| The extremely arid desert area is an important ecological barrier for the oasis.Study on this area is of great significance for protecting the oasis environment and maintaining regional ecological security.Vegetation is a natural link of soil,atmosphere and water.It plays an important role in the study of environmental and global change.Understanding vegetation distribution patterns and its formation mechanism is the basis of protection and utilization of regional plant resources,restoration and reconstruction of degraded vegetation.The NDVI index,which is a measure of the sensitivity of surface vegetation cover and vegetation growth,is recognized as the best indicator for quantitatively characterizing vegetation growth and vegetation coverage.Although domestic and foreign scholars have proposed a variety of vegetation index,many scholars further verified that the NDVI index is still considered to be a suitable index for the extraction of desert vegetation information.At present,there are few reports on the formation mechanism of vegetation patterns in desert areas outside the oasis.Thus,this study selects the desert region outside the Luntai oasis in Dina River Basin in Xinjiang as the study area,uses remote sensing,GIS,UAV,model simulation,etc.,to study the temporal and spatial patterns of vegetation and its formation mechanism.The spatial distribution pattern of vegetation is identified.The driving factors of regional vegetation pattern were analyzed from the aspects of soil,topography,climate and human interference.The simulation model is established to reveal the formation mechanism of vegetation patterns based on GRU neural network model.The research results include the following aspects:1)Spatial and temporal characteristics of vegetation pattern in desert :(1)The spatial distribution characteristics of vegetation pattern mainly include that more than 80% of the NDVI value in the study area is less than 0.2.The regions with NDVI value more than 0.2 are mainly distributed in the north of the Gobi desert region,the Non-Gobi desert region close to the oasis,and the southern Non-Gobi desert region.(2)The temporal characteristics of vegetation distribution in desert areas mainly include that: from 1992 to 2018,the average NDVI value of vegetation in desert areas fluctuated between 0.1 and 0.2.There was a significant decline from 1992 to 1998,and remained basically stable from 1998 to 2018.2)Soil type,soil moisture content and soil nutrient characteristics and their effects on vegetation pattern:(1)There are obvious differences in the influence of different soil types on vegetation pattern.The average NDVI value of desert aeolian soil is higher(mainly distributed in the Gobi desert region near the oasis),followed by calcareous meadow soil and gley meadow soil subclass(mainly distributed in the south of the Non-Gobi desert region),and the average NDVI value of brown desert soil subclass(mainly distributed in most areas of the Gobi desert region)is the lowest.(2)Soil Alkali-hydrolyzed Nitrogen(AN),Available Phosphorus(AP)and Available Potassium(AK)are negatively correlated with vegetation pattern.The relationship between soil Total Nitrogen(TN),soil total phosphorus(TP),total potassium(TK),Organic Matter(OM),p H and vegetation is not obvious.The relationship between soil salinity(TS)and vegetation is not obvious,but in Non-Gobi desert areas,soil salinity is negatively correlated with NDVI value.(3)The interaction of soil water and salt had a significant effect on the vegetation pattern,and the soil moisture content in the non-saline soil area and the weak saline soil area are positively correlated with NDVI value.(4)The soil moisture content in the Gobi desert region is obviously lower than that in the Non-Gobi desert region.With the increase of soil depth,the content of OM,AN,AP,AK,p H value and TS in the region with NDVI>0.1 in soil show a decreasing trend.The contents of AN,AP and AK in the Gobi desert region are generally higher than those in the Non-Gobi desert region.The value of soil p H in the Gobi desert region was slightly higher than that in the Non-Gobi desert region in the 0-40 cm soil layer,and slightly lower than that in the Non-Gobi desert region in the 40-100 cm soil layer.3)Effects of elevation gradient and microtopography on vegetation pattern:(1)The NDVI value increases with the increase of elevation in Gobi desert.The No-Gobi desert in the southern is flat;the NDVI value changes there are irregular.(2)In Gobi desert,the regional micro-topography has a great influence on the vegetation pattern.In the No-Gobi desert area the vegetation pattern is less affected by the micro-topography.4)Effects of climate and human disturbance on vegetation patterns:(1)During the period from 1992 to 2018,the annual rainfall and annual average temperature in the study area do not change significantly,but 2007 was significantly different from other years.The annual rainfall in the Gobi desert region was higher than that in the Non-Gobi Desert region,and the annual average temperature in Gobi desert region was lower than that in the Non-Gobi Desert region.There was a positive correlation between annual rainfall and NDVI value,and a negative correlation between annual temperature and NDVI value.(2)The NDVI value generally decreases with the increase of distance to oasis.5)Vegetation dynamic simulation:(1)The vegetation pattern in the study area is formed under the joint action of many influencing factors.The importance of influencing factors was as follows: soil type 45.69%,previous vegetation pattern 30.29%,relative position 5.42%,temperature 3.97%,precipitation 3.29%,soil AK content 2.17%,soil AN content 1.93%,soil AP content 1.64%,soil water content 1.56%,distance to oasis 1.40%,soil TS content 1.35%,terrain 1.30%.(2)The Gated Recurrent Unit(GRU)model has an obvious advantage in simulating vegetation patterns with dynamic driving factors.GRU model is selected for vegetation dynamic simulation.In order to further improve the simulation accuracy and adapt to the task of vegetation dynamic simulation driven by complex and multi-factors in this study,the GRU model was improved,and the improved model called VDM could process dynamic data and static data separately.VDM model was used to simulate the vegetation pattern in the study area,and the simulation results of VDM are better than that of GRU model.Compared with the true value,the overall accuracy of VDM reached to 81.97%,indicating that VDM model is suitable for simulating the vegetation dynamic drived by multi-factor.(3)The study comprehensively considered the regional soil,terrain,climate and other factors,and simulated the vegetation pattern based on VDM with an accuracy of more than 80%,indicating that more than 80% of the mechanism affecting the vegetation pattern distribution was explained.This paper studies the temporal and spatial pattern characteristics of vegetation in the desert area around Luntai oasis in Dina River Basin of Xinjiang,deeply analyzes the influences of soil,terrain,climate and human disturbance(Oasis expansion)on the vegetation pattern in the study area.Based on the GRU model,a VDM model is established to simulate the vegetation pattern in the study area.The results show that the vegetation pattern in the study area is mainly affected by soil factors(including soil type,soil nutrient content,soil moisture content,etc.),terrain,climate factors and human interference,and the influence degree of different factors on the vegetation pattern in the study area is different.Taking the above factors into consideration and simulating the vegetation pattern based on VDM,more than 80%of the formation mechanism of vegetation pattern is explained.This study reveals the spatial and temporal characteristics and driving mechanism of vegetation change in desert areas,which is of great significance for the protection or restoration of vegetation in the study areas. |
PDF Full Text Request