Spatial And Temporal Patterns Of Spartina Alterniflora Invasion In The Yellow River Delta Based On GEE And Sentinel Time-series Data

Biological invasion is a serious threat to global biodiversity and has developed into one of the global ecological problems due to its direct and indirect impacts on native biomes and ecosystem functions and services.Invasive plants can have dramatic effects on biotic and abiotic features of ecosystems,and can also encroach on native vegetation causing a reduction in habitat and foraging areas for many organisms such as mammals and birds.Spartina alterniflora（S.alterniflora）was introduced in China mainly for wind and sand control and water purification,but with the continuous expansion of S.alterniflora,it has become an important coastal invasive plant in China.Therefore,the analysis of the spatial distribution and landscape ecological structure of S.alterniflora is important for the effective prevention and control of S.alterniflora.In this study,Google Earth Engine was used as the running platform to analyze the spatial and temporal distribution and landscape pattern evolution of S.alterniflora in the Yellow River Delta based on Sentinel-1/2 multi-source remote sensing images from2016 to 2021.Firstly,the Sentinel-1/2 dense time series images were constructed based on the GEE platform and three SAR-derived indexes and seven spectral indexes were calculated.Secondly,the two-term Fourier function was selected to fit the seven spectral parameters by comparing the fitting effects of three different orders of Fourier functions,and a total of 35 spectral features were obtained in five categories including max-value,min-value,max-day,min-day and time interval.Similarly,a total of 19annual mean features of SAR and spectra are obtained by the annual mean calculation code of GEE platform,and a feature vector collection image containing 54 features is obtained by band fusion.Then,the complete Simple Non-Iterative Clustering（SNIC）segmentation algorithm encapsulated in the GEE platform was applied to segment the feature images,followed by the random forest algorithm to classify the S.alterniflora and obtain the spatial distribution information of S.alterniflora in the Yellow River Delta from 2016 to 2021.And the spatial and temporal changes of S.alterniflora and the evolution of landscape pattern were analyzed.Finally,a quantitative analysis of the spatial distribution drivers of S.alterniflora was conducted by integrating natural and anthropogenic factors,and guiding suggestions for S.alterniflora management were provided accordingly.The main conclusions of this paper are as follows:（1）Feature extraction of S.alterniflora based on dense time series remote sensing imagesFrom 2016 to 2021,a total of 824 scenes of Sentinel-1 images and 945 scenes of Sentinel-2 images were collected in the Yellow River Delta to construct a year-by-year10 m high spatial resolution image collection.Then,the SAR-derived parameters（SAR＿S,SAR＿D,and SAR＿N）and spectral parameters(NDVI,NDWI,m NDWI,EVI,LSWI,AWEI_sh,and AWEI_nsh)of each scene were calculated separately based on the GEE platform,and 19 annual mean features including the underlying bands were obtained by using the annual mean calculation algorithm.In addition,the two-term Fourier function was used to fit seven spectral indexes to obtain a total of 35 time-series parameters（include 10 phenological parameters）.Based on the feature vector set,we found that S.alterniflora has unique phenology features,among which the phenology features generated based on NDVI,EVI and m NDWI can significantly improve the extraction accuracy of S.alterniflora.The addition of annual mean features is also beneficial to the extraction of information of S.alterniflora.Finally,the accuracy of annual mean user of S.alterniflora reached 0.92,and the classification effect was good.（2）Construction and accuracy evaluation of classification model based on remote sensing imagesBased on the results of fieldwork and literature,a wetland classification system was constructed for the Yellow River Delta.The classification system mainly aims to extract the information of S.alterniflora and includes wetland vegetation types such as S.alterniflora,P.australis,S.salsa,and T.chinensis.Then the decoding flags were established to lay a solid foundation for subsequent classification.Subsequently,a classification method combining object-oriented and random forest was used to extract the S.alterniflora information based on the feature vector set to obtain the spatial distribution information of S.alterniflora year by year from 2016 to 2021,respectively.Finally,the accuracy validation based on the confusion matrix was performed using validation samples and classification results.The accuracy evaluation results show that the establishment of a feature vector set based on dense time series acquisition can be effective for S.alterniflora information extraction.The accuracy of producers of S.alterniflora all exceeded 93.58%.（3）Changes in spatial-temporal distribution of S.alterniflora from 2016 to 2021The area of S.alterniflora in the Yellow River Delta was 44.97 km²,48.00 km²,48.27 km²,54.76 km²,55.54 km² and 58.00 km² from 2016 to 2021,respectively,which were mainly distributed spatially on both sides of the Yellow River Estuary and gradually showed a trend of extending toward the sea.In addition,new patches of S.alterniflora appeared for the first time on the south side of the Yellow River Delta in2018,and gradually formed stable communities in 2021.From the perspective of landscape ecology,the ecological advantages of S.alterniflora landscape gradually emerged,and the landscape structure tended to be complex and connectivity gradually increased.By quantitatively analyzing the influence of natural factors on the spatial distribution of S.alterniflora,it was found that temperature and seawater depth were the main factors affecting the spatial distribution of S.alterniflora.Human activities are also the main drivers affecting the spatial distribution of S.alterniflora.Among them,the governance of S.alterniflora was the main factor that led to the shrinkage of S.alterniflora.Based on the spatial distribution drivers and practical management experience,we found that the treatment method of"cutting+tilling+flooding"can achieve the optimal treatment effect with the least environmental disturbance.