Detection of phonological transitions of spring maize in Northeast China during the last 20 years using daily NOAA/AVHRR NDVI temporal series data

Abstract:

The normalized difference vegetation index (NDVI) provides a rough measure of vegetation amount and growing condition of crops when vegetation activity is low to moderate. Based on the Leaf Collar Method, two key phenological phases, i.e., third leaf collar (TLC) and the maturity, are selected for NDVI modeling. The records on crop phenology were available from 1992 to 2013 at 103 stations in the Northeast China. However, there are large amount of missing data. Therefore, a statistic model is desirable to fill the gaps then, analyze the characteristics of the TLC and the maturity stage with the full data set. The Savitzky-Golay filter was used for noise reduction and temporal NDVI smoothing. The slope analysis was used for detection of TLC and maturity date of spring maize in the spring and in the fall, respectively. When NDVI slope values reach the turning point in certain period, the corresponding date is selected as TLC or maturity. Through comparison between observation and estimation, we find that 5-day slope method is robust to detect the changes of maize phenology. This study shows that the average estimation is 2 days earlier than observation. We then use this method to generate the TLC and mature dates for all the stations. The analyses of this full data set shows that the average TLC of spring maize in Northeast China emerges on Jun.2. The average maturity of spring maize appears on Sep. 18. The shortest growing season of 104 days appears in Jilin Province, while the longest growing season appears in Heilongjiang province of 116 days. When the latitude decreases, the annual average temperature and precipitation amount increases. Accordingly, TLC becomes earlier from Heilongjiang, Jilin to Liaoning Province. There is a significantly negative correlation between TLC that is around June and temperature of April and May. One-month time lags of climate factor, therefore, should be added to detection of phonological transitions of spring maize.