Simulated Spatial Patterns Of Soil Respiration In Pangquangou Nature Reserve Based On The MODIS Data

Carbon cycle in terrestrial ecosystems is an important component in the global carbon cycle. Soil respiration (Rs) is a major component of CO2 exchange between terrestrial ecosystems and the atmosphere. Though research of soil respiration has shifted from the initial field observation mostly at plot-scale to regional modeling construction of soil respiration at large-scale, remote sensing technology was rarely applied to the simulation of soil respiration. In this research, based on the remote sensing parameters, such as the land surface temperature (LST), which was associated with soil temperature, and vegetation index (VI), gross primary productivity (GPP), leaf area index (LAI), which were associated with vegetation growth, we explored the application of the parameters on soil respiration simulation at a larger area.The experiment was conducted in Pangquangou Nature Reserve of Shanxi province. In the study,31 sites with different types of vegetation at different altitude above sea level in the area were selected to conduct this experiment. Soil respiration was measured in July, September, and November, 2013, respectively. At the same time we downloaded the products of MODIS, including land surface temperature (LST), vegetation index (EVI and NDVI), gross primary productivity (GPP), and leaf area index (LAI). By using of correlation and regress analysis, we selected the factors which had significant effect on soil respiration to construct single factor models and compounded models estimating soil respiration. On this base the models were valuated and extrapolated to the whole area. The main conclusions were as follows:(1) The spatial and temporal variability of soil respiration rate was basically the same. The averages of soil respiration rate were 8.70 μmol CO2 m-2s-1,6.65 μmol CO2 m-2 s-1 and 2.82 μmol CO2 m-2s-1, respectively in July, September and November. Comparison of the mean for different measuring time showed a significant different (P< 0.01). The variable coefficient (CV) of soil respiration rate was 16.99%,26.19% and 36.40%, respectively, in July, September and November. The degree of spatial variation of soil respiration increased with the change of seasons from to July to November. A significant correlation between the measured soil temperature and the land surface temperature existed, showing the soil temperature could be replaced by land surface temperature for the study of soil respiration to some degree.(2) The relationship of soil respiration and both the land surface temperature and soil temperature at 10 cm depth (T10) was significant in three measurement months (P< 0.05). The correlation coefficients between Rs and LST were higher (0.55-0.72) than those between Rs and T10(0.55～0.69). Among the analysis of soil respiration with the parameters relating vegetation growth, the relationship of soil respiration with vegetation index (NDVI and EVI) was more significant than that with the others. The relationship of soil respiration with EVI was significant, but not with NDVI in July, and the relationship between soil respiration in September and November and both NDVI and EVI was significant. A further analysis showed that a model of soil respiration with single factor of the mean value (LSTav) of the four LST data on a day was better than that with the separated LST, and NDVI was better than EVI. So, we used both the LSTav and NDVI as independent variables to construct different confounded models. On this base we made precise evaluation for the models using R2, RMSE (Root Mean Square Error) and AIC (Akaike Information Criterion) values, and decided which one was the best one among the models. Comparison analysis of the measured and the modulated soil respiration showed that coefficient of determination from the confounded equations was larger than that from the single variable equations, and three fitted equations could be used to estimate soil respiration in the area.(3) The simulated spatial distribution of soil respiration in the area showed that that the high value of soil respiration appeared in northwest and south with the vegetation types of aspen and oak, and the low value occurred in the meadow located in the northeastern of the area. The distribution of the simulated soil respiration was consistent with the measured values of soil respiration. The accuracy of the model was tested by measured soil respiration in permanent sample sites in other periods. Therefore, the models based on remote sensing could be used to extent soil respiration to a large scale.