Agricultural Drought Monitoring Using Multi-source Remote Sensing Data in China
To avoid encountering 504 error when downloading big files,please take CAJ download or PDF download by chapter as your first priority.
【Author in Chinese】 Tehseen Javed；
【Author's Information】 西北农林科技大学， 水利工程， 2020， 博士
【Abstract】 Drought is an insidious hazard of nature,which is considered by many to be the most complex but least,understood of all-natural hazards.For monitoring of drought,large historical datasets are required,which involves complex inter-relationship between the climatological and meteorological data.The extraction of valuable information from such extensive data archives demands an automated and efficient way.Data mining is the answer to the above problem as it has the potential to search for hidden patterns and identify the relationship between the data.China has also been suffered from frequent droughts events,which has brought potential hazards to sustainable crop production.Therefore,the study of spatial-temporal variation characteristics of agricultural drought monitoring plays a vital role in drought relief and agricultural planning.The study analyzes the relationship between the metrological and agricultural drought,the impact of drought on vegetation phenology,and productivity and investigates the drought indices performances for prediction of agriculture drought.Multiple datasets were used in this study,daily precipitation and temperature datasets from 763 stations,between1961-2017 china meteorological bureau,satellite gridded monthly precipitation,land cover,thermal bands,normalized difference vegetation index(NDVI),and soil moisture.For assessment of drought events and there correlation the following drought indices were calculated,standardized precipitation index(SPI),standardized precipitation-evapotranspiration index(SPEI),precipitation anomaly,vegetation condition index(VCI),NDVI anomaly,enhanced vegetation index(EVI),standardized soil moisture index(SSI),multivariate standardized drought index(MSDI),and vegetation health index(VHI).To find the correction and trends of these drought indices,the following statistical analysis was performed;Pearson correlation coefficient(r),linear regression,coefficient of determination(R2),and root mean square error(RMSE)and modified Mann-Kendall(MMK).The study was divided into four phases.In the first phase of the study,drought was assessed under four different land cover types,cropland,forestland,grassland,and desert-land in China.The modified Mann-Kendall test was used to detect the significance of a trend.The Pearson correlation and coefficient of determination methods were used to find the relationship between NDVI anomaly,VCI,precipitation,and SPI.In the second phase,investigate the impacts of drought or wet conditions on the vegetation phenology and productivity across the different sub-regions(northwest(NW),north China(NC),QinghaiTibet area(QTA),and south China(SC).Daily rain gauge datasets were used to predict the air temperature,precipitation trend,and to compute the standardized precipitationevapotranspiration index(SPEI).Remote sensing-based EVI data from moderate resolution imaging spectroradiometer(MODIS)were used to evaluate the vegetation phenology.In the third phase,investigate the drought indices(SPI,SSI,MSDI,and VHI)performances for prediction of agriculture drought.In fourth phase using globally remote sensing(RS)based gridded monthly precipitation,Climate Hazards Group Infra-Red Precipitation and Station(CHIRPS),normalized difference vegetation index(NDVI)and land surface temperature(LST)datasets over 1982-2018 were utilized to derive agricultural Standardized Precipitation Index(a SPI),and Vegetation Supply Water Index(VSWI).The main findings of the study are the following:(1)The mean monthly and yearly precipitation had a general land cover type rank of forestland > grassland ≈ cropland > desert-land.A positive correlation was found between drought indices(NDVI anomaly,VCI,SPI)and precipitation for different land cover types.The NDVI anomaly and VCI were well correlated with 3-month SPI for cropland and were well correlated with 6-month SPI for forestland.VCI performed better than NDVI anomaly when correlating with SPI.The coefficient of determination(R2),were obtained for precipitation and VCI in the driest(2011)and wettest(2016)years.The R2 values for desert and grassland ranged from(0.70-0.90),and cropland and forestland were lower(0.54-0.69).Only precipitation,SPI,and VCI of cropland had significantly increasing trends.The spatial distribution patterns of precipitation,NDVI,and VCI increased with decreased elevation.The study revealed that desert and grassland had been regularly exposed to moderate or extreme droughts conditions and confirmed that desert and grassland are more sensitive to short-term drought.(2)The air temperature had significant increasing trends(p < 0.05)in all subregions.Precipitation showed a non-significant increasing trend in northwest China and insignificant decreasing trends in north China,Qinghai Tibet area,and south China.Integrated enhanced vegetation index(i EVI)and SPEI variations depicted that 2011 and 2016 were the extremely dry and wet years over 2000-2017.Rapid changes were observed in the vegetation phenology and productivity between 2011 and 2016.In 2011,changes in the vegetation phenology with the length of the growing season(ΔLGS)= was-14 ±36 days.In 2016,the overall net effect changed at the onset and end of the growing season with ΔLGS of 34±71days.The climatic sensitivity had a changing rate of 0.16 from arid to semi-arid regions and declined from semi-humid to humid regions with a decreasing rate of-0.04.A higher association was observed between i EVI and SPEI as compared to i EVI and precipitation.North China is more sensitive to climatic variation.(3)The relative soil moisture and VHI depicted similar patterns,while slight variations with precipitation.The MSDI performed well against VHI,compared to SPI and SSI.The correlation among the 1-month SPI,SSI,and MSDI with monthly VHI the maximum r-value(0.21)were obtained in northwest China with the r-values of 0.15,0.17,and 0.21,respectively.The correlation among the 3-month SPI,SSI,and MSDI with monthly VHI,the highest value was obtained in north China,followed by South China,Qinghai-Tibet area,and northwest China,with the r-values of 0.72,0.68,0.63,and 0.57 respectively.While correlation among the 6-month SPI,SSI,and MSDI with monthly VHI,maximum r(0.58)value obtained in south China,followed by Qinghai-Tibet area,north China,and northwest China(r = 0.54,0.45,0.41 respectively).The VHI shows a significant increasing trend for northwest,Qinghai-Tibet area,and south China with Mann-Kendall Z values of 2.26,4.09,and 4.70,respectively,and the insignificant increasing trend in north China.(4)Overall the three timescales(1-,3,and 6-months)of a SPI show that extreme drought events occurred in the 21 st century,and the more frequent extreme drought events occurred in the winter wheat growing season.In the wheat-growing season northwest and north China,the much frequency of extreme drought events occurred in April for three timescales of a SPI.While in Qinghai-Tibet area and south China the most frequently extreme drought events found in December and May,respectively.On the other hand,during the corn growing season in the northwest and south China,the most frequency of extreme drought events occurred in July,while in north China and Qinghai-Tibet area,August and September respectively.A higher correlation was obtained for the pair of 3-month a SPI and VISW or crop yield anomaly(YAI)in four sub-regions of China.Overall the summer corn yield shows the significant increasing trends,while wheat yield in the northeast of north China illustrations significantly increasing trends.