Multi-scale Spatiotemporal Variation of LULC and Its Relationship to Water Quality Based on High Resolution Remote Sensing Image in Fuxian Lake Watershed
Author:Li Shi Hua
Supervisor:wang jin liang
Spatial-temporal evolution of Land use and land cover(LULC)and its multiscale impacts on water environment,being of great theoretical and practical significance for optimizing land use,improving water environment quality and sustainable development of eco-environment in the basin,have become a hot issue in LULC research field.In this research,Fuxian Lake which is one of the typical plateau lakes in Yunnan Province was taken as the research area.The high resolution remote sensing images,basic geographic information and monitoring data of water quality were taken as the main data sources.Remote sensing and Geometric Information Technology were applied.A multi-scale object random forest algorithm(MSORF)is proposed to raise the classification accuracy of high resolution remote sensing images in the basin.By analyzing the multi-scale spatio-temporal evolution process and driving force of LULC in the basin,the scale dependence of water quality on the types and patterns of LULC in different pollution source areas was revealed,and the spatio-temporal relationship model of water quality and the types & patterns of LULC was built.The characteristic scales of interactions between the types & patterns of LULC and water quality in the watershed were also identified.Based on the researches above,strategies and advice on water environment protection were put forward.The main contents and conclusions of the dissertation are as follows.1.The Basin’s LULC Spatio-temporal Evolution Characteristics.(1)LULC Information Extraction.Under the guidance of Banded Support Vector Machine(BSVM),Optimized Support Vector Machine(OSVM),Pixel-wise Support Vector Machine(PSVM)and MSORF methodologies,classification experiments and comparative researches on high resolution remote sensing images in Fuxian Lake Basin in 2005,2008,2001,2014 and 2017 were carried on.The results indicates that the MSORF proposed in this paper has higher classification accuracy than PSVM,BSVM and OSVM.The accuracy can be improved by 4% to 9%.The land types with small heterogeneity(grassland,water area and exposed surface)have higher classification accuracy than those with large heterogeneity(buildings,structures and roads).(2)The Basin’s LULC Type Changes.Water,woodland,arable land and grassland are the four main types of surface cover,The total area of these four types is 95.59%,95.00%,93.89%,93.43% and 92.50% in 2005,2008,2011,2014,2017 respectively,and housing construction area,roads,gardens,structures,artificial excavation land,desert and bare surface 4.41%,5.00%,6.11%,6.57% and 7.50%.The results indicate the area of artificial surface types has increased over years,while that of the natural surface cover types is opposite.The area of artificial excavation land in 2017 is five times of that in 2005.The area of water only decreased 0.08% from 2005 to 2017.From the perspective of income and expenditure of LULC changes,the change of cultivated land,forest land and grassland is mainly reflected in the decrease of quantity,while the change of desert and bare surface,road,building area and water area is mainly represented by the transfer of spatial location.The driving forces analysis of the LULC changes shows that the slope,elevation and distance to the nearst road are the main driving factors for the LULC changes in the basin.(3)The Basin’s LULC Dynamic Degree Changes.There are significant differences in spatio-temporal scales of LULC change in the basin.The results of single land use dynamics and comprehensive land use type dynamics analysis show that there are differences in land use dynamics at different spatial-temporal scales.The comprehensive land use dynamics of the whole basin scale from 2005 to 2017 is the largest,while the riparian buffers are from 2015 to 2017,and the difference is obvious.In general,The dynamics of comprehensive land use in each sub-basin show obvious differences,and the dynamics of the sub-basin scale are higher than the whole basin scale,and land use gradually changes from extremely slow to slow.(4)The Basin’s Spatio-temporal Evolution Process and Expression.1)Based on the principle of change trajectory,in the JQuery and ECharts environment,Java Script is used as the development language to realize the dynamic visualization of LULC change trajectory network.Based on the variation trajectory analysis method,which can effectively reveal the transfer path of each class in the research period,and clarify the transfer mode between LULC types,and solve the difficulty of expressing the problem of LULC change process of more than two monitoring phase sequences.2)Constructing the spatio-temporal analysis algorithm of LULC change based on patches,and integrating timeline animation,multi-window and entity backtracking three methods comprehensively describes spatio-temporal evolution process of LULC patches,which can overcome the defects of grid model for expressing the spatiotemporal process of LULC.(5)The Basin’s LULC Pattern Changes.The analysis of the selected 8 landscape pattern indexes shows: the watershed landscape type is based on forest land as matrix.The total number of patches is increasing,which is demonstrated that the human intervention on the surface of the earth is increasingly strengthened,and the fragmentation of the landscape is intensified.The difference in patch area was weakened,the shape of patches was further complicated,the natural connectivity of patch was improved.(6)The Basin’s Land Use Intensity Change.With the increase of spatial scale,the spatial autocorrelation of land use intensity in the basin drops,which shows that with the increase of scale,the difference of adjacent land use intensity strengthens and the similarity decreases.The land use intensity of the basin is generally strong,but there is significant differences in different regions.As the north bank of Fuxian Lake is mainly the seat of Chengjiang County,the terrain is relatively flat and the level of urbanization is high.However,with the development of tourism in the south bank,the intensity of development is high,and so is the land use.However,the border of the river basin is mostly covered by forest land and in the forbidden zone,the land use intensity shows low-low aggregation.From the perspective of sub-basin scale,the land use intensity of Dongdahe River Basin,Maliao River Basin,Daicun River Basin,Luju River Basin and Niumohe River Basin shows high-high aggregation,while that of other watersheds is low,showing the distribution of high-low and low-high aggregation types.2.Relationship between Multi-scale LULC Changes and Water Quality.(1)Spatio-temporal differentiation characteristics of water quality in the watershed.1)The quality of water entering the lake is generally poor,while the quality of water in the lake centre is still type I.The annual average water quality of the rivers entering the lake is inferior to the V category.Chemical Oxygen Demand(COD)in 45 sections had the highest proportion of type I and type II,reaching 53.33%,the other types of Total Phosphorus(TP)were basically evenly distributed,accounting for 22-26%.Total Nitrogen(TN)exceeded the standard seriously,nearly 90% of the water quality is inferior V.2)The spatial variation of water quality of rivers entering the lake is significant.The urban watershed has a higher COD value,the phosphate mine watershed has a higher TP value,and the mountain watershed has a higher TN value.The changes of COD,TN and TP generally showed a growing trend in each sub-basin,and the variation was significant.(2)Multi-scale relationship between LULC types and patterns and water quality.In the analysis of the relationship between LULC type and pattern and water quality,the spatial heterogeneity of LULC and pollution sources in the basin is fully considered to affect the water quality indicators,and the multi-scale relationship analysis between LULC and water quality in different pollution sources within the basin is strengthened.It enriches the research cases of uncertainty and multi-scale relationship between LULC changes and water quality indicators.(1)There is a significant difference in the interpretation of relationship between water quality and LULC types and patterns.The area percentage of ISA and artificial excavation land,desert and bare surface plays a promoting role in each spatio-temporal scales,while the planting land and forest grass cover plays an inhibitory role.Forest cover can improve the water quality as expected.There is no obvious correlation between the increase or decrease of planting land and water quality.The impact of planting land on water quality is complex and negatively correlated with TN and TP.This indicates that the area percentage of agricultural land is not the main factor affecting TN,which is related to the application amount of pesticides and fertilizers,planting structure and distance from receiving water body and topography.(2)Sub-basin and riparian buffer scale,landscape diversity index SDI(Shannon’s Diversity Index)and SEI(Shannon’s Evenness Index),landscape fragmentation index PSSD(Patch Size Standard Deviation),MSI(Mean Shape Index)are important factors affecting water quality.Although the other landscape pattern indices ED(Edge Density),COHES(Patch Cohesion Index)and FRAC-AM(AreaWeighted Mean Fractal Dimension Index)have significant correlation with water quality,the impact of landscape index on water quality is significant at different spatial scales,and there are certain uncertainties,such as the relationship between landscape index NUMP(Number of Patch)and water quality can not be reasonably explained.2)The multi-scale relationship between LULC types and patterns and water quality is significant.(1)The 300 m riparian buffer zone in urban watersheds is the strongest spatial scale(feature scale)of LULC type effects on COD.This is consistent with the results of other plateau lake studies,and the universality of this conclusion needs further verification.(2)There are similarities between effect LULC pattern having on water quality and effect LULC type having on water quality index: urban LULC type and pattern have the greatest impact on COD,phosphate mine area has the greatest impact on TP,and village farmland has the greatest impact on TN.The difference is as follows: urban and phosphate mines have a higher interpretation of COD and TP in the buffer zone of 300 m,500m and 700 m,while in the village farmland,the sub-basin scale has the highest interpretation of water quality.The influence of LULC pattern on water quality indicators in different regions is as follows: in urban areas and phosphate mines,COD and TP are better explained in sub-basin;in village farmland,TN can be better explained in 700 m buffer zone.And TP has a higher interpretation in the buffer 300 m.Compared with the riparian buffer zone,the landscape pattern index at the sub-basin scale has a high degree of interpretation of water quality.3)Measures and suggestions on regulating water quality in Fuxian Lake and its watershed under the evolution of LULC.The result of simulation of LULC changes in the basin in 2020 using the CLUE-S model shows: the natural change scenario is bound to add a serious threat to the water quality of the rivers entering the lake,and further aggravating the non-point source pollution.While the ecological protection scenario is conducive to improving the water quality of the rivers entering the lake and effectively to improve the water environment in the basin.According to the characteristics of different regions in the watershed,the corresponding land use types and patterns,scale optimization of land use patterns,and regional ecological restoration strategies were proposed.