Post on 19-Feb-2018
International Workshop on Remote Sensing and Eco-hydrology in Arid Regions
Regional Remote Sensing Products for
Eco-hydrological Application in Arid Regions
Qinhuo Liu(qhliu@irsa.ac.cn)
State Key Laboratory of Remote Sensing Sciences
Institute of Remote Sensing and Digital Earth, CAS
September 18, 2013
Beijing, China
Outlines
1. Introduction
2. Prototype system for Land Surface Parameters
Inversion Based on Multi-source Remote Sensing
Data
3. Remote Sensing Product Generation for Eco-
hydrological Application in Arid Regions
4. Expectations and Discussions
1. Introduction
Heihe is one of the largest
inland river basin in
Northwest China:
water-stressed ecosystems,
complicated eco-hydrological
processes, more fragile to
climate change and
anthropogenic disturbance
A major research plan
“Integrated research on the
eco-hydorlogical process
of the Heihe River Basin”
launched by NSFC in 2010
the objectives of the Heihe Plan include:
To reveal the processes and mechanisms of the eco-
hydrological system in an inland river basin at different
scales(e.g. Leaf, individual plant, community, landscape,
and watershed)
to improve the research capabilities and predictability of
the evolution of hydrological, ecological, and economic
systems;
to determine the responses of eco-hydrological
processes to climate change and human activities;
and to provide fundamental theory and technical support
for water security, ecological security, and sustainable
development in inland river basins.
• Distributed Hydrological Model :DHSVM(Wigmosta et al., 1994)、SWAT(Arnold and Fohrer, 2005)and GEOtop(Rigon et al., 2006);
• Underground Water model: MODFLOW(Harbaugh et al., 2000);
• Dynamic vegetation model and Vegetation Growth model: LPJ-DGVM
(Sitch et al., 2003)、BIOME-BGC(Thornton and Running, 2002)and WOFOST(Vandiepen et al., 1989);
• Land surface process model: SiB2(Sellers et al., 1996)、CoLM
(Dai et al., 2003)and CLM 3.0/4.0(Oleson et al., 2010)。
• Parameters
• Variables
• Driven Data
which types/what scales of the eco-hydrological models ?
Key Parameters/Variables for Regional for
regional Eco-hydrological process models
variables/parameters applications Chinese satellite International satellite
precipitation 流域水循环;驱动变量 FY-3 TRMM, GPM
Evaportranspiration 流域水循环;生态耗水;水分利用效
率(生态水文耦合)
FY-3成像光谱仪、HJ-1 ASTER, MODIS, MERIS, AATSR
Flow 流域水循环;分布式水文模型 OSTM/Jason-2 Radar Altimeter,
ENVISAT Radar Altimeter-2, TerraSAR-
X, SWOT
Undergroound water 流域水循环;地表-地下水相互作用 GOCE, GRACE, GRACE-II
Soil Moisture 流域水循环;冻土水文 FY-3微波辐射计 SMAP, SMOS, ALOS PALSAR, Envisat
ASAR, BIOMASS
Snow 流域水循环;积雪水文 FY-3微波辐射计 CoReH2O, SCLP
Vegetation Classification 分布式水文模型和生态模型所需参数 HJ-1, CBERS Proba CHRIS, HYSPIRI, 其他高光谱卫
星
Land use/Cover 分布式水文模型和生态模型所需参数;
人类活动影响评价
HJ-1, CBERS TM, SPOT, MODIS, 其他多光谱卫星
Plant structure 水文模型和生态模型所需参数;灌溉
管理
HJ-1, CBERS Proba CHRIS, HYSPIRI, 其他高光谱卫
星
Vegetation coverage 水文模型和生态模型所需参数 HJ-1, CBERS TM, SPOT, MODIS, 其他多光谱卫星
Leaf area index 水文模型和生态模型所需参数;生态
系统动态
HJ-1, CBERS TM, SPOT, Proba CHRIS, MODIS, 其他
多光谱卫星
Vegetation structure 生态模型所需参数
NPP、NEP 生态系统动态;水分利用效率(生态
水文耦合)
HJ-1 DESDynI, BIOMASS, 其他高光谱卫星,
其他激光雷达卫星
DEM 分布式水文模型和生态模型所需参数 ALOS PRISM, ASTER等立体像对数据
Parameters Spatial Resolution
Temporal Resolution Period Accuracy
LULC 30m/250m 1month/1 year 2012-2015 90%
Albedo 1KM 5 day 2012-2015 90%
VI/VC 250m/1KM 5 day 2012-2015 85%
PAR 1KM Hourly, Daily 2012-2015 90%
NPP 1KM 5 day 2012-2015 80%
the Most important Regional Remote Sensing
Products for Eco-hydrological Application
Here, we only list the Ecological system parameters
how can produce a series of
high-quality remote sensing
products that can support
basin-scale integrated eco-
hydrological studies, from
the multi-source satellite
observation, ?
We have a lot of global remote
sensing product, such as
MODIS, does the accuracy
satisfy the requirements?
2. Prototype system for Land Surface Parameters
Inversion Based on Multi-source Remote Sensing Data
9
Land surface process model scale issues
Land surface process
model need different spatial
scale observation: 1KM,
5KM, 10KM, 100KM
crop
road
soil
urban
forest
Mixed pixel, terrain effect
Leaf:Component Spetrum
Vegetation canopy:3D structure
Hi-WATER
Objectives
Global Products Global scale--coarser
Developed global algorithm but not localized
Terrain effect is overlooked
Pre-knowledge is barely used
Usually based on one type of remotely sensed data
Pre-defined classification system
10
Reginoal Products River basin scale--finer
Specialized for local applications
Terrain effect is considered
Knowledge from field campaigns can be used
Based on multi-source remote sensing data
Classification system specialized for this area
Regional Remote Sensing Products for
Eco-hydrological Application the whole Heihe river basin
Hi-WATER
Design consideration
11
Parameter Retrieval from Multi-source remote sensing data for higher resolution and accuracy
Prior Knowledge from HiWATERexperiment
Topographic correction based on the DEM
New satellite datasuch as HJ-1 with720KM swath:WFI:0.43~0.90μm,
30m;
IRS: 0.75 ~12.5m,
150m/ 300m .
12
Quantitative Monitoring:
Inversion
Calibration corretionGeometric correction
Cloudy mask
Atmospheric correction
Prediction : assimilation
Qualitative application: Classification and
identification
Information
extraction
Preprocessing
Parameter Inversion algorithm development
13
Multi-source remote sensing data from different sensors
provide a chance to improve the temporal and spatial
resolution of existing hydrological products.
Pre-processing of multi-data
14
Preprocessing to normalize the multi-source remote
sensing datas: Reflectance dataset of USA(MODIS/
AVHRR), Chinese (FY3_VIRR),
European (VEGETATION) etc with
5km resolution
Dataset of USA(MODIS/AVHRR),
Chinese (FY3_VIRR), European
(VEGETATION) etc. with 1km resolution
• Geometric correction
• Resample, reprojection
• Band response function analysis
• Information augment analysis
net multisensors monosensorI I I
1 2 3ln ln ln ln( )I MSE
Preprocessing methods
15
Prototype system design
16
Institute of Remote Sensing Applications (IRSA),
CAS
Prototype system design
3. Remote Sensing Product Generation for Eco-
hydrological Application in Arid Regions
(1) Land use and Land Classification
(2) Vegetation Index/Vegetation
Coverage/ Leaf Area Index
(3) Photosynthesis Absorption
Radiation /Net Primary Production
Hi-WATER
Land use and Land Classification: What’s now?
Land
cover
dataset
Satellite
/ sensor
Time
period
Spatial
resolutio
n (m)
Number
of
classes
Classifica
tion
system
Classifica
tion
accuracy
*
GLC2000
v1.1
SPOT/
Vegetati
on
2000 1000 22 UN LCCS 39-64%
Mod12q1
C004
Terra/
MODIS2001 1000 17 IGBP 75-80%
Mod12q1
C004
Terra/
MODIS2005 500 17 IGBP 72-77%
GlobCove
r v2.2
Envisat/
Meris
2004,
2006300 22 UN LCCS 67.1%
FROM-
GLC
Landsat/
TM,
ETM+
Mainly
201030, 250 28
UN LCCS
modified64.9%
18
Hi-WATER
What’s the problem?
Less land cover types
Unbalanced classification accuracy for different regions
Accuracy is not high enough for applications in river basin scale
Classification results from different datasets are not consistent
19
Hi-WATER
Main methods for LUCC mapping
Fine spatial
resolution data
High temporal
data
High spatio-
temporal data
Data
example
Landsat-TM/ETM+,
ASTER, etc.
MODIS, AVHRR, MERIS,
etc.
HJ-1/CCD
Advantage Less mixed pixels
More precise boundary
Can capture the
phenology of vegetation
All advantages of
fine resolution and
high temporal data
Disadvanta
ge
Cannot capture the
phenology of vegetation
More mixed pixels
Classification uncertainty
is high in mixed area
N/A
20
Hi-WATER
The strategies
The advantage of high spatio-temporal resolution of HJ-1/CCD is fully taken
Getting support from multi-source remote sensing data, such as Landsat-TM/ETM+ and Google Earth images
Pre-knowledge from field campaigns are incorporated
Other information used, such as DEM
21
Hi-WATER
The advantages of high spatio-temporal resolution data
High temporal frequency can show the phenology of vegetation
22
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
小麦
大麦
油菜
玉米
Main crops NDVI variation at middle stream
5-15:wheat grows,maize
not
6-19:wheat grow s better
than maize 7-14:no difference
7-14:no difference8-3:Maize still
grows; others are
declining
8-20:Maize still
grows; wheat is
declining; rape is
harvested
9-2:only maize
left
Hi-WATER
The advantages of high spatio-temporal resolution data
Winter season is better for extracting evergreen forests
Multi-month images can partly get rid of shadow effects
24
Evergreen forests extracted from
images on Jan-Apr.
Combination from results of the
four month’s
Hi-WATER
The advantages of high spatio-temporal resolution data
Texture information from high spatial resolution data is used to differentiate grasslands and broadleaf forests
25
• Broadleaf forests have much
more texture than grass lands
and croplands
• Gray level co-occurrence matrix
can embody the difference
HJ image Texture
Hi-WATER
Multi-source remotely sensed data
Very high spatial resolution images are used to improve the samples accuracy
TM/ETM+ is used for NDBI calculation to help extract urban areas
26
Broadleaf
forests
Hi-WATER
Other information -- DEM
DEM and pre-knowledge are used to simplify the classification, such as urban areas and bare rocks below.
27
Urban areas’ altitude
usually lower than
3000m
Bare rocks’ altitude
usually higher than
3000mExtracted urban
area
Hi-WATER
Pre-knowledge from field campaigns
What can we get from field campaigns
Crops’ planting structure
Crops’ exact phenology
Main crops planted and its rotation information
Some first hand information
28
Field campaign in Zhangye area
Example of land cover filling
Land cover map of Jan. ~ Mar.
Land cover map of Apr. ~ Jun.
Land cover map of Jul. ~ Sep.
Land cover map of Oct. ~ Dec.
Accuracy evaluation• Lower stream has the highest accuracy;upper stream
is effected by shadows;Middle stream has the lowest
accuracy.
• Overall accuracy is higher than 95%,Kappa is higher
than 0.94 without considering finer crop classification.
• Overall accuracy is higher than 90%,Kappa is close to
0.9 considering finer crop classificationOA:92.19% Kappa:0.87
Types Commi
ssion
rate
(%)
Omissi
on
rate
(%)
Cropla
nds0/315 0.00 18/333 5.41
Water 0/54 0.00 13/67 19.40
Urban 1/76 1.32 15/90 16.67
Barren 45/144 31.25 0/99 0.00
OA:82.97% Kappa:0.80
typesCommis
sionrate(%)
Omissio
nrate(%)
Maize 21/93 22.58 18/90 20.00
Spring
wheat3/64 4.69 14/75 18.67
Barley 5/67 7.46 19/81 23.46
Rape 21/66 31.82 17/62 27.42
water 0/54 0.00 13/67 19.40
Urban 1/76 1.32 15/90 16.67
Barren 45/144 31.25 0/99 0.00
3. Remote Sensing Product Generation for Eco-
hydrological Application in Arid Regions
(1) Land use and Land Classification
(2) Vegetation Index/Vegetation
Coverage/ Leaf Area Index
(3) Photosynthesis Absorption
Radiation /Net Primary Production
• Normalized difference vegetation index(NDVI) is a key
parameter to describe physical and biological processes of
plants. (Asraret et al., 1984; Goward , 1992)
• Vegetation compositing technology offers a method to obtain
NDVI products with special consistency and continuity. (Huete, et
al., 1999; Lovell et al., 2003)
• BRDF compositing scheme reduces sun-target-sensor variations
with using of a BRDF model, represent major progress for
deriving biophysical variables from the radiative transfer model.(Huete, et al., 2002; Duchemin, et al., 2002; Vancutsem, et al., 2007)
• The quality of NDVI product from a single observation is always
unsatisfactory because of the ever-present atmosphere and
cloud contamination, even the data integrality.
Background for NDVI normalization
Sensor Produc
ts
Algorithm Time
resolution
Time
coverage
Spatial
resolution
AVHRR
GVI
MVC, Maximum value
compositing(Holben, 1986)
15 days 1982-present 4 km
8 kmGIMM
S
15 days 1981-2006
PAL 10 days 1981-1999
VEGETAT
ION
VGT-
S10
MVC 10 days 1998-present 1 km
VGT-
D10
BDC, BiDirectional
Compositing(Duchemin, et al., 2002)
2002-present
MODIS MOD1
3
BRDF compositing, back up
MVC,
CV-MVC, constrained- view
angle- MVC(Huete, et al., 2002)
16 days 1999-present 250m
1kmMYD1
3
2002-present
MERIS MGVI MC:mean compositing(Vancutsem, et al., 2007)
10 days 2002-present 1.2 km
Global and regional composited VI products
The commonly used global and regional composited vegetation products are listed. We can find that the MVC is still the most used algorithm, and the BRDF scheme is used more widely.
• The BRDF scheme interpolate all bidirectional reflectance
observations, of acceptable quality, to their nadir-
equivalent band reflectance values from which the VI is
computed and produced.
• To insure the interpolated quality, at least five good quality
observations is required. This limits the temporal
resolution of NDVI product and increases the change
uncertainty in the composite period, especially when the
vegetation grow fast.
• The multi-sensor composite strategy was developed to
improve the temporal resolution.
BRDF compositing scheme
Multi-sensor data compositing
• This paper was to conduct compositing and analyses on the NOAA 18
AVHRR, Terra and Aqua MODIS, and FY-3A VIRR multi-sensor
datasets.
http://fy3.satellite.cma.gov.cn/PortalSite/d
efault.aspx http://aqua.nasa.gov/;
http://terra.nasa.gov/;
http://www.oso.noaa.gov/poesstatus/;
and
http://www2.ncdc.noaa.gov/docs/podug/
Operational dates of NOAA
AVHRR, Terra and Aqua
MODIS, and FY-3A and FY-
3B VIRR.
1)multi-sensor datasets
In view of sensors homogeny
and time coverage, NOAA-15 to
NOAA-19 AVHRR, Terra and
Aqua MODIS, and FY-3A VIRR
can be used to build multi-
source dataset.
Multi-sensor data compositing
• The compositing period controls the number of
acquisitions considered. It is more pronounced as almost
4 tiles images per day for the multi-sensor datasets.
The distribution of
observation angles for
AVHRR, Terra and Aqua
MODIS, and VIRR
IN 4 or 5 days the BRDF
shape of the pixels has
been able to gain.
IN 7 or 8 days the BRDF
details information has
been able to gain
2)compositing period
Multi-sensor data compositing
• The technique employed depends on the number and quality of
observations n .
• The RossThick-LiSparseR BRDF model (Roujean et al., 1992) is used.
2)compositing period
Primary multi-sensor VI compositing methodology
),,(),,(),,( svvolvolsvgeogeoisosv kfkff
Multi-sensor data compositing
• Ri is a good indicator for the representation of the forepart i days
subset dataset to the whole N days dataset
• W e find the mean R2 can highlight the difference of Ri, and also
keep the trend
2)compositing period
The mean correlation coefficient of
accumulated data increased day by
day of each group
The 4th day mean R2 of each
group are
0.9269, 0.9314, 0.8552, so 4
days compositing period being
close to the whole group data with
the best temporal resolution
Ni
ii
kN
sizek
kkNki
sizek
kki
kN
sizek
kkNki
sizek
kki
i
RN
Rmean
NDVINDVINDVINDVI
NDVINDVINDVINDVI
R
1
22
2,
1,
2,
1,
,1
,,1
,
1
)()(
))((
Multi-sensor data compositing
• Many linear BRDF models was compared, the simple Walthall model
(Walthall et al, 1985) and Ross-thick/Li-sparse BRDF model(Roujean
et al., 1992 ) was found to perform well for most vegetation types.
(Privette et al. 1997; Leeuwen et al. 1996)
3)BRDF model of compositing
Ross-thick/Li-sparse require 7 good
observation and Walthall require 5 to
insure the fitting quality
Walthall BRDF
Ross-Li BRDF Model
The two BRDF
models was
compared from
the multi-sensor
datasets
compositing
R2 of 4 days subset and the
whole dataset composited NDVI
of each group
The Walthall BRDF
model perform better
and required less good
quality observations
Results Analysis
Composited NDVI from Jul.21
to Aug.4
• Consistency problems of BRDF back up pixels (MVC)
The MVC composited
value are always
come from VIRR and
AVHRR,nearly
92%;Need to improve the
sensor differences
BRDF back up
Algorithm
source tracker
Results Analysis• Stripe noises of CV-MVC pixels
Composited NDVI
from Aug.1 to
Aug.4
CV-MVC
source tracker
The sign map of
the CV-MVC
composited pixels
It shows the
problem is also
arise from the
sensor differences
Results Analysis• Liner band switch of AVHRR, VIRR, MODIS sensors
Improve the sensor differences
d
d
0
0
)(
)()(
The Liner band switch Data
The Bold line present the sensitivity of
Sensors
The thin lines mean the measured
spectrum curve
The Liner band switch
result of AVHRR,
VIRR, MODIS
sensors of Red and
NIR
AVHRR_MRed = 1.0465 * AVHRRRed +
0.0039
AVHRR_MNIR = 1.0975 * AVHRRNIR +
0.0093
VIRR_MRed = 1.6558* VIRR Red + 0.0132
VIRR _MNIR = 1.5999 * VIRR NIR +
0.0000
Results Analysis• Multi-sensor dataset compositing algorithm
Conclusions• Cross validation of MOD12A2
Cross validation of MOD12A2 and
multi-sensor datasets composited NDVI
by developed algorithm a)7.28-7.31
b)8.1-8.4 c)8.5-8.8 d)8.9-8.12
The R2 of Multi-sensor
subset products and
MOD12A2 are 0.8850,
0.8629, 0.9146, 0.9205,
presenting a good
results.
• Cross validation with MOD12A2 shows the R2 of multi-
sensor subset products are all higher than 0.86,
presenting a good precision.
• Multi-sensor strategy improves the temporal resolution
from 16 days to 4 days.
• The success in compositing the primary product
illustrating the multi-sensor strategy is feasible and
promising.
FVC product and validation
30m FVC 250m FVC
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0 200
FV
C
Julian Day
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0 100 200 300
FV
C
Julian DayFVC accuracy : 85%
a MODIS pixel
(250m)a HJ
pixel
(30m)
3. Remote Sensing Product Generation for Eco-
hydrological Application in Arid Regions
(1) Land use and Land Classification
(2) Vegetation Index/Vegetation
Coverage/ Leaf Area Index
(3) Photosynthesis Absorption
Radiation /Net Primary Production
Hi-WATER
The problem in NPP estimation model---light use efficiency model
53
The expression of NPP LUE model:
is the incidence solar radiation between 400-700nm that reached the canopy.PARQ
FPAR is the fraction of solar radiation absorbed by canopy.
n is the light-use efficiency.
nPARQFPARNPP
FPAR is inversed by the relationship between FPAR and VI /LAI.
1)VI has saturation problem.
2)The method based on physical model does not distinguish direct radiation and diffuse
radiation.
PAR is estimated by the relationship between PAR and solar
radiation.
1)The relationship is changed with location and the weather.
2)Most atmospheric radiative transfer model does not distinguish direct radiation and
diffuse radiation
Hi-WATER
FPAR/PAR variation under different weather condition
54
0
0.2
0.4
0.6
0.8
1
1.2
9.5 11.5 13.5 15.5 17.5
有云天晴天
Time
FP
AR
Cloudy sky
Clear sky
The observed daily change of FPAR under
cloudy sky and clear sky
The percentage changing of direct and
diffuse PAR with AOT
The daily change of FPAR under
cloudy sky has obvious difference with
that under clear sky.
The direct FPAR and the diffuse FPAR
has different characteristic.
The percentage of direct PAR and
diffuse PAR changes with different
weather condition, different seasons
and areas.
Hi-WATER
PAR estimation
55
Key process: Look-up table
LUT=f(AOD , COT , WV, OZ, α, θs )
Key parameters:Cloud
optical thickness
Instantaneous PAR estimation
on horizontal surface.
Terrain Corrections :Altitude,
Slope and Aspect
Topographic obstruction on direct PAR; sky
view factor affection on diffuse PAR .
Temporal integration :
Integrate instantaneous PAR to
hourly and daily PAR.
Hi-WATER
FPAR inversion
56
= (1 )*dir diffFPAR k FPAR k FPAR
(1 )*(1 )
[1 ( 1)* ]
b gap
dir
gap
PFPAR
a P
(1 )*(1 )
[1 ( 1)* ]
w open
diff
open
KFPAR
a K
Direct FPAR is calculated from Black
sky Albedo and canopy gap
probability.
Diffuse FPAR is calculated from
White sky Albedo and the openness
of the crown at the top of canopy.
K is the ratio of direct-to-total which
can be inversed from PAR estimation.
is the absorptivity ratio of soil and
crown canopy .
a
Hi-WATER
Product verification data
57
Downward PAR
Observation
Upward PAR
Observation
plant soilPAR PAR =APAR +APAR
*plant plantAPAR FPAR PAR
Downward/upward PAR field sites
covering three different surface types
(Forest, Grassland and farmland) on
Heihe Basin.
These observation data could be used to
verify indirectly.plantAPAR
Hi-WATER
PAR products
58
Daily direct PAR Daily total PAR
Product Period:2012.6.1-2012.8.31
Temporal resolution: Instantaneous, hourly, daily PAR
Product classification: Direct PAR, Diffuse PAR, Total PAR, Ratio of
direct-to-total.
Hi-WATER
FPAR products
59
Product Period:2012.6.1-2012.8.31
Temporal resolution: 8 day
Product classification: Direct FPAR, Diffuse FPAR, Total FPAR.
Hi-WATER
Further research and product specification
APAR validation on Heihe Basin.
The light use efficiency estimation which distinguish direct radiation
and diffuse radiation.
The research on NPP estimation methodology and its validation.
Product production.
60
Products Period Temporal
Resolution
Spatial Resolution Area
FPAR 2012-2015 8 day 1KM Heihe Basin
PAR 2012-2015 Hourly, daily 1KM Heihe Basin
K 2012-2015 Hourly, daily 1KM Heihe Basin
NPP 2012-2015 8day 1KM Heihe Basin
61
Expectations:
The first version products for 2012 will be released in the end of this year;
The second version products for 2012-2013 will be released in March 2014.
The third version products for 2012-2014 will be released in March 2015.
Discussions:
(1) How to combine accumulated earth observation data to generate the long-
term, consistent and accurate land surface products?
Synergic inversion based on information content and error propagation
using Multi-scale, multi-spectral (VIS-NIR-TIR-MIR), multi-satellites
(Geostationary and polar orbiting, USA, China, European and others)?
(2) How to validate the remote sensing inversed land surface parameters at
different scale?
Multi-scale and long term field observation and Scaling transformation?
(3) How to combine the Regional Remote Sensing Products for Eco-
hydrological process model in Arid Regions?
Assimilation?
4. Expectations and Discussions
62
Xin Li, Guodong Cheng, Shaomin Liu, Qing Xiao, Mingguo Ma, Rui Jin, Tao Che, Qinhuo Liu, Weizhen
Wang, Yuan Qi, Jianguang Wen, Hongyi Li, Gaofeng Zhu, Jianwen Guo, Youhua Ran, Shuoguo Wang,
Zhongli Zhu, Jian Zhou, Xiaoli Hu, and Ziwei Xu , HEIHE WATERSHED ALLIED TELEMETRY
EXPERIMENTAL RESEARCH (HiWATER), BAMS 2013
DongqinYou, Jianguang Wen, Qiang Liu, Qinhuo Liu, Yong Tang,The Angular and Spectral Kernel-
driven Model:assessment and application. IEEE journal of selected topics in applied earth observations
and remote sensing, JSTARS. 2013.2271502,2013.
Yu ShanShan; Xin XiaoZhou; Liu QinHuo, Estimation of clear-sky longwave downward radiation from HJ-
1B thermal data, SCIENCE CHINA-EARTH SCIENCES V: 56 N: 5 P: 829-842 , MAY 2013
Zhang, Hailong; Xin, Xiaozhou; Li, Li; An Improved Parametric Model for Simulating Cloudy Sky Daily
Direct Solar Radiation on Tilted Surfaces, IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH
OBSERVATIONS AND REMOTE SENSING V: 6 N: 1 P: 180-187, FEB 2013
Hua Li, Qinhuo Liu,Yongming Du, Jinxiong Jiang, Heshun Wang. Evaluation of the NCEP and MODIS
Atmospheric Products for Single Channel Land Surface Temperature Retrieval With Ground
Measurements: A Case Study of HJ-1B IRS Data. IEEE journal of selected topics in applied earth
observations and remote sensing, 2013, 6 (3):1399-1408.
References