| Facing the world food crisis, rice yield is very important to national grain security and resist the world food crisis. However, a huge population and limited acreage is China’s basic national conditions. Based on the basic national conditions and the status of crop production, the researches on increasing crop production, monitoring planting area and estimating yield have the vital significance for food security. The topography of Low Mountain and Hill is dominant in Hechuan, Chongqing. The model of rice yield estimation, taking full advantage of the terrain feature of Hechuan, has important implications for research on the methods of rice yield estimation, food security and climate change to Chongqing and even the South Downs Area. In addition, it also establishes a foundation for the improvement and industrialization of the accuracy of rice yield estimation in Low Mountain and Hills. And then, this model can be severed as a reference for Chinese food production, food security and environmental protection.In this article, data of EOS/MODIS-EVI from2003to2012and statistical data of rice yield and plant acreage at village level from2006to2009derived from surveys, combining information of phenology offered by Chongqing Municipal Meteorological stations, data of administrative division, land utilization and DEM was applied in the model of rice yield estimation. After the pretreatments of remote sensing data, the time-series data of EVI is reconstructed using S-G filtration, AG function fitting and D-L fitting method, and then the results are evaluated. Time-series analysis methods including Theil-Sen Median trend analysis, Mann-Kendall test method and Hurst exponent are employed to analyze the dynamic characteristics of vegetation coverage in temporal and space over10years in Hechuan. By linear regression analysis, the model of rice yield estimation is established in the entire and distinct ecological zones, the optimal model is selected and analyzed at the same time. And then these models are utilized in the forecast and error analysis the rice yield in2009. The main conclusions are as follows:(1) Three fashionable and classic reconstruction methods of time-series data of vegetation index are applied to reconstruct the EVI data. AG function fitting method has drawn the optimal results, after a direct comparison and quantitative analysis of the reconstruction results.(2) The inter-annual change characteristics of vegetation cover are as follows: the overall situation is good, the overall trend is growing, and the average annual values of EVI are varied from 0.38to0.44. Spatial distribution shows a trend of higher in the central valley area, lower in the hilly and broad valley area of northeast, the lowest in the mesa and hills area of northwest and the deep mound and low mountain area of southeast. Additionally, it shows a significant coupling with the distribution of land use and geography. Furthermore, the mean values of EVI derived from many years performs that agriculture vegetation is greater than evergreen vegetation.(3) The inter-seasonal change characteristics of vegetation cover are as follows:the overall performance of the mean EVI values show that the mean value of summer is higher, the mean values of spring and autumn is lower, and the mean value of winter is the lowest during the past10years. Evergreen vegetation mainly distributing in the northwest and southeast, since the growths of agriculture vegetation and evergreen vegetation are different, and the growth of agriculture vegetation is strong than evergreen vegetation, the EVI distributions of these areas mostly show purple (low value). Meanwhile, agriculture vegetation mainly distributing in the he northeast and central part, owing to crop growing season in summer, the EVI values of these area are high, and the distributions mostly show yellow.(4) The change characteristics of vegetation cover presented that areas of vegetation improvement is in the absolute advantage, which is much larger than areas of degradation. The trends of each ecological zone is as follows: the hilly and broad valley area of northeast and the deep mound and low mountain area of southeast are better, the central valley area is the second, the mesa and hills area of northwest is relatively poorer.(5) Sustainability analysis of vegetation cover changes in Hechuan indicates that tendency of vegetation changes of the study area, the pixels with strong consistency and continuity are in a dominant position. Sustainability of vegetation cover changes in each ecological zone performed: the deep mound and low mountain area of southeast is stronger, the central valley area and the hilly and broad valley area of northeast are strong, and the mesa and hills area of northwest is poor. According to the spatial distribution, the regions of vegetation changes significantly and heavily influenced by human activities, whose Hurst exponents are often and almost relatively small.(6) The dynamic characteristics and trends of vegetation cover change are as follows: the pixels with improvement and sustainability in the future, dominate the absolute advantage. Each ecological zone mainly distributes the pixels with the feature of no obvious improvement and sustainability. The vegetation changes trend and sustainability of the hilly and broad valley area of northeast manifests that no obvious improvement with continuous feature and no obvious degradation with continuous feature are dominate. However, the change trend and sustainability in the other three regions is in common, all of them are given priority to characteristics of obvious and no obvious improvement with continuous feature. (7) A significant positive correlation (p=0.01) was obtained by Pearson correlation analysis between the EVI spatial cumulative values and yield of rice. Under the whole region, the EVI spatial cumulative values of image phases of137and153have the highest spatial correlation with the yield of rice. The EVI spatial cumulative values of image phases of137(the jointing stage) and153(the booting stage) also have the highest correlation in the central valley area and the hilly and broad valley area of northeast; while the EVI spatial cumulative values of image phases of217(the milk stage) and233(the ripe stage) have the highest correlation in the mesa and hills area of northwest is poor and the deep mound and low mountain area of southeast. Among the various ecological zones, correlations between each growth stage of EVI spatial cumulative value indicate that the correlation of low elevations perform greater than high elevations.(8) Based on statistics and correlation analysis pattern, the rice yield linear estimation models are established using remote sensing data for the whole region and each ecological zone. The accuracy analysis of the models indicate that the estimation models of each ecological zone are more adaptable and the accuracy of forecast results are higher. The higher average elevation areas like the deep mound and low mountain area of southeast and the mesa and hills area of northwest, the best and optimal growth stages concentrate on the generative growth phase; while the lower average elevation areas like the central valley area and the hilly and broad valley area of northeast, the best and optimal growth stages concentrate on the vegetative growth phase. Especially the EVI image of phase of137(the jointing stage) is the best and optimal for the whole region and the lower elevation areas. In the end, the prediction results of rice yield in2009are superior by the optimal models of the entire and each ecological zone.In a word, the article analyzes the dynamic characteristics and trends of vegetation cover in temporal and spatial over10years and establishes the rice yield linear estimation models for the whole region and each ecological zone, after reconstructing the EVI time-series data. The results indicate that the models of each ecological zone are more adaptable. And then the results provide a basis and foundation for further study rice yield estimation and estimation industrialization in low mountains and hills. |
PDF Full Text Request