C6 Aqua 250-m eMODIS Remote Sensing Phenology Metrics across the conterminous U.S.

Phenological dynamics of terrestrial ecosystems reflect the response of the Earth's vegetation canopy to changes in climate and hydrology and are thus important to monitor operationally. Researchers at the U.S. Geological Survey (USGS), Earth Resources Observation and Science (EROS) Center have developed methods for documenting the seasonal dynamics of vegetation in an operational fashion from satellite time-series data. The phenological metrics data produced at USGS EROS provide indicators of key phenological events for the conterminous United States on a yearly basis based on Collect 6 Aqua eMODIS NDVI input data (for the 2003 - 2020 metrics). As the objective is to monitor the phenological dynamics of the vegetation canopy across large areas rather than specific plants over limited space, the use of satellite imagery provides the basis to measure large scale change at the ecosystem level. National Aeronautics and Space Administration Moderate Resolution Imaging Spectroradiometer (MODIS) sensor time series data are the source for this USGS phenological database. The readily available and consistently processed weekly eMODIS Normalized Difference Vegetation Index (NDVI) data are the key input for the phenological metrics data (Jenkerson et al. 2010). The NDVI is affected by a number of phenomena including cloud contamination, atmospheric perturbations, and variable viewing geometry of the sensor; all of which usually reduce the NDVI value. To minimize these effects on the NDVI value and at the same time to maximize the total number of observations per year, weekly composites of maximum NDVI value were used. However, composited data may still show lingering effects that tend to reduce the NDVI value and, more importantly, disturb the temporal profile of the vegetation signal. Spikes (mostly downward) are common phenomena in time-series NDVI data. These abrupt shifts in NDVI are frequently too short-lived to be a function of a real change in vegetation condition and can affect algorithms that are searching for increasing or decreasing trends representing real phenological shifts. Therefore, a temporal smoothing of the time series data is performed to reduce or eliminate spurious data spikes. For the purpose of extracting phenological metrics, a method of temporal smoothing of NDVI data was adopted that does not over-generalize the time-series profile, but eliminates spurious spikes in the NDVI while retaining sustained changes in NDVI that are representative of vegetation phenological dynamics. The weighted least-square approach for temporal smoothing (Swets et. al., 1999) was adopted for the conterminous U.S. NDVI time series to eliminate anomalously low vegetation index values and reduce time shifts caused by overgeneralization of the NDVI signal. This approach uses a moving temporal window to calculate a family of regression lines that are associated with each observation; the family of lines is then averaged at each point and interpolated between points to provide a continuous temporal NDVI signal. While interpolating values between points, a weighting factor is applied that favors peak (high value) points over valley points. Smoothed NDVI data were stacked in an ascending three year 156 NDVI composite file (52 NDVI composites per year). The three years include the previous year and the following year (e.g. 2020 phenology metrics included 2019, 2020, and 2021 smoothed NDVI). In instances where the full 52 composites are not achieved, an average for each remaining weekly composite from the processed year and two previous years are used to fill those composites in the latter year to reach 156 composites (to fill 2021, composites from years 2018, 2019, and 2020 were averaged). The smoothed NDVI data were subsequently ingested into a model developed in the Interactive Data Language (IDL) to quantify specific phenological events (see 1 - 9 below). The current suites of 250-m spatial resolution phenological metrics are as follows: 1. Start of Season Time (SOST): starting time of the onset of the growing season (in day of the year). 2. Start of Season NDVI (SOSN): NDVI value at the starting time of the onset of the growing season (unitless- based on NDVI units). 3. End of Season Time (EOST): ending time of the growing season (in day of the year). 4. End of Season NDVI (EOSN): NDVI value at the ending time of the growing season (unitless-based on NDVI units). 5. Maximum Time (MAXT): the day of the year when the NDVI reaches its maximum during the growing season (in day of the year). 6. Maximum NDVI (MAXN): the highest (or peak) value in NDVI observed in a growing season (unitless-based on NDVI units). 7. Duration (DUR): the length of the growing season-the time between the start of season and end of season (in number of days). 8. Amplitude (AMP): the difference between the Maximum NDVI and NDVI at the day of start of season (unitless-based on NDVI units). 9. Time Integrated NDVI (TIN): the cumulative value of NDVI from the start to the end of the growing season (unitless-based on accumulated NDVI units). For details about the algorithms and the data scaling for each of these seasonal phenological metrics, refer to the data creation process section of this metadata. References: Swets, D. L., Reed, B. C., Rowland, J. R., and S. E. Marko, 1999, "A Weighted Least-squares Approach to Temporal Smoothing of NDVI," In Proceedings of the 1999 ASPRS Annual Conference, from Image to Information, Portland, Oregon, May 17-21, 1999, Bethesda, Maryland, American Society for Photogrammetry and Remote Sensing, CD-ROM, 1 disc. Jenkerson, C. B., Maiersperger, T.K., Schmidt, G. L (2010) eMODIS - a user-friendly data source. U.S. Geological Survey (USGS) Open-File Report, Reston, VA, USA. 2010-1055.