This study investigated the volume and patterns of GPS data recorded after school, defined as time outdoors, and the association with objectively measured physical activity in primary school-aged children. Minute-by-minute plots of time outdoors showed little difference between the genders, but marked differences between the summer and winter seasons. Matching accelerometer data with GPS data allowed activity to be segmented into "indoors" and "outdoors" with high resolution. These data showed that physical activity was 2-3 fold higher outdoors than when indoors and that the level of physical activity indoors was consistent through the year whilst physical activity outdoors was seasonally patterned. Linear regression showed that GPS-measured time outdoors after school was a significant predictor of physical activity in this period. These data are consistent with previous research which has shown that time outdoors measured by self/proxy report or by direct observation [7, 8] is a significant predictor of children's physical activity, and suggest that GPS recording may provide a novel method to measure time outdoors in free-living young people.
The associations described in the present study are similar to previous reports. The time spent outdoors in the spring/summer by children in this UK sample is similar to reported values for children measured during springtime in Sydney, Australia, with 40% of UK children recording <30 minutes outdoors each day compared with 37% of similarly aged Australian children . In contrast, though the figures are not directly comparable, children in the present study appear to spend approximately half the time outdoors in the summer as children from Melbourne with similar though lower times outdoors in the winter . The reasons for the discrepancies are unclear, but it is possible that time outdoors was differently reported in the self report/parental proxy measures used in Australia, and also possible that time outdoors is under estimated by the GPS due to the methodological issues discussed below. Further validation studies are required to resolve these discrepancies. In agreement with other studies, boys spent more time outdoors than girls, though differences were very small [7, 8], and the duration of time outdoors was significantly lower in the winter months than in the summer. Combining GPS and accelerometer data enabled us to show that whilst physical activity outdoors was lower in the winter than the summer, physical activity indoors remained at a constant (lower) level throughout the year.
The period after school is an important time for children to be physically active [23, 24] and the hours between 3:30 pm and 6 pm on weekdays have been described as the "critical window" for children's physical activity . Accelerometry has been used to demonstrate differences in the patterns of objectively measured physical activity in this period between groups of children based on travel behaviour  or weight status , but potential environmental determinants of these differences have yet to be determined. Interpreting data from this period is complex since children leave school at variable times depending on participation in after school activities and have a wide range of possible activity patterns which may be influenced by a range of physical and social factors. The consistent association between time spent outdoors and physical activity in previous studies of primary school-aged children, and the association between proxy-reported time outdoors and weight status, indicate the importance of this measure for understanding children's physical activity. This paper supports the concept that time outside is associated with increased physical activity and that a GPS monitor potentially provides an objective method of measuring time outside. Combining accelerometer and GPS data has potential to describe associations of time spent outdoors with physical activity in different populations and locations, and also to enable the measurement of the effect of interventions aimed at increasing time outdoors. GPS technology is rapidly evolving with the development of smaller receivers built as part of other consumer products such a mobile phones, and it is likely that GPS receivers will soon be integrated with accelerometers to provide a small, single instrument for recording both activity and location. The development of standard methodologies for the combination and interpretation of these data, particularly given the episodic nature of GPS data, will greatly aid such studies.
Strengths and limitations
This study had a number of strengths. The large sample of children was recruited to be representative of the city, encompassing a range of physical and social environments. In addition there was good compliance with wearing GPS receiver, with 80.7% of children who provided accelerometer data also providing GPS data on at least 1 day and 65% of children providing data on more than one day. However there are also a number of limitations. We defined any epoch with a GPS record as being outdoors since the GPS receiver used in this work is unable to receive a signal from the satellite network within a building unless in close proximity to a window. The signal is rapidly lost when moving into the building and whilst a large majority of GPS data are likely to be recorded when the participant is outdoors, we cannot rule out that some data may be recorded indoors. In addition, GPS data are only recorded when the instrument is switched on and connected to the satellite network. Children were trained to switch the instrument on and off, and were asked to turn it off at the end of the day to preserve battery life, turning it on again when leaving school. It is possible that some children may have failed to switch the instrument on when leaving school and that time outdoors may therefore be underestimated. Figure 2 shows a small peak of higher physical activity after school in data identified as "indoors" which is likely to be data recorded outdoors but where the GPS had not connected to the satellites or was not on. The extent to which this has happened is unknown and may result in accelerometer data defined as "indoors" actually being recorded outdoors. However, since the impact of this would be to attenuate the difference between physical activity indoors and outdoors, this would only serve to strengthen the associations reported here.