Improved mapping and change detection of the start of the crop growing season in the US Corn Belt from long-term AVHRR NDVI

Long-term monitoring of crop phenology is a critical process to understand changes in cropland vegetation dynamics related to climate and human management. During the last four decades, seasonal changes in vegetation have been measured by satellite instruments, the Advanced Very High Resolution Radiometers (AVHRRs). However, an acceptable detection of crop phenology has been prevented due to its coarse spatial resolution (~8 km) pixels in which croplands are frequently mixed with non-cropland. This study introduces a novel method for detecting the start of the growing season in cropland (SOSC) from the AVHRR Normalized Difference Vegetation Index (NDVI) to examine the long-term changes in SOSC at the 8-km pixel level across the core cropping regions of the Midwestern United States (the so-called ‘the Corn Belt’) for 1982–2015. The AVHRR-based SOSC was compared with the SOSC retrieved from the pixels covering mostly croplands of the 250-m Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI. These two satellite-based SOSC shows reasonable correspondence in spatial patterns so that the annual values of the root mean square difference ranges from 3.6 to 7.4 days during the overlapping period of 2008–2015. Also, an acceptable correspondence was found between the interannual variations in the AVHRR SOSCs and in the state-level survey of crop stages. For sowing stages, correlation coefficients ranged from 0.43 to 0.68; for flowering stages those ranged from 0.51 to 0.83. By analyzing long-term SOSC changes with climate factors, we found a strong tendency that warmer pre-season condition is accompanied with earlier SOSC. In addition, larger pre-season precipitation tends to postpone SOSC as strongly as the pre-season temperature affects SOSC in a considerable portion of the core cropland. This study suggests that the postponing of sowing due to larger precipitation, which has rarely been highlighted in previous studies, requires more attentions in order for better understanding of vegetation dynamics in agriculture-climate studies.