Using PhenoCams to track crop phenology and explain the effects of different cropping systems on yield

Liu, Yujie; Bachofen, Christoph; Wittwer, Raphaël; Duarte, Gicele Silva; Sun, Qing; Klaus, Valentin H.; Buchmann, Nina

doi:10.1016/j.agsy.2021.103306

Cited by 16 publications

(12 citation statements)

References 84 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These image-based metrics of crop type and stage from PhenoCam time series can complement Gcc as opposed to replacing it though. For example, Gcc, and other greenness metrics, can be used to derive the date of peak greenness or the rate of greenup or greendown, which reflects important plant properties not present in our image classification approach such as water stress and plant vigor [6,26,48]. Added insight from complementary metrics enrich interpretation and offer decision-makers flexibility in crop management [9].…”

Section: Discussionmentioning

confidence: 99%

“…Nearsurface cameras used in the PhenoCam network offer a novel solution to this need for local-scale crop management information. Images from near-surface cameras can document the date of emergence, maturity, harvest, and tillage at the field scale with a daily temporal resolution [5,6].…”

Section: Introductionmentioning

confidence: 99%

“…In addition to producing widespread ground truth data for large-scale models, automated classification of crop phenology would be beneficial in local-scale experiments of cropping practices. Recent studies have shown phenology from PhenoCams enables high-throughput field phenotyping and tracking a variety of crop characteristics [6,26].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Classification of Daily Crop Phenology in PhenoCams Using Deep Learning and Hidden Markov Models

Taylor

Browning

2022

Remote Sensing

View full text Add to dashboard Cite

Near-surface cameras, such as those in the PhenoCam network, are a common source of ground truth data in modelling and remote sensing studies. Despite having locations across numerous agricultural sites, few studies have used near-surface cameras to track the unique phenology of croplands. Due to management activities, crops do not have a natural vegetation cycle which many phenological extraction methods are based on. For example, a field may experience abrupt changes due to harvesting and tillage throughout the year. A single camera can also record several different plants due to crop rotations, fallow fields, and cover crops. Current methods to estimate phenology metrics from image time series compress all image information into a relative greenness metric, which discards a large amount of contextual information. This can include the type of crop present, whether snow or water is present on the field, the crop phenology, or whether a field lacking green plants consists of bare soil, fully senesced plants, or plant residue. Here, we developed a modelling workflow to create a daily time series of crop type and phenology, while also accounting for other factors such as obstructed images and snow covered fields. We used a mainstream deep learning image classification model, VGG16. Deep learning classification models do not have a temporal component, so to account for temporal correlation among images, our workflow incorporates a hidden Markov model in the post-processing. The initial image classification model had out of sample F1 scores of 0.83–0.85, which improved to 0.86–0.91 after all post-processing steps. The resulting time series show the progression of crops from emergence to harvest, and can serve as a daily, local-scale dataset of field states and phenological stages for agricultural research.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Classification of Daily Crop Phenology in PhenoCams Using Deep Learning and Hidden Markov Models

Taylor

Browning

2022

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…The monitoring of phenological development as well as the detection of specific phenological stages of agricultural plants using remote sensing data is widely used [10]. Additionally, time-lapse cameras installed close to the surface (PhenoCams) are used to track phenology at very high temporal resolutions (e.g., at a daily rate) but are mostly limited to the field scale [11,12]. The great advantage of remote sensing data is their ability to cover large areas in regularly time steps.…”

Section: Introductionmentioning

confidence: 99%

Detecting Phenological Development of Winter Wheat and Winter Barley Using Time Series of Sentinel-1 and Sentinel-2

et al. 2021

View full text Add to dashboard Cite

Monitoring the phenological development of agricultural plants is of high importance for farmers to adapt their management strategies and estimate yields. The aim of this study is to analyze the sensitivity of remote sensing features to phenological development of winter wheat and winter barley and to test their transferability in two test sites in Northeast Germany and in two years. Local minima, local maxima and breakpoints of smoothed time series of synthetic aperture radar (SAR) data of the Sentinel-1 VH (vertical-horizontal) and VV (vertical-vertical) intensities and their ratio VH/VV; of the polarimetric features entropy, anisotropy and alpha derived from polarimetric decomposition; as well as of the vegetation index NDVI (Normalized Difference Vegetation Index) calculated using optical data of Sentinel-2 are compared with entry dates of phenological stages. The beginning of stem elongation produces a breakpoint in the time series of most parameters for wheat and barley. Furthermore, the beginning of heading could be detected by all parameters, whereas particularly a local minimum of VH and VV backscatter is observed less then 5 days before the entry date. The medium milk stage can not be detected reliably, whereas the hard dough stage of barley takes place approximately 6–8 days around a local maximum of VH backscatter in 2018. Harvest is detected for barley using the fourth breakpoint of most parameters. The study shows that backscatter and polarimetric parameters as well as the NDVI are sensitive to specific phenological developments. The transferability of the approach is demonstrated, whereas differences between test sites and years are mainly caused by meteorological differences.

show abstract

“…Near-surface cameras used in the PhenoCam network offer a novel solution to this need for local-scale crop management information. Images from near-surface cameras can document the date of emergence, maturity, harvest, and tillage at the field scale with a daily temporal resolution [5,6].…”

Section: Introductionmentioning

confidence: 99%

Classification of daily crop phenology in PhenoCams using deep learning and hidden markov models

Taylor

Browning

2021

Preprint

View full text Add to dashboard Cite

Near surface cameras, such as those in the PhenoCam network, are a common source of ground truth data in modelling and remote sensing studies. Despite having locations across numerous agricultural sites, few studies have used near surface cameras to track the unique phenology of croplands. Due to management activities, crops do not have a natural vegetation cycle which many phenological extraction methods are based on. For example, a field may experience abrupt changes due to harvesting and tillage throughout the year. A single camera can also record several different plants due to crop rotations, fallow fields, and cover crops. Current methods to estimate phenology metrics from image time series compress all image information into a relative greenness metric, which discards a large amount of contextual information. This can include the type of crop present, whether snow or water is present on the field, the crop phenology, or whether a field lacking green plants consists of bare soil, fully senesced plants, or plant residue. Here we developed a modelling workflow to create a daily time series of crop type and phenology, while also accounting for other factors such as obstructed images and snow covered fields. We used a mainstream deep learning image classification model, VGG16. Deep learning classification models do not have a temporal component, so to account for temporal correlation among images our workflow incorporates a hidden markov model in the post-processing. The initial image classification model had out of sample F1 scores of 0.83-0.85, which improved to 0.86-0.91 after all post-processing steps. The resulting time series show the progression of crops from emergence to harvest, and can serve as a daily, local scale dataset of field states and phenological stages for agricultural research.

show abstract

Using PhenoCams to track crop phenology and explain the effects of different cropping systems on yield

Cited by 16 publications

References 84 publications

Classification of Daily Crop Phenology in PhenoCams Using Deep Learning and Hidden Markov Models

Classification of Daily Crop Phenology in PhenoCams Using Deep Learning and Hidden Markov Models

Detecting Phenological Development of Winter Wheat and Winter Barley Using Time Series of Sentinel-1 and Sentinel-2

Classification of daily crop phenology in PhenoCams using deep learning and hidden markov models

Contact Info

Product

Resources

About