Multimodal Estimation of Wheat Phenology using Remote Sensing Imagery
Crop phenology, the study of recurring biological events in plants and their relationship to environmental factors, is an important aspect of agricultural management and research. Phenological stages (germination, tillering, flowering, grain filling) provide key insights into plant development and can significantly influence crop management practices, resource allocation, and yield forecasting. Accurate phenological monitoring is essential for optimizing the timing of irrigation, fertilization, and pest control, to ensure maximum crop productivity. Traditionally, phenological observations have been conducted manually through field surveys, which are labor-intensive, time-consuming, and prone to human error. However, with the advent of advanced remote sensing technologies, it is now possible to automate and enhance the precision of phenological monitoring through the use of Unmanned Aerial Vehicles (UAVs) equipped with multimodal sensors.
UAV platforms equipped with RGB, multispectral, and thermal infrared sensors, offer a powerful means of capturing detailed spatial and temporal information about crop growth and development. RGB imagery provides high-resolution visual information, capturing the color and structure of the crop canopy, which can be used to assess phenological stages based on visual cues (leaf color and flowering patterns). Multispectral sensors capture data across multiple wavelengths of light, including visible and near-infrared (NIR) bands, which are useful for assessing vegetation indices (Normalized Difference Vegetation Index (NDVI), Normalized Difference Red Edge Index (NDRE), and many more). These indices are critical for monitoring vegetation health, vigor, and stress levels, all of which are closely tied to phenological stages.
Thermal infrared sensors, on the other hand, measure the surface temperature of the crop canopy, providing insights into plant water status, evapotranspiration rates, and thermal stress, which can also be indicative of specific phenological stages. By integrating data from these various modalities, it becomes possible to develop a comprehensive understanding of crop phenology, allowing for more accurate and timely predictions.
Multimodal approach to phenological estimation leverages the strengths of each sensor type to capture different aspects of crop development. Integration of multiple data sources allows for the construction of robust phenological models that can accurately track the progression of crop stages over time, even in the presence of environmental variability. Additionally, by correlating phenological data with temperature, precipitation, and soil moisture, it is possible to refine these models further, enabling more precise predictions under different growing conditions.
We also use the collected data to assess a range of other crop traits, including structural, physiological, and morphological characteristics, which are essential for in-season yield estimation. Structural traits (plant height, canopy cover, biomass) can be derived from RGB and multispectral imagery through photogrammetry techniques and spectral analysis. Physiological traits (chlorophyll content, photosynthetic efficiency, water-use efficiency) are inferred from multispectral and thermal data, providing insights into the plant’s overall health and productivity. By combining these traits with crop phenological performance, it is possible to develop predictive models that estimate crop yield early in the growing season.
Faculty
Students
Collaborators
Publications
- D. Wasif, M. Q. Khan, R. Murtaza, M. Z. Ahmad, Z. Zafar, M. Shahzad, K. Berns, M. M. Fraz. “Extraction of rice phenological metrics using temporally correlated multispectral drone imagery”, In 16th International Conference on Signal-Image Technology & Internet-Based Systems (SITIS) 163-169 (2022) https://doi.org/10.1109/SITIS57111.2022.00039
- V. Mehmood, A. I. Malik, Z. Zafar, M. Shahzad, K. Berns, M. M. Fraz. “Multi-year monitoring of wheat phenology and effect of climate change in the south Asian region using Sentinel-2 NDVI time series analysis”, In Image and Signal Processing for Remote Sensing XXIX 12733, 208-219 (2023) https://doi.org/10.1117/12.2683148
- U. Athar, M. Ali, Z. Zafar, Z. Mahmood, M. Fayyaz, K. Berns, M. M. Fraz. “Temporal Analysis of Phenological Development in Wheat Genotypes Using AI Enabled UAV Multispectral Data”, In 4th International Conference on Digital Futures and Transformative Technologies (2024)