Impacts Of Drought And Flood Disasters On Rice Yield In Jiangxi Based On Remote Sensing

In recent years,global warming and frequent extreme climate change events have posed a great threat to China’s food production.Remote sensing technology(RS)has been widely used to evaluate the impact of drought and flood disasters on agricultural productivity.At present,the main research areas include crop performance monitoring,water consumption quantification,soil characterization and yield estimation.The main objective of this research is to develop an operational methodology for predicting rice yield which is utilized to assess the impacts of drought and flood disasters on agricultural productivity taking Jiangxi,China as an example.For this purpose,we first calculated the Standardized Precipitation Index(SPI)using monthly rainfall data of 83 stations from 1960to 2020 to identify the disastrous years and their spatial extent of impact.Then time-series MODIS data from 2000 to 2020,Landsat 5 TM and Landsat 8 OLI images,digital elevation model(ASTER GDEM and STRM)were obtained and utilized for rice plantation mapping using decision-tree algorithm and yield estimation.The identified 3-cropping rice plantations were converted into rice masks,and the peak values of the standardized difference vegetation index(Normalized Difference Vegetation Index,NDVI)without clouds or with minimal cloud cover were calculated after application of a maximization algorithm and accumulated to obtain the county-level cumulative NDVI(C-NDVI).Combined with the county-level annual rice yield data from 2016 to 2019,a rice yield estimation model based on remote sensing was established.By comparing the determination coefficient(R~2),root mean square error(RMSE)and prediction accuracy of the models,the best model for rice yield estimation was selected.On this basis,a new disaster indicator,i.e.,the disaster composite index(DCI)was developed for assessing the disaster impact on annual rice production in a pilot area around Nanchang city to realize the quantitative analysis of drought and flood disasters on rice yield.The main conclusions are presented as follows:(1)In the past 23 years,drought and flood disasters of varying degrees occurred almost every year in Jiangxi Province.Extreme big droughts occurred in 2003 and 2009,and extreme big floods arose in 1998,2010 and 2020.The occurrence frequency of drought and flood disasters in summer is generally higher than that in spring,and the severity of drought and flood disasters is gradually strengthened in the central and northern parts of the province.The disasters occurred mostly in Jiujiang,Shangrao,Fuzhou,Yichun and Pingxiang.(2)The difference between the rice plantation area estimated by RS and the governmental reported area from the statistic year books was within a reasonable range,the overall classification accuracy(OA)was greater than 94%,and the annual relative errors of both were less than 3%.From 2010 to 2019,the rice planting area showed first an increase and then a decrease,and reached the peak in 2016.Most of the rice planting areas are distributed in the north of the province,mainly around the Poyang Lake.(3)The rice yield estimation model was used to predict the rice yield of the province in 2020,which was 20,387,579 tons,with a unit yield of 6.2 tons/ha.The relative error was0.62%compared with the reported yield,indicating that the rice yield estimation model based on remote sensing built in this research is able to achieve reliable yield prediction and be extended to and applied for rice yield estimation elsewhere.Due to the impacts of droughts and floods,through comparison with the least affected years,the decrease in rice production in the studied province reached at maximum 698,304 t in 2010-2019.(4)A new disaster indicator,the disaster composite index(DCI),was developed for assessing the disaster impact on annual rice production in a pilot area around Nanchang City(including Nanchang County and Xinjian District).Results show that the difference between the estimated rice yield and the reported yield ranged from 0.03%to 7.8%,indicating that the developed models are capable of achieving rice yield estimation using both MODIS data for regional and Landsat data for local pilot-area scale.The more serious the drought and flood disasters,the more the rice yield affected,and there was a significant negative correlation between them.In comparison with the normal or least affected year,impacts of drought and flood were successfully evaluated using DCI for the pilot area and their negative correlation was visualized,that is,y=-56.715x+37591.945(R~2=0.890).We hence believe the developed methodology is extendable to South China and other rice plantation area in the world,which has positive significance for food security prediction and early warning.