| Grassland ecosystem is an important component of terrestrial vegetation ecosystem in the world.Grassland productivity level is the key to the research of grassland degradation,grassland ecosystem health diagnosis,ecosystem service function evaluation and carbon sequestration.Estimation and prediction of grassland yield can provide important scientific basis for reasonable determination of livestock carrying capacity;it has important theoretical and practical significance for maintaining grassland ecological balance and correctly guiding the development of animal husbandry.In this paper,we took Inner Mongolia steppe as the research area.Then,we systematically analyzed the temporal and spatial evolution characteristics and influencing factors of grassland productivity in the research area,discussed the delayed effects of temperature and precipitation on grassland productivity,and completed the remote sensing estimation and prediction of fresh grass yield by using GIS(Geographic Information System)and remote sensing methods,combining with field survey data,remote sensing data and other auxiliary data,and based on grassland NPP(Net Primary Productivity)estimates.The following are the main work and conclusions:(1)This paper estimated the grassland NPP for each month from 2000 to 2014 based on the CASA(Carnegie-Ames-Stanford-Approach)model.The Pearson correlation coefficient between the estimated results and the MOD17A3H-NPP data reached 0.935.Taking grassland NPP as the evaluation index of grassland productivity and adopting the linear trend method based on pixel,we found that the grassland productivity in Inner Mongolia showed a trend of gradual increase from west to east and from south to north;the accumulation period of NPP in grassland was mainly from April to October.Although there was a little fluctuation in the change trend from 2000 to 2014,the overall growth rate was 0.85g C/(m2·a)(P<0.01);the percentages of regions with extremely significant increase and significant increase were 29.76%and 22.26%respectively.According to the results of Hurst index and the comprehensive analysis of the change trend from 2000 to 2014,13.42%of the grassland productivity in the research area will change from the current increasing trend to the decreasing trend.(2)The paper systematically studied the correlation between NPP of grassland and natural factors in the research area,and discussed the response of grassland to human activities.Based on the correlation analysis method of pixel,it was found that the NPP of the grassland in the research area was negatively correlated with the air temperature and surface temperature on the inter-annual scale,in which the sum of thc proportions of the significant and extremely significant negative correlations accounted for 25.15%and 60.99%respectively;was positively correlated with the precipitation and evapotranspiration on the inter-annual scale,in which the sum of the proportions of the significant and extremely significant positive correlations accounted for 20.34%and 74.25%respectively;and was mainly negatively correlated with the calculation results of the improved TVDI(Temperature Vegetation Dryness Index)proposed in this paper on the continuous monthly scale of the growing season,in which the sum of the proportions of the significant and extremely significant was 49.83%.The correlation coefficient between SNPP(Sum of Net Primary Productivity)mean value and the cumulative area of newly-increased fenced rangeland and annually-planted grass was 0.689(P<0.01)and 0.81 7(P<0.01)respectively.According to the analysis,human activities such as national ecological and environment-related policies have obviously positive effects on the improvement of NPP of grassland.(3)Although the delayed effects of meteorological factors on vegetation growth had been generally accepted by researchers,quantitative research has not received sufficient attention.In this paper,the delayed effects of temperature and precipitation on grassland productivity was studied under the time-scale of three-month delay and the influence of climatic factors at different times on grassland productivity and the inter-annual time scale of different months.The results showed that the effects of temperature and precipitation on vegetation productivity in the growing season had a certain delay,but the degree of influence became weaker as time went on.At the same time,the influence of temperature and precipitation on grassland productivity in different months of the growing season had obvious spatial and temporal differences.(4)Aiming at the inconsistency between the actual samples and the pixels of remote sensing image on the spatial scale,a remote sensing estimation model for fresh grass yield based on spatial scale conversion was proposed.Taking the grassland of Xilin Gol League in 2013 as an example,using NPP and AFY(Annual Field Yield)converted by the spatial scale constructed the scatter plot,and the linear and power functions did the regression.The goodness-of-fit coefficient of the two reached 0.79 and 0.85,and root-mean-square errors were 0.0625 Kg/m2 and 0.0609 Kg/m2 respectively.The validity of the model was further verified by using Landsat 8 remote sensing images,and the model was suitable for grass yield estimation based on medium-low resolution optical remote sensing images.(5)Based on the deep learning method,the delay analysis method based on time accumulation and the remote sensing estimation model of fresh grass yield,the technical process of next year's grass yield prediction was proposed.Taking the grassland yield of Xilin Gol League in 2014 as an example,firstly.LSTM(Long Short Memory Network)and GRU(Gated Recurrent Unit)neural network models were used to predict the temperature and precipitation of the next year based on the meteorological data of the site.Then,taking the optimal temperature and precipitation prediction results as variables and using the time-based cumulative delay analysis principle,the NPP of the grassland growing season in 2014 was predicted.The goodness-of-fit coefficient of the predicted results and the estimated results based on the CASA model reached 0.92.Finally,the forecast of fresh grass yield is realized.Compared with 2013,the relative accuracy of the forecast is 76.38%. |
PDF Full Text Request