Monitor, Evaluate And Analyze Land Salinization Pattern, Process And Ecological Effects In The Yellow River Delta: Using Remote Sensing And Ecological Models

Soil salinization is a globally threat to food security and environmental quality. Moreover, as a global land degradation process, the occurrence of salinization alters the chemical composition and properties of soil, which would finally change the pattern of vegetation growth and greenhouse gas emissions (GHGs) (e.g., CO2, CH4, N2O, NO, H2S) profoundly at large scale. Given soil salinization has been an important ecological issue in global change and its serious economic and ecological consequences, it is urgently needed to monitor the extent of soil salinity accurately and timely, estimate effectively the ecological impacts caused by saline soils and investigate thoroughly the main determinants of soil salinization process. In the case, the knowledge for the pattern, mechanisms and impacts of soil salinization is useful to promote effective land managements, mitigate the trend of land degradation, provide suggestions for sustainable land use programs, amend invalid managements and restore the salinized areas. However, the distribution of land salinity is very broad and varies with the complex environmental and anthropogenic factors, presenting a highly heterogeneity at spatial and temporal scale. The traditional methods are not satisfied with the current requirements of the sustainable management and quantitative monitoring. In order to meet the new demands, new technologies were used in the study to monitor the pattern, model the ecological impacts and assess the mechanisms of the soil salinization.Remote sensing technology is the most convenient monitoring method due to its low-costing, high frequency and large coverage. However, there are many constraints existing in directly using soil spectrum, such as the interference of vegetation. Fortunately, vegetation has unique spectral responses to salinity stress, which facilitates salinization monitoring as an indirect proxy. According to the advantage, remote sensing salinization via vegetation has become a promising trend for recent decades. High spectral and temporal remote sensed data [(Moderate Resolution Imaging Spectroradiometer (MODIS)] were used in the study as they have great potentials in indicating the physiological and phonological responses of vegetation under environmental changes. Therefore, the two kinds of remote sensed data were used in the study to determine the most sensitive bands to salt stress, to construct the salinity vegetation index, which facilitate to quantitatively monitor the extent and degree of land salinization. Direct measurement of vegetation growth and GHG emissions for inventory purposes is impractical at regional or global scale. In addition, GHG emissions are highly sensitive to all factors of soil types, climate conditions and management practices which should be taken account in measurement. It is beneficial to link the available data on GHG emissions to investigate the underlying processes, such as through mathematical modeling. A process-based Denitrification-Decomposition (DNDC) based on the process and environmental factors can meet all these requirements. In this paper the DNDC model was firstly developed, calibrated, validated, with sensitivity analysis, at site scale simulations, to analyze the salinization impacts on plant growth and GHG emissions.The comprehensive and spatially explicit assessments are needed to assist government in developing ecologically sound policies. Few assessments have comprehensively quantified the impacts of multiple anthropogenic activities on salinization as environmental interferences and salinity autocorrelation are largely neglected. This study tried to perform such an assessment by identifying the nature of human impacts on salinization by a versatile GIS-based spatial autoregression (SAR) model.Our findings show that:(1) There is a great potential for using the vegetation hyperspectra of salt-sensitive and halophyte plants to monitor soil salinity. The optimum soil salinity index constructed by the most salt-sensitive bands enlarges the extent and improves the accuracy of salinity monitoring. The hyperspectra of seven typical salt-sensitive/halophyte species and their root-zone soil samples were collected to investigate the relationship between vegetation spectra and soil salinity in the Yellow River Delta (YRD) of China. According to the relationship, the most sensitive bands to salt-stress were indentified and the best salinity vegetation index was constructed. Vegetation index (VIs) examination and band investigation confirmed that the responses of the vegetation differed by species, which explained the vibrations of the VIs in many cases. These differences were primarily between salt-sensitive and halophyte plants, with the former consistently having higher sensitivity than the latter. With the exception of soil adjust vegetation index (SAVI), most VIs were found to have weak relationships with soil salinity (with average R2 of 0.28) and some are not sensitive to all species [e.g. photochemical reflectance index (PRI) and red edge position (REP)], which verified that most currently available VIs are not adequate indicators of salinity for various species. After comparisons of different hyperspectral analysis, wavelengths at 395-410 nm,483-507 nm,626-697 nm,732-762 nm,812-868 nm,884-909 nm and 918-930 nm were determined to be the most sensitive bands. By combining the most sensitive bands in a SAVI form, we finally proposed four soil adjust salinity indices (SASIs) for all species. Satisfactory relationships were observed between ECe and four SASIs for all species, with largely improved R2 values ranging from 0.50 to 0.58. The advantage of SASIs for indicating soil salinity with plant hyperspectra, especially with the halophyte hyperspectra, effectively enlarges the extent of monitoring and increasingly improves the accuracy of detecting.(2) The seasonal integral of EVI (EVI-SI) extracted from MODIS time series profile was verified as the best indicator for soil saline degrees (with R2 of 0.47), which covers the low spectral and spatial resolutions in broadband sensors and solves the problem of monitoring salinization conveniently and quantitatively at large scale. The correlation of EVI-SI (the seasonal parameter) and salinity was found to perform significantly better than non-seasonal vegetation indices derived at single time. Additionally, the correlation of EVI-SI and soil salinity was highly dependent on land cover heterogeneity, and the range of correlation coefficients was up to 0.32-0.85. The relationships of EVI-SI and ECe among different land cover types were then investigated to assess the mapping ability. Results showed that EVI-SI linearly correlated with ECe in cropland with a high model fit (R2=0.85). The relationship of EVI-SI and ECe fit best with a binomial line for the wasteland sites and EVI-SI was able to explain 70% of the variance of ECe. Despite the linear regression model fit was relatively low in mixed sites limited by spatial resolution (R2=0.32), MODIS time series VI data as well as the extracted seasonal parameters still showed a great potential to assess land salinization at large scale.(3) DNDC model performed well for modeling the impacts of soil salinization on vegetation growth and greenhouse gas emissions (GHG), and predicted the key environmental and anthropogenic factors causing the increase of global warming in face of the risk of the severe soil salinization. Model was firstly calibrated with plant biomass inversed from the remote sensed data, and the seasonal biomass from cotton and reed fields were well simulated by DNDC. The maximum absolute discrepancies between modeled and regressed biomass in the cotton and reed field were 6.73% and 16.93%. DNDC model also satisfactorily performed the trends of the daily biomass during the growing season (high r2>0.95), the daily CO2(r2:0.76) and NO2 (r2:0.94) emissions at cotton site, and the daily NEE (r2:0.88) at reed site. With the increasing of soil salinity, the accurate simulation showed that the average biomass decreased (cotton:1900-1412 kg·C·ha-1, reed:8184-5284 kg·C·ha-1), the average dSOC significantly decreased (cotton:4204-3783 kg·C·ha-1, reed:2239-1469 kg·C·ha-1). However, the changes of soil N pool and the CO2, N2O emissions were not significantly influenced under salt salinization. Through a series of sensitivity tests, the biomass, CO2, N2O emissions would be remarkably changed when the intial SOC, clay content and irrigation codition exceed the normal range, especially for the highly saline areas. The amplitude of these changes were much larger in the higly saline areas than in the lowly saline areas. Moreover, the predicted N2O emission in the reed site of the YRD was a little high (4.56 kg·N·ha-1·y-1) compared to other grassland worldwide, which may also caused by soil salinization.(4) The land salinization occurring in the YRD was proved to be an autocorrelated land ecological process. SAR sub-region model has the advantages of not missing important spatial information and obtaining the unbiased, quantitative results, which is superior to the previous assessments. GIS-SAR model fit better and performed better in quantifying human activities, compared to the conventional ordinary least square regression (OLSR) model, as SAR can deal with spatial autocorrelation in soil salinity. Natural and land use history factors were validated to have great influences to soil salinity besides human activities, so such environmental interferences should not be neglected. GIS-SAR model could perform an informative and accurate assessment by unbiasedly identifying the most important human determinants, ascertaining the nature of anthropogenic impacts on salinization. Sub-region model was verified as an effective tool of normalizing environmental interferences because more useful spatial information was provided compared to the whole region model. Among the well-defined key determinants, oil exploitation and saline aquaculture were aggregative to salinization but only in originally highly saline sub-regions, such as coastal zone and Gleyic Solonchaks (coastal saline moisture soil) area. Two agricultural activities, crop plantation and fertilization, were mainly ameliorators in most sub-regions. The most effective salinization alleviation occurred in moderately saline sub-regions, such as floodplain and Salic Fluvisols (saline moisture soil) area, which benefitted from the development of agro-forests and farm ponds. The SAR sub-region model is spatially explicit for spotting the hazardous areas and some suggestions were also provided for the policy makers.