Impacts of spatial and temporal resolution on remotely sensed corn and soybean emergence detection

Authors: Gao, Feng; Anderson, Martha; HOUBORG, RASMUS - Planet Labs Inc

Submitted to: Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/4/2024
Publication Date: 11/7/2024
Citation: Gao, F.N., Anderson, M.C., Houborg, R. 2024. Impacts of spatial and temporal resolution on remotely sensed corn and soybean emergence detection. Remote Sensing. 16(22). Article e4145. https://doi.org/10.3390/rs16224145.
DOI: https://doi.org/10.3390/rs16224145

Interpretive Summary: Crop emergence marks the beginning of crop vegetative growth. Accurately identifying when crops first emerge is vital for various agricultural tasks such as modeling growth, monitoring crop health, and estimating yields. Traditionally, this has been done by visually inspecting fields, which is time-consuming and limited in scope. Remote sensing offers a more efficient alternative by detecting changes in crop development over time. This study evaluated the performance of five remote sensing datasets in detecting crop emergence using the within-season emergence (WISE) method. The results demonstrate that Planet Fusion data outperformed the others by accurately capturing all crop emergences. The Harmonized Landsat and Sentinel-2 (HLS) time series also performed well, although slightly less accurately than Planet Fusion data. These findings provide valuable insights for utilizing remote sensing datasets in crop monitoring.

Technical Abstract: Crop emergence is critical for crop growth modeling, crop condition monitoring, and crop yield estimation. Ground collections of crop emergence dates are time-consuming and can only include limited fields. Remote sensing time series have been used to detect crop emergence. Previously, we developed a within-season emergence (WISE) algorithm, allowing us to detect crop emergence dates two weeks after the crop has emerged. This paper evaluates crop emergence detection using five remote sensing datasets (i.e., VENµS, Planet Fusion, Sentinel-2, Landsat, and Harmonized Landsat and Sentinel-2 (HLS)) with diverse spatial and temporal resolutions. The green-up dates from the remote sensing time series are assessed using ground emergence observations and planting records from the Beltsville Agricultural Research Center (BARC) fields in Maryland, USA. Specifically, we focus on the impacts of temporal and spatial resolutions in crop emergence mapping. Our results show that Planet Fusion captured all crop emergences and outperformed other datasets, with a mean difference of < 1 day and an absolute difference of < 5 days compared to ground-observed emergence dates from three years. The HLS and Sentinel-2 time series captured most crop emergences with close accuracy. Landsat detected less than half of crop emergences in recent years when both Landsat-8 and -9 were available. In our study area, temporal revisit plays a more crucial role in emergence detection than spatial resolution. Both Planet and HLS datasets are frequent enough for capturing crop emergence. Results reveal that the WISE algorithm suits different datasets even if their surface reflectance and vegetation indices differ. Aggregation methods based on mean surface reflectance, mean vegetation index, or the median of emergence dates produced similar results. Spatial resolutions from 5-30-m are suitable for field-level summaries in the study area. However, 30-m HLS lacked sub-field details in fields with mixed cropping systems. This study, assessing impacts of spatial and temporal resolution on crop emergence detection, will help to enhance future operational crop emergence mapping systems using satellite remote sensing.