A deep‐learning‐based experiment for benchmarking the performance of global terrestrial vegetation phenology models

Zhou, Xuewen; Xin, Qinchuan; Dai, Yongjiu; Li, Wanjing

doi:10.1111/geb.13374

Cited by 11 publications

(5 citation statements)

References 63 publications

(78 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Only recently a study was published where a one-dimensional convolutional neural network regression (1D-CNNR) model was developed to model global vegetation phenology using meteorological variables and satellite images as inputs. This research by (Zhou et al, 2021 ) demonstrates that the 1D-CNNR model has the potential for large-scale modeling of vegetation phenology. Future research should integrate deep learning techniques even more into phenology modeling.…”

Section: Research Trends and Future Directionmentioning

confidence: 93%

“…The idea of these modeling approaches is to use external climate data as input to predict the timing of key phenology metrics (Zhao et al, 2013 ; Hufkens et al, 2018 ). So far, phenology in the land surface models or dynamic global vegetation models (e.g., Biome-BGC (BioGeochemical Cycles) model, Lund-Potsdam-Jena model) generally adopt simple rule-based functions to account for the impacts of meteorological drivers, which can lead to large uncertainties in the modeled terrestrial ecosystem processes (Zhou et al, 2021 ). With the increase in automatically generated data (e.g., digital repeated photography in PhenoCam networks or satellite images), the amount of data available is constantly growing.…”

Section: Research Trends and Future Directionmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning in Plant Phenological Research: A Systematic Literature Review

et al. 2022

View full text Add to dashboard Cite

Climate change represents one of the most critical threats to biodiversity with far-reaching consequences for species interactions, the functioning of ecosystems, or the assembly of biotic communities. Plant phenology research has gained increasing attention as the timing of periodic events in plants is strongly affected by seasonal and interannual climate variation. Recent technological development allowed us to gather invaluable data at a variety of spatial and ecological scales. The feasibility of phenological monitoring today and in the future depends heavily on developing tools capable of efficiently analyzing these enormous amounts of data. Deep Neural Networks learn representations from data with impressive accuracy and lead to significant breakthroughs in, e.g., image processing. This article is the first systematic literature review aiming to thoroughly analyze all primary studies on deep learning approaches in plant phenology research. In a multi-stage process, we selected 24 peer-reviewed studies published in the last five years (2016–2021). After carefully analyzing these studies, we describe the applied methods categorized according to the studied phenological stages, vegetation type, spatial scale, data acquisition- and deep learning methods. Furthermore, we identify and discuss research trends and highlight promising future directions. We present a systematic overview of previously applied methods on different tasks that can guide this emerging complex research field.

show abstract

Section: Research Trends and Future Directionmentioning

confidence: 93%

Section: Research Trends and Future Directionmentioning

confidence: 99%

Deep Learning in Plant Phenological Research: A Systematic Literature Review

et al. 2022

View full text Add to dashboard Cite

show abstract

“…Thus, there may be concern whether the performance of VGD-LSTM was exaggerated due to model overfitting. Notably, some recent phenology studies adopted the spatial sampling strategy to calibrate and validate phenology models (e.g., [39]). They selected different pixels with spatial random sampling or uniform sampling to generate the training, validation, and testing sets; however, the selected pixels in the three sets may not be independent among each other owing to spatial autocorrelations, resulting in an overestimation of phenology model performance.…”

Section: A Performances Of Vgd Simulations By Vgd-lstmmentioning

confidence: 99%

“…Therefore, DL may be a promising tool to model the complex interactions of preseason temperature and precipitation on VGD. Zhou et al [39] simulated global vegetation phenology during 2001-2015 by performing a DL-based experiment. Their one-dimensional convolutional neural network regression model captures spatial patterns of VGD well, whereas the simulation of the interannual changes in VGD, even for simulation assessment at the hemispheric scale averaging over all pixels, seems to be less satisfactory (quantitative indices were not given in their Figure S11).…”

Section: Introductionmentioning

confidence: 99%

Remotely Sensed Vegetation Green-Up Onset Date on the Tibetan Plateau: Simulations and Future Predictions

Cao,

Ling,

Liu

et al. 2023

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Vegetation green-up onset date (VGD) is a key indicator of ecosystem structure and processes. As the largest and highest alpine ecoregion, the Tibetan Plateau (TP) has experienced markable climate warming during the past decades and showed substantial changes in VGD. However, the existing process-based phenology models still cannot simulate interannual variations in satellite-derived VGD. In this study, we developed a data-driven VGD model for the TP based on the Long Short-Term Memory neural network (called as VGD-LSTM). VGD-LSTM considers the complicated nonlinear relationship between VGD and multiple climatic and environmental drivers, including the time series of temperature (daytime, daily minimum, and daily mean) and precipitation, as well as non-sequential variables (elevation and geolocation). Compared with the benchmark process-based VGD model for the TP (i.e., the hierarchical model), VGD-LSTM greatly improved the simulation of interannual VGD variations. We calculated the correlation coefficients (R) between satellite-derived VGDs and VGD simulations during 2000-2018; the percentages of pixels with R values above 0.5 are increased from 15% for the hierarchical model to 41% for VGD-LSTM. The advanced trend in the satellite-derived VGD on the entire TP during 2000-2018 (-0.37 day/year) was captured well by VGD-LSTM (-0.33 day/year) but was underestimated by the hierarchical model (-0.08 day/year). According to VGD-LSTM simulations, VGDs on the TP are projected to advance by 8-10 days by 2100 relative to 2015-2020 under high Shared Socioeconomic Pathway scenarios. This study suggests the potential of artificial intelligence in phenology modelling for which the physiological processes are difficult to be fully represented.

show abstract

“…Land surface models (LSMs), such as ORCHIDEE (Krinner et al., 2005), serve as robust tools for studying the global carbon cycle. A precise representation of the leaf phenology and carbon uptake seasonality is mandatory to determine the carbon response to past and future climate change (Caen et al., 2022; Fu et al., 2013; Koven et al., 2011; Peñuelas et al., 2009; Sitch et al., 2003; Wright et al., 2017; Zhou et al., 2021). However, the standard version of ORCHIDEE poorly represents the leaf phenology of tropical rainforests since it lacks appropriate mechanisms that relate phenological signals to climate drivers (de Weirdt et al., 2012; Manoli et al., 2018).…”

Section: Introductionmentioning

confidence: 99%

Solar Radiation Triggers the Bimodal Leaf Phenology of Central African Evergreen Broadleaved Forests

Liu,

Ciais,

Maignan

et al. 2024

J Adv Model Earth Syst

View full text Add to dashboard Cite

Central African evergreen broadleaved forests around the equator exhibit a double annual cycle for canopy phenology and carbon uptake seasonality. The underlying drivers of this behavior are poorly understood and the double seasonality is not captured by land surface models (LSM). In this study, we developed a new leaf phenology module into the ORCHIDEE LSM (hereafter ORCHIDEE‐AFP), which utilizes short‐wave incoming radiation (SWd) as the main driver of leaf shedding and partial rejuvenation of the canopy, to simulate the double seasonality of central African forests. The ORCHIDEE‐AFP model has been evaluated by using field data from two forest sites and satellite observations of the enhanced vegetation index (EVI), which is a proxy of young leaf area index (LAIYoung) with leafage less than 6 months, as well as six products of GPP or GPP proxies. Results demonstrate that ORCHIDEE‐AFP successfully reproduces observed leaf turnover (R = 0.45) and young leaf abundance (R = 0.74), and greatly improve the representation of the bimodal leaf phenology. The proportion of grid cells with a significant positive correlation between the seasonality of modeled LAIYoung and observed EVI increased from 0.2% in the standard model to 27% in the new model. For photosynthesis, the proportions of grid cells with significant positive correlations between modeled and observed seasonality range from 26% to 65% across the six GPP evaluation products. The improved performance of the ORCHIDEE‐AFP model in simulating leaf phenology and photosynthesis of central African forests will allow a more accurate assessment of the impacts of climate change in tropical forests.

show abstract

A deep‐learning‐based experiment for benchmarking the performance of global terrestrial vegetation phenology models

Cited by 11 publications

References 63 publications

Deep Learning in Plant Phenological Research: A Systematic Literature Review

Deep Learning in Plant Phenological Research: A Systematic Literature Review

Remotely Sensed Vegetation Green-Up Onset Date on the Tibetan Plateau: Simulations and Future Predictions

Solar Radiation Triggers the Bimodal Leaf Phenology of Central African Evergreen Broadleaved Forests

Contact Info

Product

Resources

About