Agricultural growing season calendars derived from MODIS surface reflectance

Figures & data

Table 1. A list of the phenological transition date parameters used in this research, accompanied by their biophysical definitions, as well as their algorithm specifications.

PTD parameter name	PTD parameter definition and algorithm specification
Start of season (SOS)	Greenness onset; emergence of above ground biomass; first point at which an upward trending NDVI which precedes the NDVI maximum (peak) surpasses a given threshold
Peak period start (PPS)	Onset of green leaf area maximum; start of the period during which the NDVI maximum is likely to occur; first point above 75% of annual range in NDVI which precedes the NDVI maximum (peak)
NDVI maximum (Peak)	The NDVI maximum observed within a given growing season
Peak period end (PPE)	Onset of senescence; end of the period during which the NDVI maximum is likely to occur; last point above 75% of annual range in NDVI which follows the NDVI maximum (peak)
End of season (EOS)	End of senescence; termination of photosynthetic activity; last point at which a downward trending NDVI which follows the NDVI maximum (peak) dips below a given threshold

Figure 1. A flowchart depicting the pre-processing and data preparation steps, the basic logic of the PTD extraction algorithm, as well as the steps leading to the generation of the final data products. Intermediate data products and steps are in solid rectangles, actions are connecting arrows accompanied by italicised text, and final data products are in soft-cornered rectangles.

Figure 2. Plots of the gap-filled time series of 8-day composites of scaled NDVI (× 1000), 2001–2010 (46 per year × 10 years = 460 values), from which PTDs were detected. (a) NDVI time series from a location in Southern Kansas, providing an example of a winter wheat dominated landscape, with the SOS being detected in the October-November time period, and the EOS being detected in the June–July time period. The algorithm selects the post-dormancy resumption of growth (Miller 1999) of the crop as the peak period, and in 2001 places the SOS in March (contemporaneous with this post-dormancy re-emergence) due to the true SOS taking place in 2000, before the initiation of the time series. (b) NDVI time series from a location in central Indiana, providing an example of a corn/soy mix, with the SOS being detected in April-May, and the EOS being detected in October.

Figure 3. The global median, (a) SOS date, and (b) EOS date, as observed between 2001 and 2010. The PTDs are natively at 0.5°, but the inclusions of even a single 250 m cropped pixel in any of those half-degree grid cells would lead to a large overestimation of cropland extent. In this global view, a cropland indicator mask (GLAM-UMD; unpublished) at 0.05° has been overlaid to provide a more realistic extent of cropland area. Some grid cells (shown in grey) had no detection over 2001–2010, despite having at least some cropland present. If greater than 50% of a grid cell's time series was missing due to insufficient observations (cloud cover presence, low quality observation) that grid was not processed for PTDs.

Figure 4. For the CONUS, the median (a) SOS date, and (b) EOS date, as observed between 2001 and 2010. The GSCs are shown at their native resolution (0.5°) without any post-processing application of a finer scale cropland mask to account for true cropland extent estimation (as in Figure 3(a) and 3(b)).

Figure 6. Maps of yearly GSC 0.5° grid cells correlated with different crop progress percentages from state-level USDA-NASS yearly crop progress data, both sets from 2001 to 2010. On the right is the maximum correlation value from the five crop progress percentage thresholds, and on the left is the crop progress threshold for which that maximum correlation value exists. (a) Yearly percent planted correlated with yearly SOS; (b) yearly percent emerged with yearly SOS; (c) yearly percent harvested with yearly EOS. As expected, SOS has higher correlation with emergence than it does with planting. EOS dates, with their consistently high correlation with harvesting dates, show particular sensitivity to interannual variability in phenology.

Figure 7. The median duration of the agricultural growing season (SOS to EOS from the same growing season), 2001–2010. As in Figure 3(a) and 3(b), the GLAM-UMD cropland indicator mask at 0.05° is overlaid the native 0.5° GSCs to more accurately represent cropland extent.

Figure 9. The earliest SOS date (2001–2010) minus a corn & soybean-only compilation of the Start of Planting period (based on maximum harvested area fraction, Monfreda, Ramankutty, and Foley (2008)) for the CONUS’ Corn Belt. A negative number indicates an earlier GSC SOS date relative to the start of the planting period, while a positive number indicates a later GSC SOS date versus the start of the planting period.

Miller, Travis D. 1999. “Growth Stages of Wheat: Identification and Understanding Improve Crop Management. Texas Agricultural Extension Service, the Texas A&M University System”. SCS-1999-16.

 Google Scholar

Monfreda, Chad, Navin Ramankutty, and Jonathan A. Foley. 2008. “Farming the Planet: 2. Geographic Distribution of Crop Areas, Yields, Physiological Types, and Net Primary Production in the Year 2000.” Global Biogeochemical Cycles 22 (1): GB1022. doi:10.1029/2007GB002947.

 Web of Science ®Google Scholar