Risk Assessment and Monitoring of Winter Wheat Waterlogging Combining Ground-based Observations and Satellite-derived Data

Author:Chen Yuan is round

Supervisor:Huang Jingfeng


Degree Year:2018





Under the background of global climate change,extreme weather and climate events have increased.Frequently agro-meteorological disasters have had an increasingly severe impact on agricultural production,and seriously affected food security in China.Waterlogging is a serious agro-meteorological disaster causing crop yield losses in Southern China.The middle and lower reaches of the Yangtze River is one of the major winter wheat growing regions in China while also waterlogging prone due to its specific climatic conditions.Therefore,it is of great significance for disaster prevention and mitigation,and food security to strengthen the monitoring and evaluation of crop waterlogging in this region.3S technology provides new methods for monitoring,early warning and assessment of agrometeorological disasters.This study focuses on the risk assessment and monitoring of winter wheat waterlogging in the middle and lower reaches of the Yangtze River using 3 S technology and data fusion method,and focuses on the construction of a technical framework for monitoring and evaluating winter wheat waterlogging from multiple sources.The main research contents and results are as follows:(1)Based on disaster risk analysis theory,this study integrates the sensitivities of waterlogging hazard formative environments,hazard risk,and vulnerability of hazard-affected areas and builds a synthetic waterlogging risk assessment model.A series of environmental factors,including geographical,agricultural,and meteorological variables such as slope,low-lying degree,soil waterlogging index,climate risk probability,physical exposure of winter wheat,and disaster-resisting performance index,are combined to facilitate the building of a comprehensive winter wheat waterlogging risk assessment model.The study area is classified into high,secondary high,moderate,and low risky areas according to the constructed risk model.The results show that the secondary-risk to high-risk areas of waterlogging are mainly distributed in the plains of Hubei Province,in addition to the plain and polder areas along Yangtze River across Anhui and Jiangsu Provinces;the proposed model could more comprehensively reflect the occurrence mechanism of winter wheat waterlogging by synchronizing geographical,agricultural,and meteorological factors;the waterlogging regionalization based on the model could reasonably represent the spatial distribution and differentiate regional characteristics of winter wheat waterlogging in the study area.(2)A new downscaling-integration framework for combining the Version 7 TRMM 3B42/3B43 precipitation product and rain gauge observations is developed to obtain improved high-resolution(1 km×1 km)monthly and ten-day precipitation estimates.Firstly,an area-to-point kriging(ATPK)approach is used to downscale the original TRMM product to 1 km,so as to ensure a fair comparison with rain gauge data.Then,the downscaled TRMM precipitation datasets are integrated with the gauge observations using geographically weighted regression kriging(GWRK).As with the downscaled TRMM precipitation datasets,additional geographical factors(i.e.longitude,latitude and elevation)are also used as auxiliary variables in the local regression predictions with GWRK.Applying this approach to an experiment conducted at the middle and lower reaches of the Yangtze River in China from 2001 to 2014 shows that:the downscaled monthly TRMM precipitation data by ATPK are more accurate than the original TRMM estimates;the GWRK model employing the downscaled TRMM precipitation data and geographical factors provides better monthly precipitation estimates than the conventional ordinary kriging(OK)interpolation and the commonly used merging methods(i.e.geographical difference analysis,GDA and kriging with external drift,KED);the GWRK method reduces the influence of the inaccuracy(bias)of satellite-derived precipitation data on the precipitation estimates compared to GDA.The approach presented in this study has provided an efficient alternative for solving the scale mismatch problem between point-based gauge data and low resolution satellite data,and producing improved precipitation data at high spatial resolution.(3)Using the precipitation data observed by the regional automatic meteorological station of Anhui province,the impacts of rain gauge density and distribution on GWRK precipitation estimates obtained by combining rain gauge observations and satellite derived precipitation data were further examined in detail.The results show that:with the increase of gauge density,the accuracies of GWRK merged results of both monthly and ten-day precipitation are gradually improved,and then gradually level off.In the case of a lower gauge network density,the accuracy improvement of precipitation estimates is relatively larger;under a certain gauge density,the accuracy of merging results of monthly and ten-day precipitation are fluctuated influenced by different spatial combinations of rain gauges.The smaller the gauge network density,the greater the volatility,and vice versa;the influence of gauge distribution on the precipitation estimation accuracy of GWRK under low gauge density is greater than that for high gauge density.At small gauge densities,it could produce better precipitation merging results than that at greater gauge densities.(4)According to the differences in the variation characteristics of the MODIS-NDVI time series of typical land cover types during the winter wheat growth period,a decision tree classification model is established to extract the winter wheat planted areas from 2001 to 2015.The results of accuracy verification show that the method adopted in this study can obtain satisfactory results overall.Compared with the winter wheat planting areas recoded by the statistical yearbook at province level,the relative error of winter wheat planting areas estimated by MODIS data in the study area from 2001 to 2015 is between-4.03%and 9.54%.Compared the municipal winter wheat statistics with winter wheat planting areas extracted by remote sensing,their area values are near the 1:1 line in the scatter plots.(5)The spatial distributions of precipitation anomaly percentages of winter wheat growing season(October of last year to May this year)and standardization precipitation index(SPI)values at 6-months scale(December of last year to May this year)from 2001 to 2014 are acquired by using GWRK merged monthly precipitation data.The results show that the interannual variation trends of the waterlogging monitoring results based on precipitation anomaly percentage and SPI index are consistent;the entire study is prone to waterlogging in 2002,2003 and 2010.The monitoring results of winter wheat waterlogging from 2001 to 2014 by using GWRK merged ten-day precipitation data and winter wheat waterlogging index indicate that:compared with station monitoring results,the waterlogging monitoring results based on merged precipitation data can not only capture the occurrence of waterlogging in most stations,but also reflect the distribution of waterlogging in the entire space.The monitoring accuracy of waterlogging occurrence is above 70%in most years(10/14),and 88%and above in half of the years.The monitoring accuracy of typical waterlogging years(2001,2002,2003 and 2010)is even above 90%.The monitoring results of waterlogging affected areas in winter wheat show that different degree of waterlogging occurs every year in the winter wheat growing areas of the study area during the 2001 to 2014 period,and only a few areas are affected by waterlogging in 2004-2007 and 2011-2012.In 2001,2002,2003 and 2010,most of the winter wheat planting areas suffered from different degrees of waterlogging,the affected areas are all over 1680 thousand hectares,and the affected area reaches 2773 thousand hectares in 2010.(6)Taking Anhui province as an example,the monitoring results of waterlogging at different growth stages of winter wheat from 2013 to 2015 by using the merged precipitation data of the regional automatic meteorological observation data and TRMM satellite precipitation indicate that:the winter wheat waterlogging occurred in Anhui province in 2013-2015 is mainly mild,and the waterlogging-affected areas of winter wheat are very small in the three years;from January to May 2013,winter wheat waterlogging occurred in the grain filling period;in 2014,both wintering and grain filling stages of winter wheat had waterlogging;in 2015,winter wheat seeding,jointing and filling stages all suffered from waterlogging.The waterlogging monitoring results synergizing vegetation index anomaly show that the NDVI anomaly may not present a negative value when the precipitation anomaly percentage is high,but the variation of the NDVI anomaly time series can reflect vegetation response to precipitation to some extent.The relationship between vegetation index anomaly and precipitation anomaly is complicated.It is not enough to monitor crop waterlogging only based on the anomaly of vegetation index,and comprehensive analysis is therefore needed through the combination of various factors.