Study On Extraction And Change Detecting Methods Of Irrigated Drylands And Their Spatio-temporal Characteristics Analysis

Irrigation plays an important role in agricultural development and regional environmental change.Therefore,an accurate understanding of the structural proportion,spatial distribution pattern and space-time change characteristics of irrigated and rainfed drylands in China can provide a basis for overall accurate decisionmaking.Based on the Google Earth Engine(GEE),this study established the time series data of growing season using Landsat data with medium and high-resolution,proposed the extracting method of irrigated drylands and its dynamic change information from2000 to 2015,and obtained the patterns of spatial distribution and changes of irrigated drylands in study period.Firstly,according to the differences of natural environment,study area is divided into five areas.The characteristic indexes were analyzed and selected in each area,and the best classification methods of each area were obtained respectively.The classification methods were used to classify the data in 2000 and 2015.This direct result of 2015 is the final result,but the direct result of 2000 is the preliminary result because of the lacking of samples in 2000.The dynamic information of irrigated drylands from2000 to 2015 was extracted from the gap data of remote sensing factors of the two periods combining the direct classification results.The final distribution of irrigated drylands in 2000 was derived from the distribution data of 2015 and the changing data from 2000 to 2015.The validation shows that this result has higher accuracy than that of direct classification result in 2000.Finally,the distribution pattern of irrigated drylands in China and its temporal and spatial changing characteristics from 2000 to2015 were analyzed and summarized.The following conclusions are drawn:The classification methods of drylands based on the regional scale research can select the best distinctive feature combination and the best distinctive phases according to the own characteristics of each area.The common indexes,EVI,NDVI and NDWI,have good discrimination,among which the combination of NDVI and NDWI is the best.The effects of slope and texture factors are various in different area.The slope has a strong distinction in the north plateau area and the middle eastern area.The texture is helpful to improve the accuracy in the other three areas except the southern area,and the inverse difference moment characteristic of the gray level co-occurrence matrix can do the best.Although the zoning has reduced environment differences to a certain extent,there are still differences within each area.Therefore,environmental factors were added to weaken the difference of vegetation index caused by the difference of natural environment.It is found that the ratio of NDVI and NDWI to precipitation accumulation and the ratio of NDVI and NDWI to temperature accumulation have better classification effect.When using less phase combined data,the classification accuracy using ratio of remote sensing factors to environmental factors remain stable,but it decreases significantly using them directly.Finally,the average overall accuracy of each area is 0.8765,the average kappa coefficient is 0.685 in 2015,and the root mean square difference is 0.142 verified with the statistical data of small and medium-sized cities and counties.The final result was also consistent with the existing data.These show that these methods are reliable and can achieve better extraction accuracy.The dynamic information of irrigated drylands from 2000 to 2015 was extracted by using the gap data of remote sensing factors of the two periods and combined with the direct classification results.In this step,mean,variance,and maximum value were studied.The distribution data of irrigated drylands in 2000 was derived by using the distribution data in 2015 and the dynamic data from 2000 to 2015.It is found that the combination of the three indicators can obtain good effect in each area.The overall accuracy and Kappa coefficient of the data derived from this method in 2000 higher than the direct classification results respectively.This also indirectly proves that the result of dynamic change data is accurate and the extraction method is reliable.This method not only avoids the error accumulation caused by the overlay analysis of classification results of two periods,but also overcome the problem of insufficient classification accuracy caused by the shortage of samples and data in the past period.The distribution of irrigated drylands in China is divided into four strips from southeast to northwest,which is consistent with the distribution of precipitation.In the southeast area,the irrigated drylands present sporadic distribution;in the North China area,they present concentrated distribution;in the northeast and middle north areas,they present staggered distribution;and in the northwest area,they present a scattered distribution.Precipitation is the basic factor affecting the distribution of irrigated drylands.So,at the national scale,the proportion of irrigated drylands is negatively correlated with the precipitation in the growing season of crops.Topographic is the second factors,so irrigated drylands are distributed in low-lying plains,valleys,and basins.The convenience of obtaining water resources is the third factors.Under the influencing factors of precipitation and terrain,water sources and water conservancy facilities have become the key factors for the existence of irrigated drylands.During the study period,the distribution pattern of irrigated drylands trended to be dense,and its barycenter have shifted towards the northwest.Rainfed drylands have moved to more remote mountainous areas,and the barycenter moves towards the northeast.Some irrigated drylands in plain regions have been transformed into non drylands,and rainfed drylands have been added in remote areas such as mountains and hills.The abandoned rainfed drylands and the reclaimed irrigated drylands have increased in Northwest China.The above changes reflect the trend that irrigated drylands move to regions lacking water resources,while in the regions with relatively abundant water resources irrigated drylands are reduced.The mismatch between water resources and cultivated lands is aggravating in China.A set of extraction methods of Chinese irrigated land was proposed in this paper,which include extraction method of irrigated land in recent years,change information extraction method of irrigated drylands and extraction method of irrigated dryland in past years.The 30 m resolution distribution data of irrigated and rainfed drylands in China with high accuracy were obtained,which refine the spatial resolution of the existing data.The spatial distribution pattern and change pattern of irrigated and rainfed drylands in China and its influencing factors and laws were summarized.