Studies On Monitoring And Evaluation Methods Of Forest Soil Respiration

Soil is a large carbon pool, total storage capacity reached 1394 Pg C. Soil respiration is accounted for 60% -90% of the whole terrestrial ecosystem respiration, which may cause large change in the concentration of atmospheric CO2 with its slight variation, and it is the key process of global climate changes. Forest is the main body of terrestrial ecosystems, and its soil carbon storage accounts for 73% of global soil carbon pool. Therefore, exploring the means of monitoring and evaluation methodologies of forest soil respiration, is not only the strategic importance for the carbon sink measurement, carbon trading, and climate change mitigation, but also important guiding value for forest carbon stocks management.Firstly, the paper studied the soil respiration monitoring platform, then the spatial distribution of soil temperature, using the soil temperature to guide the monitoring points arrangement theory and the soil temperature as soft data combined with Bayesian maximum entropy method to improve the reliability of the estimation of soil respiration. Finally, we developed the software of the soil respiration analysis and evaluation, providing a convenient method for researchers to analyze and evaluate the soil respiration. This paper mainly formed the following conclusions:(1) We researched the soil respiration apparatus based on gas diffusion law, setting multilayer sensor in the instrument to get the data and establishing the CO2 diffusion model in the instrument to calculate the value of soil carbon flux, and its average error is about 10%. Then we used the AD-HOC and GPRS wireless communication technology to form the soil respiration distributed monitoring network, to achieve multi-point, synchronization, real-time monitoring. The results showed that the average error of soil respiration monitoring with only one sampling point in the experimental area was about 42%, which indicated that the single point monitoring could produce large errors, and the monitoring error could be reduced by using distributed monitoring network.(2) This paper researched on spatial distribution of soil temperature method based on BP neural network. For the uncertainty and nonlinear characteristics of soil temperature distribution, we proposed block training, combinatorial optimization improved BP neural network algorithm, and the number of hidden layer nodes of this algorithm was set to 10 and the learning rate to 0.75, which the algorithm achieved the better performance. And comparing the real and predicted values with the R2 as evaluation criteria, the R2 value of spline interpolation algorithm was 0.59, Kriging for 0.73, and BP neural network algorithm was 0.92, while the R2 value of improved BP neural network algorithm was 0.97, indicating that this algorithm could effectively improve the reliability of the spatial distribution of soil temperature.(3) Research on the soil respiration monitoring points optimization layout method based on soil temperature gradients and image segmentation algorithm, determining it was more reasonable that the amount of soil respiration monitoring points was five according to the soil temperature, and using region growing image segmentation algorithm to determine the position of the five monitoring points. Then using the average error rate as evaluation standards, the uniform grid to determine the position was 24%, and the random arrangement method was 29%, but the region growing image segmentation algorithm was 15%, indicating this method could be used to improve the reasonable of soil respiration monitoring point arrangement.(4) Study on soil respiration evaluation based on Bayesian maximum entropy method combined with soil temperature, and Bayesian maximum entropy algorithm with soil temperature as auxiliary information, using the functional relationship between soil temperature and soil respiration to integrate the soil temperature as soft data into the soil respiration estimation, which could improve the accuracy of soil respiration spatial interpolation algorithm. The average correlation coefficient (CR) of Bayesian maximum entropy algorithm (BME) was 0.745, significantly higher than the Co-Kriging algorithm (Co-OK) method that was 0.505 and the ordinary Kriging algorithm (OK) method that was 0.267; the root mean squared error (RMSE) of BME was 0.568. which was lower than the Co-OK method that was 0.851 and OK method that was 1.511; the absolute average value of mean Bias of BME method was 0.192. while the Co-OK method and OK method were 0.508 and 1.143. The experiment showed that the performance of BME method was better than the Co-OK method and OK method. Meanwhile. BME algorithm could be used to reduce the number of soil respiration monitoring points.(5) Use SQL Server database to store sensor data. ArcGIS to establish soil respiration related image layers, and Visual Studio 2008 to develop the soil respiration analysis and evaluation software, to provide the facilitate for users to study soil respiration.The results provided an optimized method for monitoring and evaluating the forest soil respiration, and the necessary technical support for further study of dynamic changes of soil carbon and the guidance of carbon sinks and forest resource management.