With the development of ocean monitoring technologies, the people are getting more and more ocean data. Because the ocean flow is the main channel for material and energy transportation in the whole ocean system, it is of great significance for all sea related fields to analyze the migration law and visualize the change of the ocean flow field based on these ocean data. However, due to the characteristics of multi-sources and heterogeneity of these ocean space-time large datasets, there are still no satisfactory models and tools for spatio-temporal data analysis and dynamic visualization of ocean flow field. It is also very hard to organize and share these heterogeneous data at an unified standard. For this reason, in order to realize the unified description of ocean flow field at semantic levels, based on the analysis of multi-dimensional, topological and indefinite characteristics of the ocean flow, and based on the idea of ontological semantic analysis, this paper proposed and constructed an ontology organization infrastructure which was composed of four tuples, O=(C,P,R,I), in which C is the collection of concepts, P is the collection of attributes, R is the collection of relationships between concepts, and I is the collection of instances. By combining the definitions of concepts, attributes, and relational structures, the paper constructed the whole ontology structure of ocean flow field. In order to give a clear ontology semantic modeling procedure, the paper took the local ontology construction of ocean current phenomena as an example, gave the definition of concepts, ontology attributes and semantic relationships of ocean flow filed based on the analysis of ontological attributes such as causes, spatiality, temporality, and mobility, providing a specification framework for the organization of ocean flow data. Taking the geostrophic current as the example, with the advantage of Web Ontology Language (OWL) on modeling and constructing ontology, the paper gave the formal expression and constructed the related ontology classes of the geostrophic current. The research showed that the ontology based ocean flow field semantic analysis method could effectively solve the heterogeneous issues in the traditional marine phenomenon description, and could provide certain references for the study of marine information sharing and integration.
The social survey questionnaires sometimes are multi-objective, and some objectives are attribute data or category variables. However, traditional spatial sampling theory is primarily used in single-target and non-attribute data. It is not suitable for the investigation required multi-type target objects. A new method based on variability model was proposed in this paper. The different types of variables can be measured by variability model on the spatial variability and used as the basis for spatial stratification to design sampling plan. Depending on the questionnaire about residents' daily travel energy consumption of Xiamen Island and historical data in this study, we calculated the values of variability of samples by the model and get the map of spatial distribution. Contrasted with the map of hierarchical combination of the integrated factors and the map of stratified sampling by experts,it got the value of variability through the pre-investigation, ultimately obtained program of sampling point distribution about the target settlement of Xiamen Island. The results showed that: (1) Contrast to experts layer, the main component layer and combination of factors layer from the perspective of variability values, combination of factors layered approach is more reasonable. This method reflects various factors that affect sampling in spatial distribution plan, which offers solution for the survey involving multiple data category and expands the application scope of “Sandwiches” model. (2) The number and distribution of samples are affected by variance in the sandwiches spatial Sampling. But they are not increased with the increase of variance, the number and the distribution of samples are affected by many factors, the size of the geographical space is one factor. (3) Variability model quantified various types of data about Sampling objectives successfully. In our study, we got more detail sampling plan based on pre-investigation in small range. The sampling accuracy is 0.0002while sample size is 35. It meets the practical requirements of the survey of Xiamen. The number of samples and the accuracy are controlled in the reasonable range. Not only we saved manpower and material resources, but also, we improved the accuracy of sampling.
Precipitation data with high accuracy and high spatial resolution are very important for improving our understanding of basin-scale hydrology, agriculture and earth science, and are essential in characterizing the behavior of a catchment. Attaining accurate and high spatial resolution precipitation data is deemed necessary for environmental, meteorological, and hydrological applications. This study proposed a statistical downscaling method based on the geographical weighted regression method (GWR) and high accuracy surface modeling method (HASM) by selecting the optimal downscaling scale and considering the errors produced in the scale-change process. GWR can address the spatially heterogeneous relationships between precipitation and its influence factors, such as digital elevation model (DEM), normalized difference vegetation index (NDVI) and slope, at different spatial resolutions, whereas HASM is used to merge the cross-scale error fields that are produced from the downscaling process and meteorological observations.The method was used to downscale the TRMM precipitation dataset over the Heihe River basin (HRB) from 0.25o to 1 km. Cross-validation method was used to validate the developed method combined with the station observations. Results showed that the proposed downscaling method performed better than the traditional downscaling method, which directly downscaled the TRMM products without considering the optimal downscaling scale. Besides, it was found that residual correction is necessary after the GWR-based downscaling method. The method proposed in this research can be used to downscale precipitation dataset with coarse resolution and could be applied to the areas with data-scare network and complex topography.
While population maps are important tools for people to perceive the regularities of population distribution, different scales of population maps cause map readers' cognitive difference in the regularities of spatial distribution of population. In this paper, eye movement parameters such as number of fixations, fixation duration and number of correct answers were selected in the population map cognitive experiment by eye movement tracking to test the significance of the difference, and the results were analyzed from the perspective of spatial differentiation regularities. By exploring the influence of different scales including province and county (city) on map readers' cognition of the distribution regularities of population, it is concluded that different scales of population maps have a significant impact on readers' perception based on the significant difference analysis. When perceiving the characteristics of spatial distribution of population and the population quantity, more details and information are provided by county (city)-scale population maps, which is beneficial to readers' understanding of the spatial differentiation regularities of population, with less average number of fixations, shorter average fixation duration, more correct number of answers for each question and higher cognitive efficiency. The impact of scale on the cognition of the population spatial distribution and the population size was discussed. The acquired cognitive rules of the scale can be used in designing the demographic maps and shortening readers' cognition time, which is convenient for readers to extract valid information from the demographic maps, thus to improve the map usability. Besides, through the analysis of eye movement parameters like the fixations points, fixation time and number of correct answers, as well as the significance test, quantitative researches of the scale effects on the population distribution were carried out. The perspective drawing of the fixations hotspot can be used to visualize the cognitive spatial differentiation of the readers. And the results are no longer limited to the simple qualitative expression, which is of great significance for the use of different scales of demographic maps to express population distribution characteristics and regularities. In addition to adopting the hierarchical mapping method to draw the population maps, this thesis also has conducted experiments on the readers' cognition of the spatial distribution regularities of population with different population density maps at different scales. Since it can reflect the population distribution more precisely and more visually, the results of this research may be further improved. And in the further work, the above population map needs further studying.
Real-time and accurate short-term traffic flow prediction, a critical technical problem in traffic control and guidance which is challenging and needs to be solved urgently in related research fields and engineering practice, still remains because of the hardship caused by the uncertainty and the temporal variability in traffic flow datasets acquired in different times. In order to improve the performance of the short-term traffic flow prediction, a new method based on similar data aggregation techniques and a modified KNN algorithm with varying K-value (KNN-SDA) was proposed and the related algorithm was also implemented and tested on actual measured datasets in this paper. Firstly state vectors were generated from the preprocessed traffic flow datasets by calculating the optimal time delay with the help of the mutual information theory. Each of our state vectors is composed of two parts, the first one of which is a regular state vector and the second one of which is a modified state vector which makes a contribution to a higher similarity between our state vectors and those in training datasets. Subsequently a historical traffic flow database of temporal series was constructed on the basis of results mentioned above for further experiments. After that, the proposed similar data aggregation techniques were applied to aggregate and clean data to obtain 144 training data sets in different times from historical traffic flow database, which would effectively improve the prediction accuracy and efficiency of the proposed algorithm. At last, the optimal K-values, each of which corresponded to a moment, were determined through the cross validation method. So far, the overall process of the KNN-SDA algorithm with varying K-value has been completed. In order to verify the performance of the proposed method, we compared the experimental results derived from our method with those from three other ones. It turns out that the KNN-SDA algorithm with varying K-value proposed in this article can improve the prediction accuracy significantly and ensure high execution efficiency as well.
China has a vast area and frequent disasters. Due to the special geographical environment and the imperfect disaster prevention and reduction system, the rural areas have made it more difficult for people to take refuge. Refugees can use effective evacuation guidance to reach the shelters in the shortest time, which will reduce unnecessary casualties and improve the efficiency of rescue evacuation. Therefore, the layout of emergency shelters should consider the timeliness. The emergency shelter accessibility refers to the accessibility between emergency shelters and refugees. That is to evaluate, when disasters occur, how difficult it is for refugees to reach emergency shelters through evacuation routes. The emergency shelter accessibility is an important measurement for its rationality in spatial layout. The Gaussian based two-step floating catchment area method(Ga2SFCA) fully considers the interaction between the demand points and the facilities, it also considers the fading relationship between attraction and distance of facilities. The network analysis is based on the actual road and the results obtained are real and objective. Combining the two methods can reduce the accessibility error caused by neglecting the interaction and actual distance between supply and demand in traditional research and effectively determine the spatial distribution of the emergency shelters. Taken Songbai town in Shennongjia as an example, this study combined the Ga2SFCA with the network analysis and took multiple evacuation times as catchment sizes to analyze the rural emergency shelter accessibility. Finally, from the perspective of the evacuation demand of refuges, we analyzed the accessibility of vulnerable people to shelters in the study area. The results show that: this method can be applied to the research on accessibility of emergency shelters in rural area, and by GIS, the differences in accessibility space distribution of emergency shelters in the study area can be revealed directly. At the same time, this method can also provide support for the formulation of scientific rural disaster prevention and mitigation planning. When planning an emergency shelter, not only the accessibility of the sites but also the scale of the facilities and the configuration of the emergency infrastructure must be fully considered. Meanwhile, the emergency marking system should play a guiding role to enhance the emergency shelter accessibility from the refugees' behavior and psychology.
The living environment of trees is complex and changeable, resulting in different morphologies of trees. In order to depict the diversity and complexity of tree morphology, a simulation method of tree growth based on model decomposition was proposed. Firstly, we built diversified 3D tree models. The actual data of trees were measured, and used to fit tree trunk and crown shape curves based on the B spline function. Using the taper equation, the trend of trunk diameter was simulated. Meanwhile Direct3D API was used to realize the 3D tree morphology modeling. Secondly, the tree model was decomposed into 9 sub-models on the basic of the tree morphological characteristics. Finally, according to the topology relationship, sub-models were organized dynamically by linking scene nodes. Growth models of DBH, tree height, under branch height, crown height and crown width were used to calculate the tree morphological parameters by multithread parallel technology on multi-core processor, which represented the crown morphology in the cardinal directions at different age stages. And sub-models were controlled to realize the growth simulation of forest tree. The result shows that the method combines the 3D tree models with tree growth models closely. Meanwhile the under branch heights, crown heights, crown widths of trees in different directions grow accurately by growth models driving. The growth states in each direction of forest trees is simulated. The FPS of 3D rendering is always more than 25 and the average FPS is around 50 during trees growth. The direction heterogeneity of tree growth is simulated.
Based on the precipitation data of 34 meteorological stations in Guizhou Plateau from 1960 to 2016, the study analyzed the spatial-temporal distribution of precipitation and its barycenter shift in Guizhou Plateau at different time scales using the Co-Kriging method, Mann-Kendall trend test method, the climate tendency rate and barycenter model. The results indicate that: (1) The average annual precipitation over the Guizhou Plateau was characterized by decreasing from the south and the east to the northwest.There were three rainy centers in the south and east of Guizhou Plateau, respectively located in the northward transport path of the southwest warm and humid airflow (Xingyi-Anshun area), the windward slope of the Miaoling Mountains (Duyun-Dushan area) and the windward slope of the Wuling Mountains (Tongren- Songtao area); And the less rainy area was located in Weining and Bijie, where is the leeward slope of the Wumeng Mountains. (2) The chronological changes of precipitation from 1960 to 2016 presented volatility, the precipitation variability of 2010s was the largest, and the precipitation variability of 1990s was the smallest; the interannual variation was more dramatic, showing a not-significant decrease trend, which presented a spatial reduction in the central and western regions and increase in eastern regions. The differences in seasonal precipitation was obviously, precipitation decreased significantly in spring and autumn, and did not increase significantly in summer and winter.The change of precipitation in each month was different. The increase in precipitation was most significant in January and March, and the decline in precipitation in April was the most significant. (3) The precipitation barycenter was distributed in the southwest-northeast direction, and there had been a clear eastward shift in the past 57 years. The decrease of precipitation in Guizhou Plateau had the possibility of being associated with the weakening of the Southwest Monsoon. The above results are of great significance to the allocation of water resources and the prevention of flood disasters in Guizhou.
The temporal-spatial distribution and influencing factors of mountain-flood disaster are a key issue in the disaster data mining. Using the historical mountain-flood catastrophe data that is learned from the National Mountain Flood Disaster Investigation Project from 1950 to 2015 in the Sichuan Province, and employing the methods in geo-statistics, geographic detector and geo-spatial analysis, this paper systematically analyzed the temporal-spatial distribution of historical mountain-flood disaster and the influencing factors in Sichuan Province. The main findings are the following : (1) The total amount of mountain-flood disasters in the Sichuan Province, from 1950 to 2015, remained stable and then increased rapidly. In addition, the catastrophe mainly occurred from May to September, especially in July every year. (2) The frequency of county disasters over Sichuan showed a decreasing trend from south to north. The average rainfall during historical mountain-flood disaster (ARD) increased exponentially from east to west, and decreased from middle to north. (3) From May to September each year and from1950s to 2010s, the center of gravity and the elliptical center of each standard deviation of the accumulated mountain-flood disaster are concentrated in the central part of Sichuan, moving to the northeast. The accumulated disaster points emerged in a pattern of southwest-northeast. (4) The spatial autocorrelation analysis indicates a positive spatial correlation between the amount of mountain-flood disaster and ARD in county area. (5) The geographic detector analysis indicates that natural factors, rainfall, human activity and other factors have a great influence on the temporal-spatial distribution of mountain-flood disaster. In particular, the main driving factors are the rainfall index, standard deviation of elevation and slope. The results provide a theoretical basis, scientific and technological support for the investigation of the temporal-spatial distribution characteristics of mountain-flood disaster in the Sichuan Province, which can also benefit the monitoring and early warning, the risk assessment, the prevention and control of mountain-flood disaster in small watersheds.
As the North-South boundary of China and the watershed between Yellow River and Yangtze River, the Qinling-Daba Mountains are characterized with complicated and transitional variations in mountain altitudinal belts. For a longtime, latitude, longitude and mountain mass effect (MEE) have been considered to affect the vertical distribution of montane vegetation or altitudinal belts. The two former factors reflect the zonality of the geological distribution of vegetation and have receiving much attention, but MEE, an important factor for montane vegetation variation, is often overlooked because of the lack of effective quantitative methods for it in the Qinling-Daba Mountains. Quantification of MEE contributes to understanding azonal factors and the transitional characteristics between warm-temperate and north subtropical zone in China. Mountain base elevation (MBE), as a topographic factor, related to MEE closely, and recognition and quantification of it is indispensable for quantifying MEE. This study aims at quantifying MBE based on STRM-1 data with 30m resolution in the Qinling-Daba Mountains. We used two methods (based on mountain characteristics and based on drainage basin division) to determine mountain base subareas and then used relief amplitude, the average altitude in each subareas and slope to calculate the MBE. The results show that: (1) There are 93 subareas based on mountain characteristics method and 209 subareas based on drainage basin division method in the Qinling-Daba Mountains. The two MBEs have similar values and distributions and reflect the characteristics of the terrain of the study areas. (2) From east to west, MBE presents a gradual upward trend. (3) Along latitudinal direction, MBE presents a tendency of increasing from the Hanjiang Valley to the Qinling ridge and the Daba ridge. (4) The MBEs vary greatly in different exposures, from 1000m to 1809m in the southern flank of the Qinling, but only 850~1300 m in the northern flank. The MBE extracting methods will provide important technical support for the quantitative research of MEE in the Qinling-Daba Mountains.
Urban population distribution and activities are always the hot research topics. Identifying the spatial-temporal variation and predicting future trends are of great significance for estimating population accurately, making policy effectively, and warning of population booming timely. With the availability of data and the development of data processing technique, multisource data with both spatial and temporal features, such as mobile signaling data, have been used in population studies. In this paper, q-statistic was firstly applied as an exploratory analysis, then Bayesian spatial-temporal models were used to evaluate patterns of urban population and make prediction of future trends. The Chaoyang, Beijing in 2017 was selected as empirical study of this model. The spatially stratified heterogeneity was detected by q-statistic in Geodetector firstly. Then we explored the overall spatial variation, overall time trend and the departures of the local trends from the overall trend of resident population in Chaoyang by use of Bayesian spatial-temporal hierarchical model. Secondly, we applied Bayesian Gaussian predictive process to predict the resident population in December of 2017 by incorporating other relevant influential factors. The results show the perfect spatial stratified heterogeneity for resident population in Chaoyang, and the overall spatial variation demonstrates an increasing trend of population from center to the outside along the main ring road in Beijing. The overall time trend is still growing all over Chaoyang district, while the local trends, which departure from the overall trend of resident population, are different between each sub-districts in Chaoyang. Moreover, the spatial distribution of predicted resident population shows a high consistency with the observed resident population, and the prediction accuracy can be well accepted on the scale of Chaoyang district. However, prediction accuracy shows obvious difference on scale of sub-districts, with the worst prediction accuracy in the capital airport area. These findings show that Bayesian hierarchical model and Bayesian Gaussian predictive process are reliable in empirical study of population evaluation and prediction by effective application of multisource spatial-temporal data. Researches in this paper can be an excellent theoretical and practical support for mining multisource spatial-temporal data and assisting multiscale analysis with Bayesian spatial-temporal model, and provide an important basis for population controlling and early warning in urban population management.
Land use mapping using remote sensing techniques supplies essential datasets for scientific researches including global climate change, regional sustainable development and so on. The evaluation information on the accuracy of the land use mapping determines the integrity, reliability, usability, controllability and shareability of the land use maps obtained by the applications of remote sensing techniques. In this paper, the methods, processes and results of multiple temporal land use mapping for China's coastal zone using remote sensing techniques were overviewed, and the land use maps in 2010 and 2015 were selected for accuracy evaluation. The validation samples were collected based on Google Earth and the confusion matrices were established for the whole coastal zone and its sub-regions, respectively. Then, the overall accuracy and Kappa coefficient were calculated. Main findings are as follows: (1) Results of land use mapping in 2010 and 2015 using remote sensing techniques achieved high accuracy. For the entire coastal zone in China, the overall accuracy came to 95.15% and 93.98%, with the Kappa coefficients of 0.9357 and 0.9229 in 2010 and 2015, respectively. (2) The accuracy of land use mapping in China's coastal zone exhibited obvious regional differences. The best accuracy was found in the coastal area of Jiangsu province in 2010, and very high accuracy were found in the coastal area of Hebei-Tianjin, Shanghai city, Hainan province and Taiwan province in 2015, while the worst accuracy was found in the coastal area of Fujian province in both 2010 and 2015. (3) The accuracy of land use mapping in China's coastal zone exhibited obvious typological differences. The very high accuracy (both producer precision and user precision) were achieved for farmland, forest, grassland and saltwater wetlands, and the high accuracy for built-up, freshwater wetlands and human made saltwater wetland, while the worst accuracy for unused land. (4) The misclassification between cultivated land and forest land, construction land and grassland is quite significant. Inland water bodies were easily misclassified into cultivated land, forest land and construction land. Artificial salt water wetlands were easily misclassified into cultivated land and construction land, and unused land. It was easy to mistakenly classify the unused land as cultivated land. These are the issues that should be paid more attention during the continuous update of the land use maps in the future. This study provides supports for the dynamic monitoring and scientific researches on coastal land use changes.
With the continuous improvement of the spatial resolution of high-altitude platform remote sensing images, multi-scale information extraction methods have been used widely. The choice of optimal segmentation scale is a key technique in multi-scale remote sensing image segmentation. Aiming at the problem of segmentation quality assessment of multi-scale remote sensing image segmentation, a spectral and shape-based segmentation quality assessment method is proposed. Firstly, the image is initially segmented using the superpixel method, and the image is over-segmentated into several regions. Secondly, the multi-scale images are generated by iteratively merging the neighboring regions according to the merging criterion, and the scale-sets structure is used to index the regions of each scale. Adjacent graphs are used to record the relationship between the regions of each scale, then the formulas for the shape compactness and the shape smoothness of each scale are given by this paper, and the homogeneity and heterogeneity of each scale are calculated by combining the spectral features and the shape features of each scale; Finally, the optimal segmentation scale is automatically selected according to the Bayesian minimum risk criterion. The experimental results show that this method can adapt to the characteristics of regions in different images, and make the choice of optimal segmentation scale more reasonable and the image segmentation effect much better. The proposed algorithm selects the optimal segmentation scale based on standard of the global optimize, It is one of the development of multi-scale information extraction technology that how to make the object reach the best in every scale.
Object recognition of remote sensing image is of great theoretical significance and application value in many fields. Faster and more accurate object identification methods are hot and difficult points in the field of remote sensing and image. In this paper, the method of deep learning is applied to remote sensing image object recognition, and a fast and accurate method of object recognition based on Faster R-CNN deep learning network is proposed. This method uses the proposal region extraction method based on RPN and the VGG16 training convolution network model, and constructs a deep convolutional neural network for the object recognition of remote sensing image. In order to verify the accuracy and performance of the method, the GPU accelerated computing model was used in the Caffe deep learning framework. Firstly, the aircraft target recognition experiment in remote sensing image was designed. The aircraft target recognition accuracy reached 96.67%. Then, after the experiment was successful, we continued to identify other target features, and selected the high-resolution remote sensing image of the oil tank, playground and overpass object for verification experiments. In the same experiment environment, the same good experimental verification results were obtained, the target recognition rate was at a high level, and the cost time of recognition in each picture was less than 0.2 seconds, which fully verified the validity and reliability of the model studied in this paper. After analysis and comparison, the conclusion is that the deep learning method based on Faster R-CNN can realize the fast and accurate recognition of the selected targets, which proves that the method has a good promotion significance in high-resolution remote sensing image target recognition applications. Therefore, the model has great application value, and it also has certain reference significance for target recognition research based on other deep learning methods.
The advantages of Ningxia Hui Autonomous Region is land, light energy, and irrigation of Yellow River, which provides an inherent condition for the growth of characteristic crops of Ningxia (e.g. watermelon, Chinese wolfberry and jujube). Accessing to planting structure information of characteristic crops quicklf and accurately is not only an important basis for regional crop monitoring, yield estimation and disaster assessment, but also an important evidence for analyzing the spatial pattern changes of characteristic crop and assessing the impact of regional characteristics on agricultural production. In recent years, with the continuous development of space technology and remote sensing satellites, more and more scholars have applied remote sensing technology to the extraction of crop planting structure information. However, the traditional remote sensing survey model is only applicable to low resolution and medium resolution remote sensing data, and domestic and abroad scholars have relatively few studies on information extraction of similar Ningxia characteristic crops (e.g. watermelon, Chinese wolfberry and jujube), and the selection of classification models and strategies is difficult to meet the demand of rapid monitoring, accurate acquisition and real-time decision-making. Based on this, this paper calculate and analyze the spectral characteristics and texture features of the three types of specialty crops under the support of GF-2 remote sensing data, and establish an SVM of object-oriented classification model, the overall classification accuracy is 94.94% and the Kappa coefficient is 0.9174. And compare the classification results with the traditional SVM classification results, The study found that the SVM of object-oriented model established in this paper has the highest accuracy and best results, the texture information makes it easier to distinguish the Chinese wolfberry and jujube,whice has theless difference in spectral characteristics, and the texture information effectively reducing the model error and missing error, and improving the model classification results.
Due to seasonal vegetation dynamics and hydrological fluctuations, classification of wetland from remote sensing images is often more difficult. In this paper, a pretreated GF-2 image in the east of Tongyu Country, Baicheng City, Jilin Province, was analyzed by Random Forest with optimized feature space. The key method is divided into two steps. The first step is to perform multi-scale segmentation and extraction of object features in the remote sensing image of the study area. For a situation that some scholars obtain the best segmentation scale subjected to subjective factors, this paper obtains the best segmentation scale by improving the global optimal segmentation method, The second step is based on optimal segmentation, to optimize the feature space of the random forest classification algorithm on the basis of the importance of features to obtain the best random forest classification results, and then the classification results of the K-NN, SVM, and CART algorithms with the same data, the same segmentation scale, the same training sample and the same feature space, and the RF algorithm with unoptimized feature space are compared. The results show that the total classification accuracy and Kappa coefficient of the RF algorithm based on optimized feature space are 93.038% and 0.9177, respectively, while the total accuracy of the classification results of K-NN, SVM and CART are 83.357% and 78.068%, respectively, 77.136%, the total accuracy of the classification results of RF algorithm with unoptimized feature space is 90.937%. Compared with K-NN, SVM and CART classification algorithms, the RF algorithm has better classification performance in GF-2 wetland image data. At the same time, the accuracy of the RF algorithm with the optimized feature space has been improved, and it can play a very important role in wetland resource management.
Estuarine region is one of the most densely populated and prosperous area around the world, and it is also an eco-environmental vulnerable area which is more fragile to human activities. The acceleration of urbanization have inevitably resulted in a series of ecological and environmental problems on estuarine region, the thermal environment is a severe part of them. Higher temperatures and extreme heat not only hamper air quality but also increase energy consumption for cooling, threatening the health of urban residents. Based on the multi-source remote sensing images, characteristics of land surface temperature under the urbanization in Minjiang River estuary area were analyzed by using remote sensing techniques and statistical methods. With the help of Moran's I index, spatial clustering characteristic and scale effect of LST were examined. Further, the correlations between LST and different landscapes were found in quantitative analysis. The results showed that: (1) Built-up land area increased sharply from 1993 to 2013, showing a slow-rapid-steady and increasing process. A large number of large-scale edge-expansion was the primary growth type, meanwhile urban sprawl was mainly in east, west and south directions. (2) The area of sub-high and high temperature zone increased markedly, while the sub-low and middle temperature zone reduced; and there was no significant change in low temperature zone. Moreover, the spatial distribution of the high temperature region was consistent with built-up land expansion. (3) The LST exhibited an obvious disturbance characteristic; the temperature near city center presented dramatic changes and the temperature fluctuation in suburb was relatively smoother than Fuzhou city proper. On the other hand, the LST had a significant spatial clustering characteristic, and the spatial pattern of LST had a scale effect. (4) The dominance of built-up land significantly strengthened surface temperature, while increasing the dominance of vegetation and water could cool temperature. Cropland displayed no sign of cooling effect, the LST tended to be stable as the percentage of cropland increased. The results of the study can provide a useful reference for improving urban thermal environment and developing sustainable cities in estuarine regions.
