Assessing the impacts of natural disasters on rice production in Jiangxi, China. Issue 5 (4th March 2022)

• Record Type: Journal Article
• Authors: Fu, Xiao; Zhao, Guannan; Wu, Weicheng; Xu, Bin; Li, Jie; Zhou, Xiaoting; Ke, Xinxin; Li, Yuan; Li, Wenjing; Zhou, Cuimin; Jiang, Jingheng; Zhang, Ming; Liu, Yixuan; Tu, Qihong
• Is Part Of: International journal of remote sensing. Volume 43:Issue 5(2022)
• Page Start: 1919
• Page End: 1941
• Publication Date: 2022-03-04
• Subjects: drought and flood disasters -- SPI -- NDVI -- decision tree classification -- rice area extraction -- yield estimation models; Remote sensing -- Periodicals; Télédétection -- Périodiques; 621.3678
• DOI: 10.1080/01431161.2022.2049914
• Languages: English
• ISSNs: 0143-1161

Abstract

ABSTRACT: Agricultural productivity is affected by natural disasters, such as drought and flood, which can be assessed by remote sensing (RS) technology. The main objective of this research is to develop an operational methodology for predicting rice yield that 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 1960 to 2020 to identify the disastrous years and their spatial extent of impact. Then, time-series MODIS data from 2000 to 2020, Landsat images, digital elevation model (DEM) 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 for calculating the county-level cumulative NDVI (normalized difference vegetation index) to match the reported county-level annual rice yield for building remote sensing-based rice yield models taking the averages of 2016–2019 as test. By comparing the determination coefficient (R 2 ), root mean square error (RMSE) and prediction accuracy of the models, we found that the best model for rice yield estimation is Y D = 4.899 × 10 −6 × NDVI 2 + 2.891 × NDVI + 98511.218 (R 2 = 0.898). This model was applied to estimate the county-level annual rice production of 2014 and 2015, and provincial annual rice production of 2010–2019 for validation and prediction. The agreement with the government-reported yield reaches 97–98% at provincial level and the models were hence considered reliable. After verification, the models were applied to other remained years after 2010 for estimating the provincial annual rice production. Then, a new disaster indicator, that is, the disaster composite index (DCI) was developed for assessing the disaster impact on annual rice production in a pilot area around Nanchang City. Results show that developed models are capable of achieving rice yield estimation using both MODIS data for regional and Landsat data for local scale. In comparison with the normal or least affected year, impacts of drought and flood were successfully evaluated using DCI for pilot area and their relationship was created. We hence believe the developed methodology is extendable to South China and other rice plantation area in the world.