Remote sensing captures varying temporal patterns of vegetation between human-altered and natural landscapes

Abstract

Global change has led to shifts in phenology, potentially disrupting species interactions such as plant-pollinator relationships. Advances in remote sensing techniques allow one to detect vegetation phenological diversity between different land use types, but it is not clear how this translates to other communities in the ecosystem. Here, we investigated the phenological diversity of the vegetation across a human-altered landscape including urban, agricultural, and natural land use types. We found that the patterns of change in the vegetation indices (EVI and NDVI) of human-altered landscapes are out of synchronization with the phenology in neighboring natural California grassland habitat. Comparing these findings to a spatio-temporal pollinator distribution dataset, EVI and NDVI were significant predictors of total bee abundance, a relationship that improved with time lags. This evidence supports the importance of differences in temporal dynamics between land use types. These findings also highlight the potential to utilize remote sensing data to make predictions for components of biodiversity that have tight vegetation associations, such as pollinators.

Keywords: Agricultural; Bees; EVI; Land use change; MODIS; NDVI; Remote sensing; Urban.

Figure 1. Maps illustrating the change in NDVI from March to September in East Contra Costa County, California.

A subset of bee collection sites (eight of each land use type) are marked on the map to illustrate the distribution of land use types. Triangles represent agricultural sites, squares represent urban sites, and circles represent natural sites. As the year progresses, higher NDVI values are associated with different land use types.

Figure 2. Different land use types exhibit different patterns of change throughout the year, from 2000 to 2014.

Each point represents the mean EVI for all pixels of the same land use classification within the 50 × 50 km region encapsulating the study site. Agricultural sites (A), have two peak EVI values, urban sites (B) remain relatively even, and natural sites (C) have one peak EVI value.

Figure 3. Standard deviation (A) and range (B) of EVI across different land use types vary significantly.

Agricultural sites (green) consistently have the highest standard deviations and ranges, natural (yellow) have the lowest, and urban is in between (red).

Figure 4. Relationship between total bee abundance and EVI across seasons.

In this example, total bee abundance is correlated with the EVI value from the same location 16 days before the closest day of collection. Each panel represents a different land use type. Different seasons (early spring, late spring, early summer, and later summer) are plotted with different symbols. A simple linear regression of all points within each panel is fit, but is only statistically significant for the natural land use type (p < 0.001).