Establishing PhenoCam-based monitoring of wheat and rice phenology in India with satellite and meteorological data

PhenoCams have become a common tool for vegetation phenology monitoring in North America and Europe, but their use in Indian agriculture remains very limited. Most crop phenology research in India still depends on satellite imagery or field surveys. These approaches provide valuable information, but they often lack high temporal resolution and cannot capture short-term changes in crop conditions. To address this gap, we evaluated the use of a PhenoCam to monitor wheat and rice phenology at a field site in Roorkee, India across two growing seasons. Our aim was to assess how high-frequency imagery can complement field observations, satellite, and meteorological data for crop phenology assessment.

We installed an infrared-enabled PhenoCam on a 6.5 m tower overlooking a winter wheat field during the rabi season (2023-24 and 2024-25) and a rice field during the kharif season (2024). Images were captured automatically at fixed intervals and were processed using the PhenoAI framework, which is a deep learning Python framework designed for automated time-series data processing. Greenness indices such as GCC and NDVI were derived from the processed images. For the wheat study, we conducted a cross-platform evaluation over two consecutive seasons (2023–2025) by combining PhenoCam data with in-situ observations, Sentinel-2, and PlanetScope imagery. PhenoCam achieved the highest timing agreement with field observations, with a mean absolute error (MAE) of 2.6–3.5 days. Sentinel-2 followed with MAE values of 2.4–4.2 days, while PlanetScope showed larger errors of 4.1–5.6 days due to radiometric noise and cloud cover. GCC was most sensitive to early green-up, whereas NDVI provided stable tracking of the full growth cycle (R² > 0.90).

During the rice season, we focused on how crop phenology responds to local weather conditions. We collected meteorological data from a co-located automated weather station. We examined climate–phenology relationships using a combination of exploratory correlations and mixed-effects model analysis. Minimum air temperature and PAR showed the strongest overall negative correlations with canopy greenness (r = −0.42 and r = −0.37). Stage-wise analysis indicated that tillering responded positively to temperature (r = 0.45), while booting and heading showed negative responses. A log response ratio (lnRR) meta-analysis identified flowering as the most climate-sensitive stage, with significant lnRR effects for 4 out of 8 climate variables, followed by tillering (3/8) and germination (2/8).

Overall, these results show that PhenoCam imagery can resolve inter-annual shifts in wheat phenology, identify climate-sensitive stages in rice, and validate satellite-derived phenology at daily scale. As one of the first agricultural PhenoCam deployments in India, this work demonstrates the value of near-surface imaging for bridging field and satellite observations. It reduces temporal gaps during cloudy conditions, provides ground reference for satellite calibration, and reveals stage-specific climate responses relevant for climate-resilient crop management in India.