Remote sensing of crop lodging: A multi-sensor approach
Due to the COVID-19 crisis measures the PhD defence of Sugandh Chauhan will take place (partly) online in the presence of an invited audience.
The PhD defence can be followed by a live stream.
Sugandh Chauhan is a PhD student in the research group Natural Resources (NRS). Her supervisor is prof.dr. A.D. Nelson from the Faculty of Geo-Information Science and Earth Observation (ITC).
Crop lodging: the permanent bending of the crop stem from its vertical position (stem lodging) or displacement of root anchorage (root lodging), is a major yield-reducing factor in cereal crops such as wheat. It can also delay harvest, increase drying costs and deteriorate grain quality in cereal crops, thus affecting the likelihood of achieving a premium price. Therefore, accurate spatio-temporal information about crop lodging and its susceptibility is critical for improving yield estimates, informing insurance loss adjusters and facilitate decision making.
The conventional measures to assess lodging are primarily based on visual inspection of crop health and the use of mechanistic mathematical models which can be time-consuming and challenging to apply over large areas. Remote sensing (RS) data can be a valuable supplement or even replacement to these conventional methods, delivering spatial and temporal information about crop lodging over synoptic scales. However, the use of RS for crop lodging assessment is still in a nascent stage. An understanding of the RS-based metrics derived from the satellite data and their utility for lodging detection, characterisation and susceptibility analysis was lacking in the literature. In this context, this PhD study aimed to address the problem of lodging assessment using RS satellite data from different sensors, including Sentinel-1, Sentinel-2 and multi-incidence angle RADARSAT-2 data.
We defined five objectives that aimed at investigating the potential of spaceborne RS data for lodging detection, its characterisation and susceptibility mapping in wheat. The first objective was to carry out a systematic literature review that could relate field/lab-based approaches to RS-based methods, review and identify the research gaps in existing RS-based crop lodging studies and provide perspectives for future research. Our review found only 22 peer-reviewed articles published between 1951-2018, most of which focused on qualitative analysis of lodging. The review also enabled us to identify several unanswered research questions. Buildings upon our findings from the review, the subsequent objectives characterized lodging in three ways: detecting lodging stages, classifying lodging severity and identifying the time of lodging incidence. The final objective dealt with susceptibility analysis. The second objective investigated the use of Sentinel-1, low incidence angle RADARSAT-2 and high incidence angle RADARSAT-2 data for estimating crop angle of inclination as an indicator of lodging stage (moderate, severe and very severe). Our results demonstrated the higher sensitivity of low incidence angle RADARSAT-2 data (R2CV= 0.87) for estimating crop angle of inclination and highlighted the importance of Sentinel-1 data for operational assessment of crop lodging stages. The third objective presented a SAR-based approach for the classification of lodging severity based on lodging score. We found that lodging severity was best classified using low incidence angle RADARSAT-2 (overall accuracy 72%) while the model developed using Sentinel-1 data could identify 60% of the lodging severity cases in the study site. The next objective examined the utility of dense time-series Sentinel-1 data in combination with multi-spectral Sentinel-2 data for identifying the time of lodging incidence in wheat. It also evaluated the effect of lodging on backscatter/coherence and spectral reflectance response. Our results showed that with the temporal analysis, it was possible to indicate a plausible window of the main lodging event and the red edge (740nm), NIR (865nm) and VH backscatter could best distinguish between healthy from lodged wheat. The last objective investigated the role of SAR data for estimating a safety factor against root lodging as an indicator of lodging susceptibility in wheat. We found that the safety factor correlated well with the lodging observed in the fields and was detectable using the satellite data (with 73-84% accuracy), confirming that it could be used as an early indicator of lodging susceptibility.
