| The seasonal cycle of crop growth is sensitive to environmental change. The recurrence timing of crop phenology reflects the farmland ecosystem’s responses to environmental shocks, casting important implications to the relationships between climate change and agriculture. Rice is a most important grain crop in China. Improvements in the understanding of spatiotemporal patterns of rice’s phenological changes are not only essential in the management of the rice production system, but also indispensable in ensuring food security and in stimulating agricultural policy. Characterized as being a timely and objective means, satellite remote sensing has the capacity to effectively detect phenological changes of agricultural crops over large areas. However, the remotely sensed phenological data is derived from the emissivity information of surface vegetation at the pixel level, which is apparently different from field phonological observations logged at long-term agricultural experimental stations. This important discrepancy shows the necessity of calibration of the remotly sensed crop phenology prior to any applications.Therefore, the MODIS phenological products during 2007-2011 were used in this thesis to investigate the spatiotemporal changes of rice growing season in Heilongjiang province in Northeast China. Other data sources include crop phenological records at long-term agricultural experimental stations, climatic dataset, spatial distribution of rice areas, etc. in Heilongjiang province. A statistical model was built to calibrate the start and end of the rice growing season(SOS and EOS, respectively). The spatiotemporal patterns of the length of rice growing season(LOS) were analyzed and priorities in future research were identified. The main findings include:(1) The single-factor statistical calibration model, which only considers the accumulated temperature as the independent variable, out-performs the multi-factor model which simultaneously considers accumulated temperature and precipitation. Moreover, linear models have higher goodness of fit than nonlinear models. Model comparisons show that accumulated temperature has a higher significance of correlation than precipitation with the detection difference of the end of rice growing season between remote sensing and ground observation(i.e., ?EOS), although both accumulated temperature and precipitation are found significantly correlated with ?EOS. Taking 2007 as an example, the significance level of the single-factor model involving accumulated temperature is as high as p = 4 ? 10-6.(2) The calibrated length of the rice growing season in Heilongjiang shows a distinct spatial pattern along the southwest to northeast diagonal. The LOS of rice in northeastern Heilongjiang tends to be longer than the southwest. The LOS ranges between 118 and 129 days in the northeast, while in southwest it ranges at 96-106 days, with an average difference of 6 days. Furthermore, the LOS tends to be longer at higher latitudes and longitudes at a gradient of 2.1 and 1.2 days per degree, respectively. It has been clear that crops need a longer season to reach maturity in cooler environments. The mostly Japonica varieties of rice in Heilongjiang show no exception to this rule.(3) A shorter growing season is evident for rice cultivation in Heilongjiang during 2007-2011. Analyses reveal frequent delays in SOS, contrasting to a largely constant EOS. Overall, LOS of rice in Heilongjiang has been shortened at a rate of 1.87 days per year during 2007-2011. The trend of a shorter season can spatially be attributed in most of the Sanjiang Plain and Hulan Basin in northern Songnen Plain, although a few areas observe longer seasons, such as upper Nengjiang and Gannan regions. |
PDF Full Text Request