| Moss biocrust is an important desert plant in arid and semi-arid regions.It has important ecological significance for wind protection and sand fixation,and improving soil moisture and nutrition.Chlorophyll,as the first stage of photosynthesis in the process of plant growth,plays an important role in controlling the energy transfer and material circulation of plants.The biomass of plants is an important indicator of plant physiological conditions,plant stress,and characterization of nutritional status,providing an important basis for the relationship between plants and the environment.The nitrogen content of plants is an important indicator for evaluating plant growth,and it is also an important factor for studying global changes and migration of nutrients.Hyperspectral remote sensing technology can monitor the dynamic changes of vegetation in real time in a short period of time and supplement the spectral research results of bryophyte biological crusts in arid regions,providing a scientific basis for the development and protection of desert ecosystems in arid regions.This research area is located in Gurban,which passes through Beishawo of Fukang City in the south of the special desert.It uses ASD spectroscopy to collect the hyperspectral data of bryophyte crusts,and the crust chlorophyll,biomass,and total nitrogen measured by the laboratory.Measured values,an estimation model based on hyperspectral data and biochemical parameters of moss biological crusts was established.Analyze the basic spectral characteristics of moss biological crusts in the study area,and the correlation between the measured spectral reflectance and the three biochemical parameters.Select the relationship between the bryophyte crust,chlorophyll,biomass,and total nitrogen.Higher characteristic bands.PLSR,BP neural network and support vector machine(SVM)models were used to establish an estimation model of bryophyte chlorophyll,biomass and total nitrogen based on measured spectral data.The correlations between biochemical parameters of moss biological crusts and reflectivity of remote sensing images were measured,and sensitive bands were selected through correlation analysis.An estimation model of moss bioclastic biochemical parameters based on Landsat-8 remote sensing image was established,and the coefficient R2 was determined.The RMSE of modeling total root mean square error and the residual prediction bias RPD were used as indicators of model accuracy evaluation,and the model was evaluated.Finally,the best model for predicting bryophyte biological crust based on hyperspectral data and remote sensing image data was obtained.The results show:(1)Using the measured spectral data,analyze the basic spectral features of moss biological crusts and the spectral curve features of three mathematical transformation forms of spectral reflectance: first differential,second differential and reciprocal logarithm.The measured spectrum was analyzed and the results showed that the reflectance values were low in the 350-700 nm band,there was an absorption valley near 500 nm,there was a reflection peak near 600 nm,there was an absorption valley near 680 nm,and the reflection near 700 nm The rate began to rise,the slope of the spectral curve was larger,and the trend of the curve was steeper and steeper.In the range of 900 to 1300 nm,the curve showed a trend of fluctuating rise.The reflectivity value showed a continuously fluctuating change,and the reflectance value reached the highest near 1000 nm;in the band Between 1300 and 1400 nm,the spectral reflectance showed a sharp decline,but the reflectance level was still higher than 350 to 700 nm.There are three significant absorption valleys at 1200 nm,1400 nm,and 1900 nm,respectively.It can be concluded that first-order differential and second-order differential are the best forms of spectral reflectance.By using the two transform methods,the subtle characteristics of spectral reflectance can be amplified,which is beneficial to observation.There are small spikes at 400 nm,750 nm,1000 nm,1100 nm,1150 nm,and 1800 nm,indicating that these bands have a good correlation with biochemical parameters of bryophyte crusts.(2)Correlativity analysis of chlorophyll,biomass,and total nitrogen of bryophyte crusts with first-order differential and second-order differential of spectral reflectance.Correlation analysis results showed that the correlation coefficient of spectral reflectance first-order differential and chlorophyll content reached an extreme at 400 nm,and the correlation coefficient was 0.42.At 1000 nm,the correlation coefficient of spectral reflectance first-order differential and biomass reached an extreme value.The correlation coefficient is-0.68;the correlation between the first-order differential of the spectrum and the total nitrogen content shows extremum at 750 nm,1400 nm,1650 nm,1800 nm and 2300 nm,and the maximum correlation coefficient is-0.5.From the correlation coefficient of spectral reflectance second-order differential and biochemical parameters of bryophyte crusts,the correlation coefficient of chlorophyll content is better,with the largest correlation coefficient reaching above 0.4 and the biomass appearing in the 1000 nm wave band.The correlation coefficient reached 0.68,and the absolute values of the correlation coefficients at the other wavelength bands of 500 nm,1850 nm,and 1900 nm all exceeded 0.4.In the correlation between the second-order differential of the spectrum and total nitrogen,the wavelength range near 750 nm,1300 nm,1650 nm,1700 nm,and 2200 nm is the main spectral response area of total nitrogen,and the correlation is good.The absolute value of the correlation coefficient can be maximized.It is 0.47.(3)Using PLSR,BP and SVM methods to establish an estimation model of biochemical parameters of bryophyte crusts based on measured spectra.Biomass is the best biochemical predictor of bryophyte crusts in terms of model stability and predictability.The parameter,BP,is the best model for estimating chlorophyll,biomass,and total nitrogen of bryophyte crusts,where the coefficient of determination R2 is 0.53,the root mean square error is 0.05,and the residual prediction bias RPD is 1.91.(4)The reflectivity values of the seven characteristic bands extracted from the Landsat-8 image were selected and correlated with the chlorophyll,biomass,and total nitrogen of the bryophyte crust,respectively.The parameters with the highest correlation among the feature bands were biomass.The correlation coefficient reached 0.63.In the third wave band,the correlation coefficient between chlorophyll and biomass was high,and the correlation coefficient values were 0.38 and 0.53.There was no correlation between the 7th band and the three biochemical parameters in the bryophyte crust.Chlorophyll reached a significant correlation level only between the 3rd and 4th bands;the total nitrogen content was correlated with the 1st and 2nd band values and the 5th and 6th band values,respectively.The wavebands related to biochemical parameters of moss biological crusts were selected for modeling analysis.(5)The PLSR,BP,and SVM estimation models based on remote sensing images were used to model the reflectivity values of the sensitive bands in the Landsat-8 remote sensing images as model input parameters.The results show that the BP model based on remote sensing images is the best for moss crusts.The model,the coefficient of determination of the model R2 is 0.43,the root mean square error RMSE value is 0.29,and the residual prediction bias RPD value is 2.01.From this,it can be concluded that the total nitrogen content is the best parameter for estimating bryophyte crusts. |
PDF Full Text Request