| Soil organic matter (Organic Matter, OM) is the important part of thewater-level-fluctuation zone (WLFZ). It contains the main elements that required for plantgrowth, and is the energy source of soil microbial activity. As an important indicator of soilfertility, The impact of OM content on wetland ecosystem productivity is very significant.Simultaneously, soil organic matter content has a more sensitive response to climate change, tobe able to indicate the climate change to a certain extent. The WLFZ as a transition zone betweenthe terrestrial and aquatic ecosystems, could improve the water ecosystem productivity andmaintain the dynamic balance of regional ecological system, which has a great ecological servicefunction value. Therefore, researches of OM distribution and dynamic change in WFLZ is ofgreat significance. Due to special hydrological conditions of seasonal alternate wetness anddryness, and disturbance of human activities, the soil organic matter content has a high level ofvariability in time and space, and its distribution and influence mechanism is more complex. Onthe other hand, soil organic matter has unique spectral response characteristics in the visible andnear infrared bands. Using remote sensing technology could determinate soil fertility in regionalscale rapidly, and more quickly master the spatial distribution pattern of soil organic matter,which can provide theory and technical support for further study on estimation of soil organiccarbon storage and analysis of carbon cycling in wetlands.In this paper, selecting typical WLFZ in Guanting reservoir, analyze the spatial distributioncharacteristics of soil organic matter and the influencing factors, which will provide basic dataand some basis for soil quality evolvement process. By means of ASD spectrometer, soil spectralreflectance was measured indoor. According to the spectral response characteristics of soilorganic matter, use partial least squares regression (PLSR) method to estimate the soil OMcontent in study area. And based on the estimation model, use hyper spectral image data ofsampling area to estimate OM content, and last got the map of its spatial distribution pattern,which will provide technical support for synchronous dynamic remote sensing of soil fertility inlarge scales. The results showed that:(1) The OM content of WLFZ soil in Guanting reservoir was fairly low, and changed in the range of1.64~26g/kg, with average value was13.16g/kg. The coefficient of variation was50.59%. Which showed that the distribution of soil nutrients in WLFZ had higher heterogeneity,because of the dry-wet alternate hydrological conditions and human activities. The OM contentof frequent flooding area was15.74g/kg in average, higher than the contrasting long-termoutcrop area of10.12g/kg. And the coefficient of variation was41.38%in frequent flooding area,lower by comparison with those in control zones with54.98%. It indicated that the soilnutrient retention ability in frequent flooding area was stronger, and differences of soil organicmatter between different sampling points relatively small. Under different plant communities,OM content of Phragmites australis and Typha angusitifolia communities was the highest, theaverage value was17.088g/kg; the lowest content was in Populus simonii Carr and Elymusdahuricus Turcz communities, the average value was9.12g/kg; then was the Dockleafknotweed and Cirsium setosum (Willd.) Bied communities, OM average value was15.49g/kg. Different depth of soil, OM content was different. The overall trend was from thesurface downward gradually reduced. Organic matter of each soil layer showed significantdifferences (P<0.05). The soil C/N changed in the range of1.64~18.95, the average value was8.95. That demonstrated the C/N of study area was relatively low, organic matter had a higherhumification degree, organic nitrogen tend to accumulate, potential of soil anaerobicdecomposition to produce CH4and CO2was larger and faster. In the vertical distribution, C/Nincreased first and then decreased as the depth of soil profile change, and reached the maximumvalue in30cm.(2) Soil organic matter and total phosphorus, total N and C/N were positively correlatedsignificantly, coefficients were0.62(P<0.01) and0.57,0.60(P<0.05), respectively. This showedthat the soil organic matter and total phosphorus, total N and C/N had same change trend, andexisted interaction. Secondly, soil organic matter and moisture were negatively correlated (r=-0.51) at the0.05level, indicating that soil moisture in the study area had a significant impact onthe organic matter content. Temperature had a significant influence on the distribution of soilorganic matter in the study area, the correlation coefficient was-0.508, significance levelP=0.031<0.05. Vegetation cover and soil organic matter content was significantly positivelyrelated on the significance level of0.05, indicating that the vegetation factor was also one of theimportant factors affecting the distribution of soil organic matter. (3) Soil spectral reflectance characteristics of the study area showed that, in the wholewavelength range, soil spectral curve was relatively flat, overall waveforms were similar, whichdisplayed as convex parabolic curve. Spectral reflectance of soil after air dry was significantlyimproved, especially after grinding sieving. The higher soil organic matter content, the lower thespectral reflectance. In the same soil profile and different depth, spectral curves were nearlyparalleled, and reflectance increased with the increase of soil depth. The original spectralreflectance was transformed as first order differential form, second order differential form andabsorbance conversion, and the spectral curves after those transformations were easier to detectthe small changes in soil spectra. The standard deviation of soil reflectance at each wavelengthindicated that impact of organic matter on soil reflectance mainly in the400-1358nm wavelength,and caused the biggest difference in600-800nm. The correlation analysis between OM contentand spectral value of reflectance and their transformations showed that OM content and secondorder differential spectral of absorbance has the highest correlation, and the maximumcorrelation coefficients were at501nm and616nm wavelength band, respectively0.84and-0.83.(4) The PLSR model accuracy test results showed that the second order differential of soilabsorbance had a better estimation for organic matter. The coefficient of determination R2cvinmodeling group sample reached a maximum value of0.84, root mean square error RMSEcvwassmallest, only2.02. And the root mean square error RMSEPin validation group was3.02, R2Pwas0.81. The model accuracy was92.09%. According to the cross validation and accuracyevaluation indexes, the best model was:y=9.41+24999x425+120814x501-323059x616+1174740x834+1167050x861independent variables were second order differential of soil absorbance at425nm,501nm,616nm,834nm,861nm band, respectively.
|
PDF Full Text Request