| Forest is the main part of terrestrial ecosystem, and the carbon cycle of forest plays a highly important role in terrestrial carbon cycle system. In this study we optimized and verified the BEPSHourly model on site scale. We also explored the application method that combines the BEPSHourly and BEPSDaily model on a regional scale.Firstly, we collected the drive and validation data for BEPSHourly model on site scale, then iterative method was used to optimize the main photosynthesis parameters:the maximum carboxylation rate(Fc max) and the maximum electron transport rate(Jmax). Based on the research results on biomass allocation of different stand types in Northeast, we optimized the biomass allocation parameters of different stand types. Further, considering the influence of the forest canopy secondary permeation phenomenon and multiple reflection, we optimized the radiation transfer model. With the optimized radiation transfer model, then we can calculate the net radiation more accurately. Accordingly, the analog capability of BEPSHourly model on estimating latent and sensible heat fluxes can be enhanced. Using the data assimilation algorithm to calibrate the predicted soil temperature by BEPSHourly model, we reduced the error accumulating over time. Through this model, the soil temperature and the changes of snow depth on earth surface were simulated and verified, and the canopy temperature was simulated. We also did sensitivity analysis on the main influence factors of NPP, such as leaf area index and meteorological factor (temperature, rainfall, wind speed, total solar radiation and relative humidity).Moreover, we collected the regional drive data for BEPSDaily model on regional scale. Since the existing MODIS LAI products can’t meet the application requirements, we proposed a novel time series forest LAI estimation method that is suitable for small and medium-sized region scale. We compared the characteristic and applicability of BEPSDaily and BEPSHourly model respectively, particularly analyzing the reliability of passing the main photosynthesis parameters Vc max and JmaX from BEPSHourly model to BEPSDaily model. Based on this reliability analysis, we established the method for BEPSHourly and BEPSDaily model to collaborate. With the optimized BEPSHourly model on site scale, the law of GPP and NPP changes within a day was simulated and analyzed. What’s more, we achieved the estimation of regional NPP and the space distribution of carbon sink/source with the optimized regional BEPSDaily model.In conclusion, the results of the research indicate:1. Optimal Vc max and Jmax for the mixed forest in Northeast China were 41.1 μmol·m-2·s-1 and 82.8 μmol·m-2·s-1 respectively, with the minimal RMSE and the maximum R2 of 1.10 g C-m 2·d-1 and 0.95. After Vc max and J max optimization, BEPSHourly model can simulated the seasonal variation of GPP better.2. The proportion of leaves in aboveground biomass of broad-leaved forest is 4 percent, and that of limbs is 96 percent. For mixed forest, the proportions for leaves and limbs are respectively 5 percent and 95 percent. For coniferous forest, the proportions for leaves and limbsare respectively 6 percent and 94 percent. There was a significant linear correlation between the aboveground biomass and the underground biomass. So we optimized the ratio of underground and aboveground biomass into a linear equation.3. After the optimization by taking the influence of the forest canopy secondary permeation phenomenon and multiple reflection into consideration, the R2 of the simulated latent heat flux increased from 0.769 to 0.792, and the RMSE decreased from 50.77W/m2 to 47.84W/m2. The R2 of the simulated sensible heat flux also increased from 0.684 to 0.705, and the RMSE decreased from 48.42 W/m2 to 45.86 W/m2. The simulated value was significantly correlated with the measured value of latent and sensible heat fluxes. The changes of the latent and sensible heat fluxes within a day was a unimodal curve, and it up to its peak at noon and down to its valley at night or before dawn. Comparing to the sensible heat flux, the latent heat flux changes more significantly on seasonal scale. The latent heat flux on growing season is much higher than that on non-growing season. On one hand, it means that the latent heat flux is positively correlated with plant growth. On the other hand, the sensible heat flux seems to be off-peak when the plant is at vigorously growing stage.4. The RMSE between simulated soil humidity and measured soil humidity was reduced from 0.1198 to 0.0293 after the measured data was introduced and assimilated at an interval time of 0.5 hour, and the simulation result was obviously improved. The higher frequency the assimilated data was introduced, the smaller RMSE was, i.e., the better the simulation result was. When the frequency was down to 15 days, the assimilation algorithm did not show any significant improvement.5. There was a non-linear positive correlation between NPP and LAI, and the sensitivity S was 0.292, class III. At first, NPP increased as the temperature increased, but when the temperature kept increasing, the NPP started to decrease. Its sensitivity S was 0.954, class IV. NPP remained unchanged when the rainfall increased. This means that the current rainfall which changes from-30% to 30% did not inhibit or promote the plant growth. Its sensitivity S was 0.0005, class I. NPP was non-linear positively related with the solar radiation, and its sensitivity S was 0.310, class Ⅲ. NPP did not change when the wind speed changed and its sensitivity S was 0.0015, class Ⅰ. NPP was non-linear positively related with relative humidity, and its sensitivity S was 0.159, class Ⅱ.6. In this research, we extracted the normalized growth curve of LAI from MODIS LAI product to estimate the daily LAI in regional scale, and we estimated the maximum value of LAI from traditional remote sensing statistical model, then we simply multiplied these two value to get the time series LAI. We applied this method on small and medium regional scale LAI estimation, and compared the results with the measurements. The comparison results showed that the RMSE from two different sample plots in mixed forest was 0.40 and 0.49 respectively, and the RMSE from one sample plot in broad-leaf forest was 0.59. So this method is simple and fast, and it can provide effective time series LAI data for other research in small and medium regional scale.7. The main photosynthetic parameters optimized by BEPSHourly model can be directly introduced into the BEPSDaily model to achieve higher simulation ability; GPP and NPP have significant seasonal variations, presenting a positive correlation with the vegetation growth and temperature increase; At the regional scale, using the optimized BEPSDaily model, the average values per unit area of GPP and NPP of Maoershan in 2011 were 1265.56 g C·m-2·a-1 and 628.40 g C·m-2·a-1，respectively;The primary productivity of different forest types for ascending order was deciduous broad-leaved forest, mixed forest and coniferous forest. |
PDF Full Text Request