Bin Chen

Dr. Bin Chen

Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, 11A Datun lu, Chaoyang district, Beijing
orcid

Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 October 2022 in Remote Sensing
Verified
Journal Article
Open access
PDF file
YLYuanyuan Liu
Yuanyuan Liu
SWShaoqiang Wang
Shaoqiang Wang
JCJinghua Chen
Jinghua Chen
BCBin Chen
Bin Chen
XWXiaobo Wang
Xiaobo Wang
LSLeigang Sun
Leigang Sun
As the second largest rice producer, India contributes about 20% of the world’s rice production. Timely, accurate, and reliable rice yield prediction in India is crucial for global food security and health issues. Deep learning models have achieved excellent performances in predicting crop yield. However, the interpretation of deep learning models is still rare. In this study, we proposed a transformer-based model, Informer, to predict rice yield across the Indian Indo-Gangetic Plains by integrating time-series satellite data, environmental variables, and rice yield records from 2001 to 2016. The results showed that Informer had better performance (R2 = 0.81, RMSE = 0.41 t/ha) than four other machine learning and deep learning models for end-of-season prediction. For within-season prediction, the Informer model could achieve stable performances (R2 ≈ 0.78) after late September, which indicated that the optimal prediction could be achieved 2 months before rice maturity. In addition, we interpreted the prediction models by evaluating the input feature importance and analyzing hidden features. The evaluation of feature importance indicated that NIRV was the most critical factor, while intervals 6 (mid-August) and 12 (mid-November) were the key periods for rice yield prediction. The hidden feature analysis demonstrated that the attention-based long short-term memory (AtLSTM) model accumulated the information of each growth period, while the Informer model focused on the information around intervals 5 to 6 (August) and 11 to 12 (November). Our findings provided a reliable and simple framework for crop yield prediction and a new perspective for explaining the internal mechanism of deep learning models.

20 February 2020 in Journal of Geophysical Research: Biogeosciences
Verified
Journal Article
Open access
PDF file
SWShaoqiang Wang
Shaoqiang Wang
YLYue Li
Yue Li
WJWeimin Ju
Weimin Ju
BCBin Chen
Bin Chen
JCJinghua Chen
Jinghua Chen
FYFengting Yang
Fengting Yang
Photosynthetic rate is a key source of uncertainty in the modeling of the terrestrial carbon cycle. Recent studies have utilized leaf chlorophyll content (Chl) as a proxy for leaf photosynthetic capacity in croplands and deciduous forests, with little investigation into this relationship for other plant function types and for different leaf ages. In this study, we evaluated the relationship between the maximum rate of carboxylation (Vcmax25) and the maximum electron transport capacity (Jcmax25) at 25 °C with both leaf nitrogen and Chl from different leaf ages (current and previous year) in Masson's pine (Pinus massoniana Lamb.) and slash pine (Pinus elliottii Engelm.) species in a subtropical evergreen coniferous forest. Our results showed small changes in leaf nitrogen over the growing season. In contrast, Vcmax25, Jmax25, and Chl displayed larger seasonal variations. Vcmax25 was more related to leaf Chl than leaf nitrogen in both previous year's and current year's leaves, likely due to the variable partitioning of leaf nitrogen between and within photosynthetic and nonphotosynthetic fractions. Leaf Chl and month after budding (MAB) were the main predictors for Vcmax25 based on the random forest regression analysis. These findings highlighted the problem in using leaf nitrogen as a proxy for Vcmax25 where there is a dynamic nitrogen investment (i.e., with leaf ontogenesis, or between different species) and illustrated the value of using leaf Chl (as retrievable from remotely sensing) and MAB to constrain Vcmax25 in process‐based models to improve the simulation of photosynthetic rates in evergreen coniferous forests.

1 October 2023 in Atmospheric Environment
Verified
Journal Article
Closed access
SCShiliang Chen
Shiliang Chen
BCBin Chen
Bin Chen
SWShaoqiang Wang
Shaoqiang Wang
LSLeigang Sun
Leigang Sun
HSHao Shi
Hao Shi
LWLizhe Wang
Lizhe Wang

18 February 2025 in Environmental Research Letters
Verified
Journal Article
Open access
TYTe Yang
Te Yang
BCBin Chen
Bin Chen
SWShaoqiang Wang
Shaoqiang Wang
XXXiyan Xu
Xiyan Xu
Paulo ArtaxoPaulo Artaxo
Paulo Artaxo
XYXin Yu
Xin Yu
The Amazon, the world’s largest tropical forest, plays a critical role in the global carbon cycle. It has a large carbon pool and acts as a major carbon sink. However, in 2023–2024, a compound heatwave-drought (CHWD) event (HD2023) hit the Amazon region, resulting in extreme temperatures and soil moisture deficits, threatening the region’s carbon sink capacity. Using advanced multisource satellite data and meteorological reanalysis, we quantified the impact of various climatic factors on vegetation productivity during HD2023 and analyzed its progression. Our findings showed that HD2023 led to a 530 Tg C decline in gross primary productivity (GPP) and 0.003 reduction in near-infrared reflectance of vegetation. The strongest phase of the event spanned 5 months, causing persistently high temperatures and reduced precipitation, leading to a continuous decline in soil moisture and marked reduction in GPP. The most severe decrease in GPP occurred in January 2024. The event originated in the northwest and gradually spread to the southeast. Soil moisture was the dominant factor in the decline of photosynthesis across vegetation types, whereas high solar radiation mitigated the impact of drought in evergreen broad forests and savannas. Moreover, the sensitivity of GPP to CHWD varied across vegetation types, ranking as grassland > savanna > evergreen broad forest. This study assessed the impact of HD2023 on regional carbon flux in the Amazon. As climate projections indicate future increases in climate extremes over the Amazon, it is important to identify the drivers of this impact on the carbon cycle of the Amazon.

18 November 2022 in Ecological Modelling
Verified
Journal Article
Closed access
BCBin Chen
Bin Chen
PWPengyuan Wang
Pengyuan Wang
SWShaoqiang Wang
Shaoqiang Wang
WJWeimin Ju
Weimin Ju
ZLZhenhai Liu
Zhenhai Liu
YZYinghui Zhang
Yinghui Zhang

26 October 2020 in Journal of Geophysical Research: Biogeosciences
Verified
Journal Article
Open access
PDF file
BCBin Chen
Bin Chen
MAM. Altaf Arain
M. Altaf Arain
JCJing Chen
Jing Chen
SWShaoqiang Wang
Shaoqiang Wang
HFHongliang Fang
Hongliang Fang
Jane LiuJane Liu
Jane Liu
Terrestrial gross primary productivity (GPP) quantifies the photosynthetic uptake by vegetation and it is the largest component of the terrestrial carbon cycle. The separation of sunlit and shaded leaves has been shown to be an effective leaf‐to‐canopy upscaling method for modeling vegetation GPP. In this study, the performance of the Integrated Carbon‐Canadian Land Surface Scheme (IC‐CLASS) based on Farquhar's photosynthetic model and the two‐leaf approach was compared against that of Moderate Resolution Imaging Spectroradiometer (MODIS) GPP algorithm using the light use efficiency (LUE) approach, validated with eddy covariance (EC) measured GPP over a variety of terrestrial ecosystems in Canada. There were systematic differences between the IC‐CLASS simulated GPP and the MODIS GPP product in spatial distribution patterns. The differences were due to inherent shortcomings of the LUE modeling approach where a constant maximum LUE value is specified for each biome type, ignoring the variation of shaded leaf contribution to total GPP. The IC‐CLASS model separates the sunlit and shaded leaves and the bias in simulating GPP was minimized. The IC‐CLASS performed better than the MODIS GPP algorithm compared with monthly and annual GPP derived from EC flux data at 13 Canadian Carbon Program sites. The differences between the IC‐CLASS and the MODIS GPP estimates were larger in more clumped canopies (i.e. forests), because of the increase in the shaded leaf fraction. Different LUEs in sunlit and shaded portions of the canopy should be considered for effective and reliable estimation of GPP at regional scale.