Effects Of Irrigation Treatments On Rice Hyperspectrum And Rice Growth Monitoring Under Heat Stress

Irrigation is an agricultural measure that is often used during the occurrence of heat stress to rice.Groundwater and surface water are the most important and commonly used irrigation water sources in actual crop production.Hyperspectral technology is widely used in crop growth monitoring because of its high spectral resolution,large amount of spectral information,and accurate spectral data.This study used well water(W),pond water(P)and control group(CK)treatments to irrigate the paddy field when the heat stress of rice occurred,combining rice physiological data and rice canopy reflectance spectrum data,studying the effects of different irrigation treatments on rice canopy spectrum,analyzing the chlorophyll content,dry matter accumulation,yield and rice canopy spectrum of rice of different irrigation treatments under heat stress,screening the sensitive bands through the full-band normalized difference spectral index(NDSI)and continuous wavelet transform(CWT)methods to of rice chlorophyll content,dry matter accumulation and yield,analyzing the changes of sensitive bands under different irrigation treatments,and establishing estimation model of chlorophyll content,dry matter accumulation and yield of rice under different irrigation treatments under high temperature background.This study used hyperspectral data to explain the effects of different irrigation treatments under heat stress of rice,and provided data support for remote sensing inversion of chlorophyll,dry matter and yield of rice under heat stress,with a view to providing theoretical support for the research on irrigation methods to alleviate the heat stress of rice.The main conclusions are as follows:1.The original canopy reflectance spectrum of crops in the visible light band,the magnitude of the reflectance of different irrigation treatments is P treatment>CK treatment>W treatment,in the near infrared band,the magnitude relationship is W treatment>CK treatment>P treatment.The difference between the original reflection spectrum of rice in the780-900nm band and the 970nm band is the largest.The red peak value of the first-order derivative spectrum of W treatment is the highest,followed by CK treatment,and the lowest is P treatment;as the development stage goes from heading to maturity,the red edge parameters appear"blue shift"phenomenon,the relationship of the red edge parameters between different treatments is P processing>CK processing>W processing.2.Correlating the chlorophyll content of the three developmental stages with the rice canopy reflection spectrum of the corresponding period,the study showed that the 550-680nm band range is the chlorophyll sensitive band,and the model fitting effect of the heading stage between different developmental stages is the best.The partial least squares regression(PLSR)model constructed by sensitive wavelet coefficients CWT(3,735),CWT(2,736)and CWT(1,737)at the heading stage has the best fitting effect on chlorophyll,the model testing r~2=0.84;The PLSR models constructed by sensitive wavelet coefficients CWT(5,569),CWT(2,654)and CWT(7,562)at the filling stage has the best fitting effect on chlorophyll,the model testing r~2=0.76;The PLSR model constructed by the sensitive wavelet coefficients CWT(1,815),CWT(7,575)and CWT(3,578)at maturity has the best fitting effect on chlorophyll.3.Correlating the final dry matter accumulation and the canopy reflectance spectra of the three developmental stages,the study found that the sensitive band of dry matter is in the 600-670nm band.The PLSR model established by the spectral indexes NDSI(694,726),NDSI(492,819)and NDSI(689,715)has a better fitting effect on the final dry matter accumulation than the ordinary function model,the model testing r~2=0.79;The PLSR model established by the sensitive wavelet coefficients CWT(7,987),CWT(6,797)and CWT(5,774)has a better fitting effect on the final dry matter accumulation than the ordinary function model,the model testing r~2=0.84.4.Correlating the final yield and the canopy reflectance spectra of the three developmental stages,the study found that the sensitive band of yield is 740-900nm.The PLSR model established by the spectral indexes NDSI(431,891),NDSI(693,697)and NDSI(752,452)has a better fitting effect on the final output than the ordinary function model,the model testing r~2=0.81;The PLSR model established by the sensitive wavelet coefficients CWT(6,838),CWT(6,841)and CWT(7,975)has a better fitting effect on the final output than the ordinary function model,the model testing r~2=0.83.5.Comparing the empirical model of vegetation index and the wavelet analysis model,the models established by both have higher accuracy and can accurately retrieve the physiological parameters of rice,and the fitting effect of the wavelet analysis model is better.This is because compared with the NDSI model,the CWT model decomposes the spectral signal and mines more hidden spectral information,so the adaptability has a certain improvement compared to the NDSI model.