scispace - formally typeset
Search or ask a question
Posted ContentDOI

A global land aerosol fine-mode fraction dataset (2001–2020) retrieved from MODIS using hybrid physical and deep learning approaches

TL;DR: Yan et al. as discussed by the authors developed a new satellite-based global land daily aerosol fine-mode fraction dataset (Phy-DL FMF) by synergizing the advantages of physical and deep learning methods at a 1° spatial resolution by covering the period from 2001 to 2020.
Abstract: . The aerosol fine-mode fraction (FMF) is potentially valuable for discriminating natural aerosols from anthropogenic ones. However, most current satellite-based FMF products are highly unreliable. Here, we developed a new satellite-based global land daily FMF dataset (Phy-DL FMF) by synergizing the advantages of physical and deep learning methods at a 1° spatial resolution by covering the period from 2001 to 2020. The Phy-DL FMF dataset is comparable to Aerosol Robotic Network (AERONET) measurements, based on the analysis of 361,089 data samples from 1170 AERONET sites around the world. Overall, Phy-DL FMF showed a root-mean-square error of 0.136 and correlation coefficient of 0.68, and the proportion of results that fell within the ±20 % expected error window was 79.15 %. Phy-DL FMF showed superior performance over alternate deep learning or physical approaches (such as the spectral deconvolution algorithm presented in our previous studies), particularly for forests, grasslands, croplands, and urban and barren land types. As a long-term dataset, Phy-DL FMF is able to show an overall significant decreasing trend (at a 95 % significance level) over global land areas. Based on the trend analysis of Phy-DL FMF for different countries, the upward trend in the FMFs was particularly strong over India and the western USA. Overall, this study provides a new FMF dataset for global land areas that can help improve our understanding of spatiotemporal fine- and coarse-mode aerosol changes. The datasets can be downloaded from https://doi.org/10.5281/zenodo.5105617 (Yan, 2021).

Summary (2 min read)

1 Introduction

  • POLDER ended its mission in 2013, whereas the Moderate Resolution Imaging Spectroradiometer 50 has operated for about 20 years and continues to perform well.
  • O pe n A cc es s Earth System Science Data.

2 Materials and methods

  • The MODIS sensor onboard Terra has provided long-term observations on a global scale every day since February 2000 (Levy et al., 2010), available at the Atmosphere Archive & Distribution System Distributed Active Archive Center.
  • Table S1 summarizes details about the MODIS data used in this study.

2.2 AERONET data

  • The AERONET is a worldwide, sun–sky photometer network providing ground-level aerosol properties, recently updated to Version 3 (Holben et al., 1998).
  • Since there is not enough Level 2.0 data for use as training data for modelling purposes, here, the authors used the Level 1.5 SDA FMF dataset generated from data from 1170 global AERONET sites covering the period of 2001 to 2020 as the ground truth for further modelling and validation .
  • These AERONET sites are spread around the world, enabling the construction of a universal model and allowing for a more thorough validation of the new FMF product.

2.3 Meteorological data

  • Due to the impact of meteorological factors on FMF (Yan et al., 2021a), five meteorological variables (i.e., 2-m air temperature, planetary boundary layer height, surface pressure, 10-m U/V wind components, and relative humidity) were obtained from ERA5 .
  • ERA5 is the fifth-generation product produced by the European Centre for Medium- Range Weather Forecasts, with hourly data available since 1950 and at a 0.25° spatial resolution.
  • Given the overpass time and 105 spatial resolution of MODIS data, only monitoring-time meteorological data collected from 10:00 to 11:00 local time were used and resampled to 1°×1° to obtain daily averages.

2.4 Combining physical and deep learning models (Phy-DL) for retrieving FMFs

  • This method follows the SDA to build a 110 LUT for the satellite-based FMF retrieval.
  • O pe n A cc es s Earth System Science Data.

2.5 Other global FMF products for comparison

  • Phy-DL-derived FMFs were compared with the following FMF products from three other satellite missions (Table S2): a. POLDER/GRASP FMF: Launched in December 2004, POLDER-3 onboard the Polarization and Anisotropy of Reflectances for Atmospheric Sciences coupled with Observations from a Lidar satellite was operational from March 2005 to October 2013, making multi-150 angular polarization measurements.
  • By capitalizing on the small and fairly neutral polarized reflectances (Deuze et al., 2001), POLDER/GRASP is able to provide the fine-mode AOD (fAOD, radius < 0.35 µm) in two categories: “high-precision” and “models”.
  • Because “high-precision” fAODs perform better than “models” fAODs (Wei et al., 2020), the authors used monthly “highprecision” POLDER/GRASP fAODs and AODs (both at 490 nm) at a spatial resolution of 1° for calculating FMF (at 490 nm) (FMF=fAOD/AOD).
  • O pe n A cc es s Earth System Science Data.

3.1 Phy-DL FMF validation

  • Figure 3 shows the validation of the Phy-DL FMFs against AERONET FMFs.
  • Figure 3b shows the biases of the Phy-DL FMFs (estimated FMF minus AERONET FMF) as a function of the AERONET FMFs.
  • Over 90% of Phy-DL FMFs fell within the ±20% EE envelope at some sites in the Amazon, southern 180 Africa, and southeast Asia.

3.3 Comparison between Phy-DL, DL-based, and Phy-based FMFs

  • To analyze the differences in FMFs obtained by different methods, FMFs generated by the Phy-DL method, deep learning (DL) method (meaning no Phy-based FMF as input), and Phy-based method (i.e., the LUT-SDA) from 2008 to 2017 275 were compared using AERONET FMFs as the ground truth.
  • O pe n A cc es s Earth System Science Data.

4. Data availability

  • The FMF data are in the Geotiff format on a daily scale.
  • This study developed an improved long-term global aerosol FMF (at 500 nm) dataset (2001–2020) over land by combining physical and deep learning approaches called Phy-DL FMF.
  • Geographically, Phy-DL FMFs captured the low FMFs well over the Saharan region, Central Asia, Australia, and southern South America, while Phy-based FMFs showed significant overestimations.
  • O pe n A cc es s Earth System Science Data.

Financial support.

  • This work was supported by the National Natural Science Foundation of China (42030606, 41801329 and 91837204), the National Key Research and Development Plan of China (2017YFC1501702), the Open Fund of State Key Laboratory of Remote Sensing Science (OFSLRSS201915) and the Fundamental Research Funds for the Central Universities.
  • O pe n A cc es s Earth System Science Data.

Did you find this useful? Give us your feedback

Figures (7)

Content maybe subject to copyright    Report

1
A global land aerosol fine-mode fraction dataset (20012020) retrieved from
MODIS using hybrid physical and deep learning approaches
Xing Yan
1*
, Zhou Zang
1
, Zhanqing Li
2*
, Nana Luo
3
, Chen Zuo
1
, Yize Jiang
1
, Dan Li
1
, Yushan Guo
1
,
Wenji Zhao
4
, Wenzhong Shi
5
, Maureen Cribb
2
1
State Key Laboratory of Remote Sensing Science, College of Global Change and Earth System Science, Beijing Normal
5
University, Beijing, 100875, China
2
Department of Atmospheric and Oceanic Science and ESSIC, University of Maryland, College Park, MD, 20740, USA
3
School of Geomatics and Urban Spatial Informatics, Beijing University of Civil Engineering and Architecture, Beijing 102612,
China
4
College of Resource Environment and Tourism, Capital Normal University, Beijing, China
10
5
Department of Land Surveying and Geo-Informatics, The Hong Kong Polytechnic University, Hong Kong, China
*Correspondence to: Xing Yan (yanxing@bnu.edu.cn); Zhanqing Li (zli@atmos.umd.edu)
15
Abstract. The aerosol fine-mode fraction (FMF) is potentially valuable for discriminating natural aerosols from anthropogenic
ones. However, most current satellite-based FMF products are highly unreliable. Here, we developed a new satellite-based
global land daily FMF dataset (Phy-DL FMF) by synergizing the advantages of physical and deep learning methods at a 1°
spatial resolution by covering the period from 2001 to 2020. The Phy-DL FMF dataset is comparable to Aerosol Robotic
20
Network (AERONET) measurements, based on the analysis of 361,089 data samples from 1170 AERONET sites around the
world. Overall, Phy-DL FMF showed a root-mean-square error of 0.136 and correlation coefficient of 0.68, and the proportion
of results that fell within the ±20% expected error window was 79.15%. Phy-DL FMF showed superior performance over
alternate deep learning or physical approaches (such as the spectral deconvolution algorithm presented in our previous studies),
particularly for forests, grasslands, croplands, and urban and barren land types. As a long-term dataset, Phy-DL FMF is able
25
to show an overall significant decreasing trend (at a 95% significance level) over global land areas. Based on the trend analysis
of Phy-DL FMF for different countries, the upward trend in the FMFs was particularly strong over India and the western USA.
Overall, this study provides a new FMF dataset for global land areas that can help improve our understanding of spatiotemporal
fine- and coarse-mode aerosol changes. The datasets can be downloaded from https://doi.org/10.5281/zenodo.5105617 (Yan,
2021).
30
https://doi.org/10.5194/essd-2021-326
Open Access
Earth System
Science
Data
Discussions
Preprint. Discussion started: 30 September 2021
c
Author(s) 2021. CC BY 4.0 License.

2
35
1 Introduction
Evaluating the impact of anthropogenic aerosols on climate change and human health relies on the ability to separate the
proportion of anthropogenic aerosols from the total aerosol loading (Anderson et al., 2005; Zheng et al., 2015). Although
satellite remote sensing can provide global-scale data on aerosol content that are represented by the aerosol optical depth
(AOD), accurate monitoring of anthropogenic aerosols is still a major challenge. This is because a key parameter called the
40
aerosol fine-mode fraction (FMF), which is used for discriminating anthropogenic aerosols from natural ones (Bellouin et al.,
2005), has been regarded as highly unreliable according to satellite-based AOD retrievals, especially over land (Levy et al.,
2013; Yan et al., 2017; Liang et al., 2021; Yang et al., 2020; Zang et al., 2021a).
Satellite-based FMF retrievals based on physical methods have been performed previously,; currently, five global-scale
FMF products exist (Figure 1) that exhibit different temporal resolutions from 1 to 16 days (Levy et al., 2007; Garay et al.,
45
2020; Chen et al., 2020a). Of these, POLarization and Directionality of the Earths Reflectances (POLDER) can perform multi-
angle and multi-spectral polarized measurements, which provide unique advantages in the retrieval of aerosol FMF (Dubovik
et al., 2011; Dubovik et al., 2019). Therefore, in recent years, several POLDER-based FMF retrieval methods have been
proposed (Zhang et al., 2016; Zhang et al., 2021), such as the generalized retrieval of aerosols and surface properties (Dubovik
et al., 2014). However, POLDER ended its mission in 2013, whereas the Moderate Resolution Imaging Spectroradiometer
50
(MODIS) has operated for about 20 years and continues to perform well. Currently, only the MODIS Dark Target (DT) method
has been used to generate global aerosol FMF products over both land and ocean. However, the MODIS DT-derived FMF
over land is highly unreliable and is not recommended for use even though it has evolved to the Collection 6.1 (C6.1) level
(Levy et al., 2013; Chen et al., 2020a). To improve the accuracy of MODIS land-based FMF retrievals, improvements have
been made to physical approaches, such as the Look-Up-Table-based Spectral Deconvolution Algorithm (LUT-SDA) (Yan et
55
al., 2017; Yan et al., 2019) and the Yonsei Aerosol Retrieval algorithm (Choi et al., 2016). Using the LUT-SDA model in
previous research, we developed a 10-year global land FMF dataset (Yan et al., 2021b) with moderately improved retrieval
accuracy (root-mean-square error, RMSE = 0.22). No multi-angle and multi-spectral polarized information, Lipponen et al.
(2018) noted that MODIS-based FMF retrievals using physical methods still suffer from these major limitations.
https://doi.org/10.5194/essd-2021-326
Open Access
Earth System
Science
Data
Discussions
Preprint. Discussion started: 30 September 2021
c
Author(s) 2021. CC BY 4.0 License.

3
60
Figure 1: Overview of the time periods covered by different satellites that provide global-scale FMF products. Acronyms used in this figure:
AATSR: Advanced Along-Track Scanning Radiometer; MISR: Multi-angle Imaging SpectroRadiometer; MODIS: Moderate Resolution
Imaging Spectroradiometer; POLDER: POLarization and Directionality of the Earth’s Reflectances; VIIRS: Visible Infrared Imaging
Radiometer Suite.
65
In recent years, deep learning approaches have been applied to satellite satellite-based atmospheric research (Zang et al.,
2021b; Yan et al., 2020a; Yuan et al., 2020; Shen et al., 2018; Ong et al., 2016), including FMF retrieval (X. Chen et al., 2020).
Compared with classical machine learning methods, deep learning is more capable of approximating nonlinear relationships
(Yan et al., 2021c). For example, X. Chen et al. (2020) used a convolutional neural network (CNN) to develop a deep learning
model for MODIS FMF retrievals called the Neural Network based AEROsol retrieval (NNAero) method. The NNAero-
70
derived FMF is a significant improvement over the MODIS DT-derived FMF, with the RMSE decreasing from 0.34 (DT) to
0.1567 (NNAero). However, this method has only been applied and validated over northern and eastern China, and not globally.
As an important limitation, Zhang et al. (2016) noted that satellite-measured multispectral reflectance of ground-based data
alone was not sufficient to retrieve FMFs with high accuracy. This limitation increases the difficulty of preparing training data
for the deep learning approach. The physical method may provide a means to alleviate this deficiency, raising the question of
75
whether combining the physical method and deep learning can improve the FMF retrieval accuracy.
To address the above issues, we synergize the advantages of the physical method and deep learning to retrieve aerosol
FMFs over land on a global scale using MODIS data. We tested and validated this hybrid model using two decades of data
(20012020) and produced a new long-term FMF dataset called Phy-DL FMF (physical-deep learning FMF). Contrary to
previous studies, the proposed hybrid model considers both physical characteristics and nonlinear relationships to constrain
80
the FMF calculation. This long-term dataset shows good promise for shedding light on the impacts of human activities on
atmospheric aerosols, providing a foundation for understanding the variations in fine mode aerosols on a global scale.
https://doi.org/10.5194/essd-2021-326
Open Access
Earth System
Science
Data
Discussions
Preprint. Discussion started: 30 September 2021
c
Author(s) 2021. CC BY 4.0 License.

4
2 Materials and methods
2.1 MODIS data
85
The MODIS sensor onboard Terra has provided long-term observations on a global scale every day since February 2000
(Levy et al., 2010), available at the Atmosphere Archive & Distribution System Distributed Active Archive Center. In this
study, MODIS C6.1 L1B MOD02SSH data (i.e., top-of-the-atmosphere (TOA) reflectances from Band 1 to Band 7), MODIS
C6.1 L3 MOD09CMG data (surface reflectances from Band 1 to Band 7), and MODIS C6.1 L3 MOD08 daily data were
obtained from 2001 to 2020 for retrieving FMFs. Table S1 summarizes details about the MODIS data used in this study.
90
2.2 AERONET data
The AERONET is a worldwide, sunsky photometer network providing ground-level aerosol properties, recently updated
to Version 3 (Holben et al., 1998). To retrieve FMFs from AERONET solar extinction data, (O'neill et al., 2001a; O'neill et
al., 2001b; O'neill et al., 2003) developed the Spectral Deconvolution Algorithm (SDA) method. The FMFs based on this
inversion method (i.e., SDA FMF) have been included in the standard AERONET data offering, with an estimated uncertainty
95
of 0.1 (O'neill et al., 2001b; O'neill et al., 2003). Since there is not enough Level 2.0 data for use as training data for modelling
purposes, here, we used the Level 1.5 SDA FMF dataset generated from data from 1170 global AERONET sites covering the
period of 2001 to 2020 as the ground truth for further modelling and validation (Figure S1a). These AERONET sites are spread
around the world, enabling the construction of a universal model and allowing for a more thorough validation of the new FMF
product.
100
2.3 Meteorological data
Due to the impact of meteorological factors on FMF (Yan et al., 2021a), five meteorological variables (i.e., 2-m air
temperature, planetary boundary layer height, surface pressure, 10-m U/V wind components, and relative humidity) were
obtained from ERA5 (Figure S1b-f). ERA5 is the fifth-generation product produced by the European Centre for Medium-
Range Weather Forecasts, with hourly data available since 1950 and at a 0.25° spatial resolution. Given the overpass time and
105
spatial resolution of MODIS data, only monitoring-time meteorological data collected from 10:00 to 11:00 local time were
used and resampled to 1°×1° to obtain daily averages.
2.4 Combining physical and deep learning models (Phy-DL) for retrieving FMFs
The physical model used in this research is the LUT-SDA (Yan et al., 2017). This method follows the SDA to build a
110
LUT for the satellite-based FMF retrieval. Based on Eq. (1), a set of hypothetical values for the Ångström exponent (AE)
https://doi.org/10.5194/essd-2021-326
Open Access
Earth System
Science
Data
Discussions
Preprint. Discussion started: 30 September 2021
c
Author(s) 2021. CC BY 4.0 License.

5
derivative (
), the AE (
), the AE of fine-mode AOD (
f
), and AOD (at 500 nm) are input to the SDA for the FMF
calculation (
), thus creating the LUT:
2 1/2
' ' ' '
1
{( b*) [( b*) 4c*(1 a)] }
2(1 - a)
(1)
cc
f c c c
cc
c
fc





115
where a, b*, c*,
, and
c
are fixed parameters described by O'neill (2010). In this study, the MODIS MOD08 DT-based
AOD and AE are used as input to the LUT-SDA for the global land physical-model-based FMF (Phy-based FMF) retrieval
(Yan et al., 2021b).
The deep learning model used in this study is called EntityDenseNet (Yan et al., 2020). The EntityDenseNet incorporates
the Entity Embeddings method (Guo and Berkhahn, 2016) that can directly process spatial or time-based features, such as
120
location, season, and month. It includes one input layer, two hidden layers, and one output layer. Each hidden layer has one
fully connected layer, one rectified linear unit (ReLU) layer, one batch normalization (BN) layer, and one dropout layer. The
feed-forward operation of each hidden layer can be written as
1 1 1
a { [W (a ) b ]} (2)
n n n n
BN f D

where n is the layer number,
a
n
is the output vector from layer n, D() is the dropout layer for the thinning vector
a
n
,
1
W
n
125
and
1
b
n
are weights and biases, respectively, at layer n+1,
[]f
is the ReLU activation function, and BN is the batch
normalization function.
In this study, we combine Phy-based FMF into EntityDenseNet along with satellite measurements and meteorological
data to reduce FMF retrieval biases. As shown by Yan et al. (2021b), the global land Phy-based FMF is still unreliable. Due
to its unknown and known error sources (e.g., MODIS-derived AE) and nonlinearity in the data itself, a linear model may not
130
be able to correct these errors. In addition, current physical retrieval methods do not use all the information provided by satellite
observations for aerosol size information retrievals (Zang et al., 2021). Lipponen et al. (2018) showed that satellite TOA
reflectance and geometry data can significantly improve the aerosol size data retrieval accuracy of the machine learning model.
Some studies have also suggested that surface reflectance and meteorological factors can also impact the FMF retrieval
accuracy (Yan et al., 2021a; X. Chen et al., 2020). Thus, besides Phy-based FMF, we input MODIS TOA reflectance data,
135
geometry data, surface reflectance, and meteorological data into EntityDenseNet for the final Phy-DL FMF calculation (Table
S1). In the deep learning model training process, 70%, 20%, and 10% of all input data are randomly separated into groups of
data for training, validation, and testing, respectively. The validation data are used for the hyperparameter optimization (node
numbers and dropout rate in each hidden layer) of the deep learning model. The testing data are used to evaluate the
performance of the trained deep learning model. When the trained model is finally optimized by the validation and testing data,
140
we apply this trained deep learning model to reconstruct global land FMFs for the period of 2001 to 2020.
https://doi.org/10.5194/essd-2021-326
Open Access
Earth System
Science
Data
Discussions
Preprint. Discussion started: 30 September 2021
c
Author(s) 2021. CC BY 4.0 License.

Citations
More filters
07 Jul 2014
TL;DR: In this paper, an analysis of satellite data and global climate model simulations suggests that dust aerosol levels over the Arabian Sea, West Asia and the Arabian Peninsula are positively correlated with the intensity of the Indian summer monsoon.
Abstract: The Indian summer monsoon is influenced by numerous factors, including aerosol-induced changes to clouds, surface and atmospheric heating, and atmospheric circulation. An analysis of satellite data and global climate model simulations suggests that dust aerosol levels over the Arabian Sea, West Asia and the Arabian Peninsula are positively correlated with the intensity of the Indian summer monsoon.

265 citations

Posted ContentDOI
09 Mar 2020
TL;DR: Zheng et al. as discussed by the authors used a multi-species atmospheric Bayesian inversion approach to attribute satellite-observed atmospheric CO variations to its sources and sinks in order to achieve a full closure of the global CO budget during 2000-2017.
Abstract: . Atmospheric carbon monoxide (CO) concentrations have been decreasing since 2000, as observed by both satellite- and ground-based instruments, but global bottom-up emission inventories estimate increasing anthropogenic CO emissions concurrently. In this study, we use a multi-species atmospheric Bayesian inversion approach to attribute satellite-observed atmospheric CO variations to its sources and sinks in order to achieve a full closure of the global CO budget during 2000–2017. Our observation constraints include satellite retrievals of the total column mole fraction of CO, formaldehyde (HCHO), and methane ( CH4 ) that are all major components of the atmospheric CO cycle. Three inversions (i.e., 2000–2017, 2005–2017, and 2010–2017) are performed to use the observation data to the maximum extent possible as they become available and assess the consistency of inversion results to the assimilation of more trace gas species. We identify a declining trend in the global CO budget since 2000 (three inversions are broadly consistent during overlapping periods), driven by reduced anthropogenic emissions in the US and Europe (both likely from the transport sector), and in China (likely from industry and residential sectors), as well as by reduced biomass burning emissions globally, especially in equatorial Africa (associated with reduced burned areas). We show that the trends and drivers of the inversion-based CO budget are not affected by the inter-annual variation assumed for prior CO fluxes. All three inversions contradict the global bottom-up inventories in the world's top two emitters: for the sign of anthropogenic emission trends in China (e.g., here - 0.8 ± 0.5 % yr −1 since 2000, while the prior gives 1.3±0.4 % yr −1 ) and for the rate of anthropogenic emission increase in South Asia (e.g., here 1.0±0.6 % yr −1 since 2000, smaller than 3.5±0.4 % yr −1 in the prior inventory). The posterior model CO concentrations and trends agree well with independent ground-based observations and correct the prior model bias. The comparison of the three inversions with different observation constraints further suggests that the most complete constrained inversion that assimilates CO, HCHO, and CH4 has a good representation of the global CO budget, and therefore matches best with independent observations, while the inversion only assimilating CO tends to underestimate both the decrease in anthropogenic CO emissions and the increase in the CO chemical production. The global CO budget data from all three inversions in this study can be accessed from https://doi.org/10.6084/m9.figshare.c.4454453.v1 (Zheng et al., 2019).

12 citations

Journal ArticleDOI
TL;DR: In this paper , seasonal characteristics and long-term variations in aerosol optical parameters in Hong Kong from 2006 to 2021 using AERONET data and satellite-based observations based on the extreme-point symmetric mode decomposition (ESMD) model.
Abstract: This study analyzes seasonal characteristics and long-term variations in aerosol optical parameters in Hong Kong from 2006 to 2021 using AERONET data and satellite-based observations based on the extreme-point symmetric mode decomposition (ESMD) model. The dominant aerosol types in Hong Kong are mixed aerosols and urban/industrial aerosols with fine-mode sizes, and slightly absorbing or non-absorbing properties. Aerosol optical depth (AOD), Angstrom exponent (AE) and single scattering albedo (SSA) varied seasonally with a lower AOD but higher AE and SSA in summer, and elevated AOD but lower AE and SSA in spring and winter. The long-term variations show the year 2012 to be a turning point, with an upward trend in AOD and AE before 2012 and then downwards after 2012. However, for SSA, a rising trend was exhibited in both pre- and post-2012 periods, but with a larger gradient in the first period. The ESMD analysis shows shorter-term, non-linear fluctuations in aerosol optical parameters, with alternating increasing and declining trends. The examination of the relationships between AOD and meteorological factors based on the extreme gradient boosting (XGBoost) method shows that the effects of weather conditions on AOD are complex and non-monotonic. A lower relative humidity, higher wind speed in southwest directions and lower temperature are beneficial to the abatement of aerosol loads in Hong Kong. In conclusion, the findings of this study enhance the understanding of aerosol properties and the interactions between aerosol loading and meteorological factors.

3 citations

TL;DR: Zhang et al. as discussed by the authors used the bagging trees ensemble model, based on 1 km aerosol 18 optical depth (AOD) data and multiple environmental covariates, to produce monthly FEC AOD products in the arid and semi-arid areas.
Abstract: : 13 Aerosols are a complex compound with a great effect on the global radiation 14 balance and climate system even human health, and concurrently are a large uncertain 15 source in the numerical simulation process. The arid and semi-arid area has a fragile 16 ecosystem, with abundant dust, but lacks related aerosol data or data accuracy. To solve 17 these problems, we use the bagging trees ensemble model, based on 1 km aerosol 18 optical depth (AOD) data and multiple environmental covariates, to produce monthly 19 advanced-performance, full-coverage, and high-resolution (250 m) AOD products 20 (named FEC AOD, Fusing Environmental Covariates AOD) in the arid and semi-arid 21 areas. Then, based on FEC AOD, we analyzed the spatiotemporal pattern of AOD and 22 further discussed the interpretation of environmental covariates to AOD. The result 23 shows that the bagging trees ensemble model has a good performance, with its 24 verification R 2 always keeping at 0.90 and the R 2 being 0.79 for FEC AOD compared 25 with AERONET. The high AOD areas are located in the Taklimakan Desert and the 26 Loess Plateau, and the low AOD area is concentrated in the south of Qinghai province. 2 Taklimakan Desert, while the AOD in the southern Qinghai province almost shows no 30 significant change between 2000 and 2019. The annual variation characteristics present 31 that AOD is the largest in spring (0.267) and the smallest in autumn (0.147); the AOD 32 pattern in Gansu province is bimodal, but unimodal in other provinces. The farmland 33 and construction land are at high AOD levels compared with other land cover types. 34 The meteorological factors demonstrate a maximum interpretation of AOD on all set 35 temporal scales, followed by the terrain factors, and the surface properties are the 36 smallest, i.e., 77.1%, 59.1%, and 50.4% respectively on average. The capability of the 37 environmental covariates for explained AOD varies with season, with an sequence 38 being winter (86.6%) > autumn (80.8%) > spring (79.9%) > summer (72.5%). In this 39 research, we pathbreakingly provide high spatial resolution (250 m) and long time 40 series (2000-2019) FEC AOD dataset in arid and semi-arid regions to support the 41 atmosphere and related study in northwest China, with the full data available at 42 Network ground observation the and MxD08 AOD satellite products collected accuracy FEC AOD; spatiotemporal change is analyzed; environmental of FEC AOD

2 citations

DOI
TL;DR: Chen et al. as discussed by the authors used the bagging trees ensemble model, based on 1 km aerosol optical depth (AOD) data and multiple environmental covariates, to produce a monthly
Abstract: Abstract. Aerosols are complex compounds that greatly affect the global radiation balance and climate system and even human health; in addition, aerosols are currently a large source of uncertainty in the numerical simulation process. The arid and semi-arid areas have fragile ecosystems with abundant dust but lack related high-accuracy aerosol data. To solve these problems, we use the bagging trees ensemble model, based on 1 km aerosol optical depth (AOD) data and multiple environmental covariates, to produce a monthly advanced-performance, full-coverage, and high-resolution (250 m) AOD product (named FEC AOD, fusing environmental covariates AOD) covering the arid and semi-arid areas. Then, based on the FEC AOD products, we analyzed the spatiotemporal AOD pattern and further discussed the interpretation of environmental covariates to AOD. The results show that the bagging trees ensemble model has a good performance, with its verification R2 values always remaining at 0.90 and the R2 being 0.79 for FEC AOD compared with AERONET AOD product. The high-AOD areas are located in the Taklimakan Desert and on the Loess Plateau, and the low-AOD areas are concentrated in southern Qinghai province. The higher the AOD, the stronger the interannual variability. Interestingly, the AOD reflected a dramatic decrease on the Loess Plateau and an evident increase in the south-eastern Taklimakan Desert, while the southern Qinghai province AODs showed almost no significant change between 2000 and 2019. The annual variation characteristics show that the AOD was largest in spring (0.267±0.200) and smallest in autumn (0.147±0.089); the annual AOD variation pattern showed different features, with two peaks in March and August over Gansu province but only one peak in April in other provinces/autonomous regions. Farmlands and construction lands have high AOD levels compared to other land cover types. Meteorological factors demonstrate the maximum interpretation ability of the AODs on all set temporal scales, followed by the terrain factors, while surface properties have the smallest explanatory abilities; the corresponding average contributions are 77.1 %, 59.1 %, and 50.4 %, respectively. The capability of the environmental covariates to explain the AOD varies seasonally in the following sequence: winter (86.6 %) > autumn (80.8 %) > spring (79.9 %) > summer (72.5 %). In this research, we provide a pathbreaking high spatial resolution (250 m) and long time series (2000–2019) FEC AOD dataset covering arid and semi-arid regions to support atmospheric and related studies in northwest China; the full dataset is available at https://doi.org/10.5281/zenodo.5727119 (Chen et al., 2021b).

1 citations

References
More filters
Journal ArticleDOI
TL;DR: The operation and philosophy of the monitoring system, the precision and accuracy of the measuring radiometers, a brief description of the processing system, and access to the database are discussed.

6,535 citations


"A global land aerosol fine-mode fra..." refers background in this paper

  • ...The AERONET is a worldwide, sun–sky photometer network providing ground-level aerosol properties, recently updated to Version 3 (Holben et al., 1998)....

    [...]

Journal ArticleDOI
TL;DR: The Collection 6 (C6) algorithm as mentioned in this paper was proposed to retrieve aerosol optical depth (AOD) and aerosol size parameters from MODIS-observed spectral reflectance.
Abstract: . The twin Moderate resolution Imaging Spectroradiometer (MODIS) sensors have been flying on Terra since 2000 and Aqua since 2002, creating an extensive data set of global Earth observations. Here, we introduce the Collection 6 (C6) algorithm to retrieve aerosol optical depth (AOD) and aerosol size parameters from MODIS-observed spectral reflectance. While not a major overhaul from the previous Collection 5 (C5) version, there are enough changes that there are significant impacts to the products and their interpretation. The C6 aerosol data set will be created from three separate retrieval algorithms that operate over different surface types. These are the two "Dark Target" (DT) algorithms for retrieving (1) over ocean (dark in visible and longer wavelengths) and (2) over vegetated/dark-soiled land (dark in the visible), plus the "Deep Blue" (DB) algorithm developed originally for retrieving (3) over desert/arid land (bright in the visible). Here, we focus on DT-ocean and DT-land (#1 and #2). We have updated assumptions for central wavelengths, Rayleigh optical depths and gas (H2O, O3, CO2, etc.) absorption corrections, while relaxing the solar zenith angle limit (up to ≤ 84°) to increase poleward coverage. For DT-land, we have updated the cloud mask to allow heavy smoke retrievals, fine-tuned the assignments for aerosol type as function of season/location, corrected bugs in the Quality Assurance (QA) logic, and added diagnostic parameters such topographic altitude. For DT-ocean, improvements include a revised cloud mask for thin-cirrus detection, inclusion of wind speed dependence on the surface reflectance, updates to logic of QA Confidence flag (QAC) assignment, and additions of important diagnostic information. At the same time, we quantified how "upstream" changes to instrument calibration, land/sea masking and cloud masking will also impact the statistics of global AOD, and affect Terra and Aqua differently. For Aqua, all changes will result in reduced global AOD (by 0.02) over ocean and increased AOD (by 0.02) over land, along with changes in spatial coverage. We compared preliminary data to surface-based sun photometer data, and show that C6 should improve upon C5. C6 will include a merged DT/DB product over semi-arid land surfaces for reduced-gap coverage and better visualization, and new information about clouds in the aerosol field. Responding to the needs of the air quality community, in addition to the standard 10 km product, C6 will include a global (DT-land and DT-ocean) aerosol product at 3 km resolution.

1,628 citations


"A global land aerosol fine-mode fra..." refers background or methods in this paper

  • ...1 MODIS aerosol product (Levy et al., 2013) no longer includes global scale FMF, so we used FMF at 550 nm from the previous collection (C5) for comparison purposes (Levy et al., 2007)....

    [...]

  • ...1) level (Levy et al., 2013; C. Chen et al., 2020)....

    [...]

  • ...…is used for discriminating anthropogenic aerosols from natural ones (Bellouin et al., 2005), has been regarded as highly unreliable according to satellite-based AOD retrievals, especially over land (Levy et al., 2013; Yan et al., 2017; Liang et al., 2021; Yang et al., 2020; Zang et al., 2021a)....

    [...]

Journal ArticleDOI
TL;DR: The Global Fire Emissions Database (GFED) as mentioned in this paper has been used to quantify global fire emissions patterns during 1997-2016, with the largest impact on emissions in temperate North America, Central America, Europe, and temperate Asia.
Abstract: . Climate, land use, and other anthropogenic and natural drivers have the potential to influence fire dynamics in many regions. To develop a mechanistic understanding of the changing role of these drivers and their impact on atmospheric composition, long-term fire records are needed that fuse information from different satellite and in situ data streams. Here we describe the fourth version of the Global Fire Emissions Database (GFED) and quantify global fire emissions patterns during 1997–2016. The modeling system, based on the Carnegie–Ames–Stanford Approach (CASA) biogeochemical model, has several modifications from the previous version and uses higher quality input datasets. Significant upgrades include (1) new burned area estimates with contributions from small fires, (2) a revised fuel consumption parameterization optimized using field observations, (3) modifications that improve the representation of fuel consumption in frequently burning landscapes, and (4) fire severity estimates that better represent continental differences in burning processes across boreal regions of North America and Eurasia. The new version has a higher spatial resolution (0.25°) and uses a different set of emission factors that separately resolves trace gas and aerosol emissions from temperate and boreal forest ecosystems. Global mean carbon emissions using the burned area dataset with small fires (GFED4s) were 2.2 × 1015 grams of carbon per year (Pg C yr−1) during 1997–2016, with a maximum in 1997 (3.0 Pg C yr−1) and minimum in 2013 (1.8 Pg C yr−1). These estimates were 11 % higher than our previous estimates (GFED3) during 1997–2011, when the two datasets overlapped. This net increase was the result of a substantial increase in burned area (37 %), mostly due to the inclusion of small fires, and a modest decrease in mean fuel consumption (−19 %) to better match estimates from field studies, primarily in savannas and grasslands. For trace gas and aerosol emissions, differences between GFED4s and GFED3 were often larger due to the use of revised emission factors. If small fire burned area was excluded (GFED4 without the s for small fires), average emissions were 1.5 Pg C yr−1. The addition of small fires had the largest impact on emissions in temperate North America, Central America, Europe, and temperate Asia. This small fire layer carries substantial uncertainties; improving these estimates will require use of new burned area products derived from high-resolution satellite imagery. Our revised dataset provides an internally consistent set of burned area and emissions that may contribute to a better understanding of multi-decadal changes in fire dynamics and their impact on the Earth system. GFED data are available from http://www.globalfiredata.org .

1,135 citations

Journal ArticleDOI
TL;DR: In this article, the authors validate the MODIS along-orbit Level 2 products by comparing to quality assured Level 2 AERONET sunphotometer measurements at over 300 sites, and find that >66% (one standard deviation) of MODIS-retrieved aerosol optical depth (AOD) values compare to AERO-observed values within an expected error (EE) envelope of ±(0.05 + 15%), with high correlation (R = 0.9).
Abstract: . NASA's MODIS sensors have been observing the Earth from polar orbit, from Terra since early 2000 and from Aqua since mid 2002. We have applied a consistent retrieval and processing algorithm to both sensors to derive the Collection 5 (C005) dark-target aerosol products over land. Here, we validate the MODIS along-orbit Level 2 products by comparing to quality assured Level 2 AERONET sunphotometer measurements at over 300 sites. From 85 463 collocations, representing mutually cloud-free conditions, we find that >66% (one standard deviation) of MODIS-retrieved aerosol optical depth (AOD) values compare to AERONET-observed values within an expected error (EE) envelope of ±(0.05 + 15%), with high correlation (R = 0.9). Thus, the MODIS AOD product is validated and quantitative. However, even though we can define EEs for MODIS-reported Angstrom exponent and fine AOD over land, these products do not have similar physical validity. Although validated globally, MODIS-retrieved AOD does not fall within the EE envelope everywhere. We characterize some of the residual biases that are related to specific aerosol conditions, observation geometry, and/or surface properties, and relate them to situations where particular MODIS algorithm assumptions are violated. Both Terra's and Aqua's–retrieved AOD are similarly comparable to AERONET, however, Terra's global AOD bias changes with time, overestimating (by ~0.005) before 2004, and underestimating by similar magnitude after. This suggests how small calibration uncertainties of

1,069 citations


"A global land aerosol fine-mode fra..." refers background or methods in this paper

  • ...The MODIS sensor onboard Terra has provided long-term observations on a global scale every day since February 2000 (Levy et al., 2010), available at the Atmosphere Archive & Distribution System Distributed Active Archive Center....

    [...]

  • ...Although this MODIS FMF product is not reliable over land (Levy et al., 2010), it was used in numerous previous studies (Ramachandran, 2007; Vinoj et al....

    [...]

  • ...The AODs over the bright surface used for the Phy-DL FMF retrieval were significantly overestimated, with the worst performance compared to other vegetated land cover types (Levy et al., 2010; Petrenko and Ichoku, 2013)....

    [...]

  • ...Although this MODIS FMF product is not reliable over land (Levy et al., 2010), it was used in numerous previous studies (Ramachandran, 2007; Vinoj et al., 2014) including for PM2....

    [...]

Journal ArticleDOI
TL;DR: An overview of the as-built instrument characteristics and the application of MISR to remote sensing of the Earth is provided.
Abstract: The Multi-angle Imaging SpectroRadiometer (MISR) instrument is scheduled for launch aboard the first of the Earth Observing System (EOS) spacecraft, EOS-AM1. MISR will provide global, radiometrically calibrated, georectified, and spatially coregistered imagery at nine discrete viewing angles and four visible/near-infrared spectral bands. Algorithms specifically developed to capitalize on this measurement strategy will be used to retrieve geophysical products for studies of clouds, aerosols, and surface radiation. This paper provides an overview of the as-built instrument characteristics and the application of MISR to remote sensing of the Earth.

947 citations

Frequently Asked Questions (1)
Q1. What are the contributions in "A global land aerosol fine-mode fraction dataset (2001–2020) retrieved from modis using hybrid physical and deep learning approaches" ?

Overall, this study provides a new FMF dataset for global land areas that can help improve their understanding of spatiotemporal fineand coarse-mode aerosol changes.