of the ellipse and introduce concerns similar to the minimum convex polygon approach which could result in measuring the extreme extent of travel and capture large geographic areas that are not visited by an individual. Therefore, our study used a 2-SDE polygon to delineate activity space in line with previous research [42].
2.4 Statistical Analysis
Bivariate analysis of variables derived from the daytime activity log, and distance measures of land use were carried out with areas of activity space and counts of daily steps. Furthermore, ANOVA tests controlling for different waves were conducted for the two dependent variables by gender, and the self-reported time for different activities by gender by wave, respectively. Linear mixed models with the fixed effects as the observations from different waves and the random effects as individual participants were developed to identify factors associated with areas of activity space and counts of daily steps, respectively.