Study On Phenophase Extraction Of Spring Maize In Liaoning Province Based On MODIS NDVI Data

Maize is the most important grain crop in Liaoning Province.Access to maize phenophases can improve its field fine management level,and it is significant for monitoring seasonal changes of maize and ensuring food security.Access to maize phenophases by remote sensing is mainly based on the vegetation index time-series,and it has the advantages of wide coverage,long monitoring time,higher frequency acquisition.MODIS data has high spectral resolution,temporal resolution and spatial resolution.It is an ideal data to extract the spring maize phenophases.The study is based on the MODIS NDVI data of 2007,2008 and 2009,the fitting methods of Asymmetric Gaussians functions(AG),Double Logistic functions(DL)and Savitsky-Golay filter(S-G)are used to reconstruct the NDVI time series of spring maize.The root mean square error(RMSE)of reconstructed and original time-series is used to evaluate the fidelity ability of each fitting method.The fitting degree(R2)of reconstructed curve is used to evaluate the smoothness ability.The study proposes and uses the average weigh comprehensive evaluation index to evaluate the comprehensive ability of each fitting method.The dynamic amplitude threshold method,inflection point method and maximum value method are used to extract the key phenophases(seedling,jointing,tasseling and maturity stage)of spring maize in Liaoning Province,and the phenophases data observed at 12 agrometeorological stations in Liaoning are used to verify the extraction result.The main conclusions are as follows:First,the fidelity and comprehensive ability of AG method are better than the other two methods.Its average RMSE of each year is 0.035,0.050 and 0.041,and its average weigh comprehensive evaluation index is 2.11.The smoothness ability of S-G method is better than the other two methods.Its average R2 of each year is about 0.99.Second,the average extraction effect of all key phenophases and the extraction effect at the jointing,tasseling and maturity stage based on the AG method are better than those of the other two fitting methods.Its mean absolute error(MAE)of each year is less than 5d,and its root mean square error(RMSE)is less than 4d.The S-G method has the best extraction effect at the seedling stage,whose seedling stage RMSE is 4.1d,4.0d and 2.9d for each year.Third,the dynamic amplitude threshold method is better than dynamic threshold method.It can effectively correct abnormally low values and make the extraction results better.Fourth,as a whole,the key phenophases of spring maize in Liaoning are gradually delayed from south to north,that is,with the increase of the latitude,its growing period is gradually delayed.Spring maize growing period in the coastal planting area is relatively short.There are late sowing and late harvesting in northern Liaoning and hilly areas in eastern Liaoning.