| China is one of the most serious countries in the world which is suffering from landsandification. The velocity of sandification now is slower than that in the 1990s, and the trend ofsandy land expansion in the whole country has been controlled, however, the areas of sandylands are still expanding in some regions and the whole situation of sandification is still serous.It is a long and arduous task to take the powerful measure for combating sandification.Since 1994 China has implemented three monitoring projects of the national sandificationto aware the types, degrees, areas, distributions and changes of sandy lands, analyze the reasonof dynamics, and provide reliable foundation for macroscopical decision making ofsandification combation. Remote sensing has played an active role in the national sandificationmonitoring. Based on the great information amount, large observation extent, high precisionand quick velocity of remote sensing, it will be more beneficial to successfully implementsandification monitoring to further disscuss the techniques, theories and methods ofsandification monitoring using remote sensing and strengthen the analyse and evaluation ofsandification monitoring results.The paper systematically analyzed the current research situation and trend of sandy landmonitoring using remote sensing at home and abroad, taking Dengkou County of InnerMogolia for example, and Minqin County of Gansu Province and Gonghe County of QinghaiProvince involved, disscussed the rules of spectral characteristics of sandy lands, establishedthe technical system of information extraction and change detection of sandy lands, andanalyzed the dynamic, landscape pattern and grain effect to reveal the process and rule ofsandy land evolvement. The paper deeply understood sandy land monitoring mechanisms usingremote sensing, solved the problems of the large extent, time and energy consuming, increasedthe efficiency of information collection, strengthened the acquaintance ability of sandy landevolvement, and provided the technical support for decision making of sandification combation.The paper has obtained several achievements as follows.(1) Spectral characteristics of sandy lands. Using the spectral instrument of ASD, the fieldspectra of main indicators of sandification process, various surface coverage types and various degrees of sandy lands were measured. The quantitative relationship of spectral characteristicsand sandification was established, the spectral variability rule was found, and the criteria wasprovided for classifying and quantitative analyzing sandy lands in the respect of spectrum. Itwas brought forward that two types of surface features including single surface features,namely various vegetation types and soil types, and compound surface features, namely sandyland with various vegetation coverages and salina with various degrees were measured bydifferent spectral sampling methods. The analyses of spectral characteristics showed thatspectral data of sandy lands changed with vegetation coverage, vegetation type, soil, landcover,etc. With the increasing vegetation coverage, the reflectances of sandy lands decreased.Affected by vegetation and soil moisture, the reflectances in the spectrum of 1300～2500nmwere lower than that in the spectrum of 750～1300nm, the reflectance diffence of sandy landswith different vegetation coverages increased.(2) Information extraction of sandy land. The multi-layer information extraction methodof sandy lands was developed. Sandy lands were first extracted, and then the degrees of sandylands were classified. Using multi-temporal Landsat ETM+images, overrating and underratingsandy lands were avoided. Based on image and spectral characteristic analyses, knowledgeswere introduced using multi-layer extraction method with different formats in different layersto simplify the relationship between various land types, clarify the layer relationship betweenvarious layers during classification process, and realize the separation of sandy lands and otherlands. The basic criteria of sandy land classification was vegetation coverage, detailedly that inclassification standard of the national desertification investigation. By regression models ofNDVI and vegetation coverage, sandification degrees were divided with the thresholds ofNDVI, and the method was an effective way to evaluate sandy land degrees.(3) Change detection of sandy land. To quickly detect the extension or reversion of sandylands, the hybrid method of principal component analysis and supervised classification wasadopted to detect sandy land changes in ten years using two dates of Landsat ETM+/TMimages with principal component analysis producing change areas and supervised classificationproviding detailed change types. It could identify the extension or reversion of sandy lands andprovide scientifica basis for adopting control measures. It also benefited to sandy land mapupdating. Based on the current baseline classification map, the change area was classified to avoid a great lot of labours of field investigation and validation, reduce workload, and increasethe work efficiency. Combination method of multi-time and multi-band in principal componentanalysis was adopted. The principal component bestly reflecting change effect was chosen. Thethresholds were used to separate change and no change areas. Supervised classification wasused to classify the change area and get rid of false change areas.(4) Dynamics and landscape pattern analyses of sandy lands. Based on two periods oflandscape classification maps of 30 mm resolution, markov transformation matrix andlandscape index were used to analyse dynamics and landscape pattern of sandy lands, anddisclose the process and rule of sandy land evolvement. Markov model disclosed thetransformation of different degrees of sandy lands and the transformation of sandy lands andother landscape elements. The whole area of sandy lands decreased and shifting sandy landsincreased. Sandy land management in Dengkou County was still confronted with rigorouschallege. Semi-fixed and fixed sandy land should be renewed culture, and comprebensivesandification control should be regarded to avoid new sandy land come into being. Inclass-level and landscape-level, number of patch (NP), mean patch size (MPS), largest patchindex (LPI), area-weighted mean shape index (AWMSI), area-weighted mean patch fractaldimension (AWMPFD), Shannon diversity index (SHDI), and Shannon evenness index (SHEI)were calculated. It showed that fixed sandy lands and farmlands became more fragmented, andlandscape heterogeneity increased; the fragmentation and heterogeneity of semi-fixed andshitting sandy lands, waterbodies and salina decreased; in landscape-level fragmentation andheterogeneity decreased, landscape pattern developed unevenly, and patch shape wassimplified.(5) Grain effect of sandy lands. By resampling method of majority rule and independentaggregation, landscape distribution maps within grain range of 30～300mm were obtained. Theanalysis of landscape index and spatial autocorrelation coefficient (Moran's I) change withgrain size reflected the influence of grain effect on landscape pattern, to benefit to the choice ofbest scale, transform the results of different scales, and understand change rule of landscapepattern and process. Waterbody and salina with small MPS and scattered distribution wereincorporated to larger landscape types around and disappeared. The grain effect of landscapeindex showed that NP, AWMSI and AWMPFD decreased, MPS of various landscape types increased, and LPI increased but some fluctantly with grain increasing. Moran's I analysisshowed that landscape patches within grain size of 30～180mm distributed collectively, and thatof 210～300mm distributed dispersedly.
|
PDF Full Text Request