Title : Analysis of the impacts of natural disasters on food production by remote sensing taking Jiangxi, China as an example
Natural disasters, typically, drought and flood, exert direct impact on crop growing performance and productivity. The objective of this research was to conduct such an impact analysis using meteorological, statistic and multi-resolution remote sensing data taking Jiangxi, China as an example. We first calculated the Standardized Precipitation Index (SPI) based on the monthly rainfall data from 83 meteorological stations of the period 1960-2020 covering the whole province to identify the drought and flood years. Then time-series MODIS vegetation data, digital elevation model (DEM) and its derived slope were employed to define the staple food, i.e., the paddy plantation area of three different cropping rice, namely, early rice (Apr-July), middle cropping rice (Jun-Sept) and late rice (Jul-Oct) by Decision-tree approaches using Landsat images and Google Earth for verification. And then, an exploration on the relationship between the vegetation indices such as NDVI, EVI and LAI and the reported rice yield (Y) was conducted to build remote sensing-based yield model. Results show that among the all the test models, those coupling the accumulated county-level average of the peak NDVI of three cropping rice of the period from 2014 to 2019 with the reported county-level annual mean rice yield are most effective for estimating the annual rice yield of each year for the whole province. The derived models are shown as follows: y = 106489.574+0.01x+9.231E-11x2 (R2= 0.894) or y = 93087.727+0.015x (R2 = 0.888), where y is the predicted county-level annual total rice yield and x the accumulated county-level peak NDVI of three cropping rice. With these models, the predicted province-level annual rice yield is in a good agreement with the government reported annual rice yield with a little difference of about 0.59-2.18%. Taking 2019 as an example, the predicted county-level annual yield is well consistent with the reported county-level annual yield, that is, Yp = -1889.157 + 0.966Xr (R2 = 0.885), where Yp is the predicted county-level yield and Xr the reported county-level yield. As revealed by SPI analysis on the rainfall data from 1960-2020, it is noted that in the very recent five years’ period, 2016 and 2017 were the normal years while 2018 was flooded and 2019 suffered from both flood and drought. In comparison with 2016 and 2017, we found that a reduction of 50,000 and 910,000 t of rice production in the province respectively in 2018 and 2019. We also noted that the government had slightly overstated the rice production by about 161,000-262,000 t in these two years. In conclusion, SPI-based analysis and time-series of remote sensing processing and modeling allow us to achieve staple crop yield prediction and analyze the impacts of natural disasters on the former.