research article wireless sensor network of typical land...

Research ArticleWireless Sensor Network of Typical LandSurface Parameters and Its Preliminary Applications forCoarse-Resolution Remote Sensing Pixel

Baocheng Dou123 Jianguang Wen23 Xiuhong Li123 Qiang Liu123 Jingjing Peng234

Qing Xiao23 Zhigang Zhang3 Yong Tang3 Xiaodan Wu3 Xingwen Lin3 Dongqin You3

Hua Li3 Li Li3 Yelu Zeng3 Erli Cai1 and Jialin Zhang1

1College of Global Change and Earth System Science Beijing Normal University No 19 Xinjiekou Wai Street Haidian DistrictBeijing 100875 China2Joint Center for Global Change Studies Beijing 100875 China3State Key Laboratory of Remote Sensing Science The Institute of Remote Sensing and Digital EarthChinese Academy of Sciences and Beijing Normal University No 20 North DaTun Road ChaoYang District Beijing 100101 China4Institute of Remote Sensing and GIS Peking University Beijing 100871 China

Correspondence should be addressed to Jianguang Wen wenjgradiaccn

Received 3 July 2015 Accepted 3 April 2016

Academic Editor Yao Liang

Copyright copy 2016 Baocheng Dou et alThis is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

How to obtain the ldquotruthrdquo of land surface parameter as reference value to validate the remote sensing retrieved parameter inheterogeneous scene and coarse-resolution pixel is one of the most challenging topics in environmental studies In this paper adistributed sensor network system named CPP-WSN was established to capture the spatial and temporal variation of land surfaceparameters at coarse-resolution satellite pixel scale around the Huailai Remote Sensing Station which locates in the North ChinaPlain The system consists of three subnetworks that are RadNet SoilNet and VegeNet Time series observations of typical landsurface parameters including UVR PAR SWR LWR albedo and land surface temperature (LST) from RadNet multilayer soilmoisture and soil temperature from SoilNet and fraction of vegetation cover (FVC) clumping index (CI) and leaf area index(LAI) from VegeNet have been obtained and shared on the web Compared with traditional single-point measurement the ldquotruerdquoreference value of coarse pixel is obtained by averaging or representativeness-weighted averaging the multipoint measurementsacquired using the sensor network The preliminary applications which validate several remote sensing products with CPP-WSNdata demonstrate that a high quality ground ldquotruthrdquo dataset has been available for remote sensing as well as other applications

1 Introduction

With the increase of datasets from coarse-resolution remotesensing satellites researchers have developed many quanti-tative products of land surface parameters and have beenconcerned with the need to quantify the accuracy of theirapplication [1 2] Thus several validation plans have beenimplemented to verify these products at both the local scaleand global scale such as BigFoot Validation of the LandEuropean Remote Sensing Instruments (VALERI) and LandProduct Validation (LPV) [3ndash5] These works have evaluated

the accuracy of remote sensing products under abundant landcovers in long time series but mainly use single-point obser-vation for remote sensing pixel validation The representa-tiveness of the in situ point observation with regard to theremote sensing pixel scale observation is questionable espe-cially for products with coarse spatial resolution [6 7] Themismatch between point observation and pixel observationin other words the scale effect becomes the main challengefor the validation of remote sensing products in coarse reso-lution Advanced observation methods should be employedwith consideration of the heterogeneity of coarse pixel

Hindawi Publishing CorporationInternational Journal of Distributed Sensor NetworksVolume 2016 Article ID 9639021 11 pageshttpdxdoiorg10115520169639021

2 International Journal of Distributed Sensor Networks

Study area1km SIN pixel

(km)

(km)

(km)

01250 025 05 075 1

115∘48998400E

115∘48998400E

20∘

30∘

40∘

30∘

40∘

50∘

South China Sea Islands200 0 200 400

N

N

Heihe S Station nameForest Typical land surface typesLand cover type

Evergreen needleleaf forestEvergreen broadleaf forestDeciduous needleleaf forestDeciduous broadleaf forestMixed forestClosed shrublandsOpen shrublandsWoody savannasSavannas

GrasslandsPermanent wetlandsCroplandsUrban and built-up landsCroplandnatural vegetation mosaicSnow and iceBarren or sparsely vegetated landsBodies of water

20∘

15∘

10∘

5∘

120∘115∘110∘

140∘130∘120∘110∘100∘90∘80∘70∘

130∘120∘110∘100∘90∘

8004000400

Figure 1 Location and underlying characteristics of research area (revised from [7])

Wireless sensor network (WSN) which is typically com-prised of a collection of sensors with their own power supplywireless communication data storage and data processingcapability collects and transmits data from remote field sitesback to base station [10ndash12] Networks of embedded devicesthat work together to provide enhanced monitoring acrossspatial and temporal scales are growing in popularity [13]Wireless sensor networks are increasingly applied in researchfields including earth observation environmental monitor-ing agriculture resource management public health publicsecurity transportation and military [14] With the ability oflarge-scale coverage and long time series running WSN hasthe potential to contribute to the solution of the scale effectchallenge in remote sensing research

Recently new remote sensing products of energy bud-get vegetation characteristics and soil characteristics havebeen generated to meet the needs in the climate changeenvironment monitoring and agricultural yield estimationWSN is an effective way for simultaneously monitoring theseparameters at ground With its flexibility automation andrelatively low cost to synchronously acquire data in multiplesites the WSN is recently adopted in validationcalibrationof remote sensing products [15ndash17] For this purpose it isdesirable that each node could contain multiple sensors tomeasure different kinds of parameters In this paper weexplore the establishment of the wireless sensor networkswhich monitor the heterogeneity of coarse spatial resolutionpixel with consideration of different categories of landsurface parameters (hereafter referred to as CPP-WSN)

This paper firstly presents the China validation networkand Huailai remote sensing station where the CPP-WSNlies Then the parameter selection and sensor specificationsof CPP-WSN are discussed Construction of CPP-WSNincluding design of CPP-WSN nodes optimized layout atpixel scale and automatic data transmission and sharing

are described in the third part of this paper And finallythe validations of a variety of remote sensing products areintroduced as preliminary applications of the CPP-WSN

2 Huailai Remote Sensing Station

There are many ecohydrology and flux sites around China(Figure 1) and these sites have provided various data forenvironmental monitoring and remote sensing products val-idation With advanced observational techniques abundantdata accumulation and ability of carrying on multiscaleexperiment the Huailai Remote Sensing Station and around(for short HuailaiS) located in Huailai Hebei provinceChina (40349∘N 115785∘E) becomes one of the ideal studyareas for remote sensing algorithmsproducts validation andscale effect research The HuailaiS is mainly covered by irri-gated corn and unirrigated cornThe differences in irrigationcondition and soil type distribution make the single landcover area heterogeneous but continuously changing whichcharacterize the problem of scale transformation from point-measured parameters to pixel scale reference value

The sinusoidal grid is one of the most widely usedgrid schemes for global remote sensing products For 1 kmsinusoidal pixel covering HuailaiS (the green parallelogramshown on the right of Figure 1) a 2 km lowast 15 km layout areafor CPP-WSN (the red square shown on the right of Figure 1)is chosen in consideration of the adjacent pixelsrsquo influence

3 CPP-WSN Observations and Data Collection

31 Specifications of CPP-WSN and Sensor IntercomparisonConsidering the needs of validating various remote sensingproducts and the different heterogeneity conditions of differ-ent parameters three categories of parameters are observed

International Journal of Distributed Sensor Networks 3

Table 1 Specifications of WSNsrsquo parameters and sensors in CPP-WSN

WSN Number of nodes Categories of parameters Observed parameters Sensors

RadNet 6 Multiband radiation

Ultraviolet irradiance (UVR) CUV5Photosynthetically active radiation (PAR) PQS1Shortwave irradiance (SWR)

CNR4Shortwave albedoLongwave irradiance (LWR)Net radiationLand surface temperature (LST) SI-111

SoilNet 21 Multilayer soil parameters5 cm 10 cm 20 cm and 40 cm soilmoisture EC-5

5 cm 10 cm and 20 cm soil temperature PT100

VegeNet 6 Vegetation parametersFraction of vegetation cover (FVC)Leaf area index (LAI)Clumping index (CI)

Camera

includingmultibandupwarddownward radiationmultilayersoil moisture and temperature and vegetation structureparameters Different layouts are required to capture the het-erogeneity of each kind of parametersTherefore threeWSNswere established namely RadNet SoilNet and VegeNetThe specifications of the three WSNs are shown in Table 1includingWSNname parametersrsquo category observed param-eters and selected sensors The three vegetation structuralparameters that is FVC LAI and CI are estimated fromthe camera-acquired true color image based on the improvedLang and Xiang (LAILX) algorithm [18 19]

Sensor intercomparison is commonly performed beforethe establishment of observation network or experimentsuch as the First International Satellite Land Surface Clima-tology Project Field Experiment (FIFE) the Energy BalanceExperiment (EBEX-2000) and the Heihe Watershed AlliedTelemetry Experimental Research (HiWATER) [20ndash22] Toacquire reliable radiation data whenmultiple sensors observesimultaneously intercomparison is essential to ensure allof these sensors have consistent performance Thus anintercomparison experiment was carried out for RadNetsensorsparameters including CUV5UVR CNR4SWRCNR4LWR PQS1PAR and SI-111LST The intercompari-son scene and acquired data of CNR4SWR are shown inFigure 2 All CNR4 sensors lay on the same horizontal planewith identical solar radiation condition (Figure 2(a)) andtime series observations from various CNR4s were collectedfor comparison (Figure 2(b)) The comparisons of othersensorsparameters were conducted using the same method

The measured parameters were collected and temporallyaveraged for each individual sensor The maximum relativedifferences among sensors of UVR SWR LWR PAR andLST are 483 53 371 11 and 054 respectivelySensorparameter differences indeed exist and are consid-erably large for PAR SWR UVR and LWR which cannotbe ignored Thus it is necessary to analyze and correctthe discrepancies among the sensors The normalizationcoefficient defined as the mean ratio of each sensor observed

value to the average value of all sensor observations isadopted to adjust the sensor observation (1) Consider

119870 =

sum119899

1

(VV)119899

(1)

where 119870 is the normalization coefficient 119899 is the numberof sensor observations V is observation value and V is theaveraged observations of all sensors

Table 2 lists the normalization coefficients of employedsensor parameters The corrected observation value equalsthe original observation value multiplied by the normaliza-tion coefficient

32 CPP-WSN Construction Figure 3 shows the CPP-WSNsystem for coarse scale pixel observation and near real-timedata sharing consists of the following components WSNnodes data receivingprocessing server data sharing serverand data sharing website The observations of WSN nodesare collected by data receivingprocessing server via a 3Gnetwork that is General Packet Radio Service (GPRS) andthen processed in near real-time After quality control andstandardization data together with the metadata is importedinto the data sharing database Finally the near real-timeCPP-WSN dataset is shared with self-explanatory metadataon the data sharing website

According to the needs of validating remote sensing prod-ucts in coarse spatial resolution three scientific objectives areproposed and accomplished to establish the CPP-WSN

321 Design of CPP-WSN Nodes As the variation in vege-tation density is the main driving factor for the variation ofradiation parameters in the study area the nodes of RadNetand VegeNet are bonded together while SoilNet node wasdesigned separately and both RadNetVegeNet node andSoilNet node consist of bracket sensors data collector solarpanel and battery just as Figure 4 shows The minimumheight of bracket and solar panel should be higher than


(a) Scene of intercomparisonSh

ortw

ave i

rrad

ianc

e (W

lowastm

minus2 )

200

400

600

800

1000

151

400

162

900

152

900

154

400

155

900

161

400

145

900

144

400

Time

CNR4_1CNR4_2CNR4_3

CNR4_4CNR4_5CNR4_6

(b) Observed shortwave irradiance

Figure 2 Intercomparison scene (a) and measured data (b) of CNR4SWR

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

Data sharing serverData receivingprocess server

Data sharing website

Pixel scale wireless sensor network

Data standardizing

GPRS

Firewall

Figure 3 Archetype of CPP-WSN

Table 2 Normalization coefficients of RadNet parameters

Sensor 1 Sensor 2 Sensor 3 Sensor 4 Sensor 5 Sensor 6UVR 09972 10279 09987 09868 09796 10113SWR 09734 09994 10053 09990 10264 09989LWR 10161 10189 09973 09932 09938 09818PAR 10177 09422 10030 10391 09557 10522LST 10019 09965 10014 09986 10007 10009


Solar panel

Data collector

Soil sensors

Radiation sensors

Camera sensor

(a) (b)

Bracket = 25 m

Bracket = 4m

Figure 4 Scene of RadNetVegeNet node (a) and SoilNet node (b)

the height of canopy for solar energy supplyThus the bracketheight of SoilNet node is set to 25m because the maximumcanopy height of observed area is about 2mwhile the bracketheight of RadNetVegeNet node with above canopy bracketheight of wide field of view radiometer is set to 4m tosupply a large enough field of view for effective parametersobservation

The observing period depends on the variation frequencyof parameter for example the instantaneous LST mea-surement for synchronously validating the remote sensingretrieved LST changes rapidly However the power supplyof solar panel is limited Therefore there is a compromisestrategy between the observing period and power supplylimitation The observing period of RadNet and SoilNet isset to 5 minutes which is a typical frequency of global fluxobservation station to ensure a sustainable continuous obser-vation in day and night The VegeNet camera sensor workstwice a day at 12 orsquoclock and 17 orsquoclock respectively Aftersampling the observed voltages are preliminarily processedto parameters

322 Optimized Layout to Capture the Land Surface Het-erogeneity Different categories of parameters show differentheterogeneity within coarse pixel According to the influenceof underground water soil texture irrigation conditionand vegetation the soil moisture and temperature have thehighest spatial variation Therefore dense grid layout shouldbe selected for SoilNet nodes especially considering the lackof a priori information of soil moisture and temperature dis-tribution in the study area The radiation (including albedo)and vegetation parameters are mainly determined by theplanting structure and vegetation growth and it is preferredthat they share the same layout for relevance and mutuallysupportive study In contrast to soil parameters the radiationand vegetation parameters have lower spatial variation and

093

094

095

096

097

098

099

1

Cum

ulat

ive r

epre

sent

ativ

enes

s

2 3 4 5 6 7 8 9 101Point

Figure 5 Cumulative representativeness for the 1 km pixel withdifferent node number

abundant a priori information retrieved from remote sensingimage So the heterogeneity of RadNetVegeNet could becaptured using fewer nodes As the radiation sensors (CUV5CNR4 PQS1 and SI-111) are expensive and fragile it is alsoa practical need to reduce the number of RadNet nodesso long as the radiation parameters at the coarse-resolutionpixel scale can be satisfactorily represented by the nodesIn this consideration the representativeness-based samplingmethod [23] is adopted to determine the node number andlocation of the RadNetVegeNet in CPP-WSN The methodselects the point of highest representativeness as the locationfor WSN node and multiple nodes are selected one by oneto increase the cumulative representativeness for 1 km pixelFigure 5 reveals the cumulative representativeness for 1 kmpixel increases with node number and approaches a steadystate after the sixth point Therefore the coarse-resolutionpixel could be well-represented by six nodes with cumulativerepresentativeness high to 991


(km)

(km)

N

N

N

SoilNet nodeRedNet node1km SIN pixelStudy area

Spot 5 2012-08-29RGB

Red band 3

Green band 4

Blue band 2

115∘44998400E 115∘46998400E 115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘50998400E115∘42998400E

115∘42998400E 115∘44998400E 115∘46998400E 115∘48998400E 115∘50998400E

(km)

40∘24998400 N

40∘26998400 N

40∘22998400 N

40∘20998400 N

40∘16998400 N

40∘18998400 N

40∘18998400 N

40∘20998400 N

40∘22998400 N

40∘24998400 N

40∘26998400 N

40∘16998400 N

4321050

10750502501250

(km)10750502501250

Figure 6 Sampling layout of the RadNetVegeNet and SoilNet

As shown in Figure 6 6-node RadNetVegeNet and 21-node SoilNet were set up respectively The red points showthe layout of RadNetVegeNet and the green points showthe layout of SoilNet The 1 km pixel ldquotruerdquo value of radiationparameters could be obtained by the weighted average of themeasurements at the 6 nodes and the weights are calculatedaccording to the representativeness as well as covariance ofthe nodesThe SoilNet consists of both the gridded nodes andthe nodes attached with RadNet The 1 km pixel ldquotruerdquo valueof soil parameters could be yielded by the simple averageor the geostatistic interpolation of the measurements at allnodes

323 Automatic Data Transmission and Sharing Near real-time data transmission is not only important for data userbut also essential for WSN maintenance to monitor the nodestatus and track bugs GPRS is best-effort service implyingvariable throughput and latency that depend on the numberof other users sharing the service concurrently as opposed to

circuit switching where a certain quality of service (QoS) isguaranteed during the connection [24] It is one of the generalways for real-time and stablewireless data transmission Sincethe data amount of CPP-WSN is large compared with generalwireless sensor networks a 3G network via GPRS is adopted

For energy saving the WSN nodes send observationdata once per hour and the data are listened and received byremote data server After the data are received a time seriescontinuity check is performed and a gap-filling methodis taken to make time-continuous data by identifying andfilling the missing or invalid data A further quality controlprocess is conducted to check if the data is effective andmark the data with a quality flag The data of RadNet andSoilNet are packaged and their self-description metadataare produced every month The data of VegeNet is truecolor photo of the canopy it is further processed to separategreen leaf from background and then used to estimatethe FVC LAI and CI A dataset is composed of both thedata and metadata and shared via an open access website(httprsesdslrsscnWSNProjbusinessproductdatalistjsp)


Full inversion

Magnitude inversion

Overall accuracy

Full inversionMagnitude inversion

RMSE = 0025 R2 = 071

RMSE = 0020 R2 = 071

RMSE = 0023

R2 = 072

0

01

02

03

04

05M

CD43

B3

01 02 03 04 05016-day ground albedo truth at pixel scale

(a) MCD43B3

Strict clear sky

Clear sky

Overall accuracy

Clear skyStrict clear sky

01 02 03 04 050In situ truth at VIIRS pixel scale

0

01

02

03

04

05

VII

RS in

vers

ion

RMSE = 0014

R2 = 0802Bias = 0001

RMSE = 0021

R2 = 0553Bias = 0005

(b) VIIRS

Overall accuracy

GLASS inversion

01 02 03 040In situ truth at 1km pixel scale

0

01

02

03

04

GLA

SS in

vers

ion

Bias = minus0013R2 = 0557

RMSE = 0021

(c) GLASS

Figure 7 Validation of three land surface albedo products

4 Preliminary Applications of CPP-WSN Data

The main objective of CPP-WSN is to provide groundldquotruthrdquo at the coarse pixel scale over heterogeneous landsurface to assess the accuracy of remote sensing products oralgorithms The RadNet data has been available since May2013 and the SoilNet has been available sinceDecember 2012Many researchers have used the data for various validationactivities

41 Remote Sensing Albedo Products Validation Land sur-face albedo observed by RadNet was upscaled to satel-lite pixel scale using weighted average according to therepresentativeness-based sampling method and then vali-dated three albedo products that is MCD43B3 GLASSalbedo and NPP VIIRS albedo products in 2014 Figure 7(a)shows the scatterplot of the validation of 16-day 1 km pixelscale in situ observation andmeanMCD43B3 albedowhich iswith 1 km spatial resolution and 16-day temporal composition


Retrieval PAR in cloudy sky Retrieval PAR in clear sky

y = 09629x + 24821

R2 = 08014

MBE = minus1921397

RMSE = 6378553

y = 09993x minus 13494

R2 = 0958

MBE = minus13536

RMSE = 3262103

0

50

100

150

200

250

300

350

400

450

500

Retr

ieva

l PA

R (W

m2 )

500100 150 200 250 300 350 400 450500

Measured PAR (Wm2)

(a)

y = 10333x + 11627

R2 = 09344

RMSE = 2513 Wm2

350100 150 200 250 300500

Measured PAR (Wm2)

0

50

100

150

200

250

300

350

Retr

ieve

d PA

R (W

m2 )

(b)

Figure 8 Validation of PAR algorithm [8]

windowTheWSNmeasurement is also upscaled to 1 kmpixeland averaged on the 16 days corresponding to MCD43B3product Figure 7(b) shows the scatterplot of the validationof VIIRS albedo which is in 750m spatial resolution and 1-day temporal resolution Only the inversion results on clearsky conditions are selected The in situ WSN measurementsare processed to the identical spatial and temporal resolutionFigure 7(c) shows the scatterplot between 1 km pixel scalein situ observation and GLASS daily albedo in clear skyconditions All the three albedo products show similar andacceptable accuracy and the RMSEs are 0023 0021 and0021 respectively for MCD43B3 VIIRS and GLASS VIIRSalbedo on ldquostrict clear daysrdquo shows less RMSE of 0014 thanthat of the ldquoclear daysrdquo of this site [25]

42 Remote Sensing PAR Algorithm Validation ObservedPAR data from RadNet was compared with the instanta-neous PAR estimated from the geostationary and polar-orbiting satellite data in spatial resolution of 5 km Figure 8(a)shows the scatterplot between derived instantaneous PARand measured instantaneous PAR under different climateconditions Figure 8(b) shows the scatterplot between themeasured instantaneous PAR and the retrieved instantaneousPAR according tomeasuredAOD in clear sky conditionsTheaccuracy of retrieved PAR in clear sky is higher than retrievedPAR in cloudy sky With the assistance of measured AOD inclear sky the retrieved instantaneous PAR gets even higheraccuracy and the RMSE decreases to 2513 Wm2 [8]

43 Remote Sensing LST Products Validation Themultipointmeasurements of ground LST were averaged as the pixelscale value to validate different LST products includingNPP VIIRS LST Terra MODIS LST and Aqua MODIS LSTFigure 9(a) shows the scatterplot between ground LST andNPP VIIRS LST Figure 9(b) shows the scatterplot betweenground LST and Terra MODIS LST Figure 9(c) shows

the scatterplot between ground LST and Aqua MODIS LSTThe daytime and nighttime LST were verified separatelyThe NPP VIIRS LST shows higher accuracy than the TerraMODIS LST and Aqua MODIS LST of this site

44 Leaf Area Index (LAI) Products Validation VegeNet atHuailai Station was used to validate three MODIS C5 LAIproducts including MOD15A2 MYD15A2 and MCD15A2Figure 10 shows the time series comparison between the pixelscale LAI reference values derived from WSN measurementand the corresponding LAI of the three MODIS productsThe accuracy of MOD15A2 and MCD15A2 is higher thanMYD15A2 of this siteThemeanRMSEofMODIS LAI is 033and the relative uncertainty is 122 [9]

5 Conclusions

The CPP-WSN with consideration of the heterogeneity ofdifferent parameters was established for the research ofparametersrsquo scale effect and the validation of remote sensingproducts And it achieves the near real-time data trans-mission and standard data sharing for three categories ofparameters Time series observations of land surface typicalparameters including UVR PAR SWR LWR albedo NRand LST from RadNet multilayer soil moisture and soiltemperature from SoilNet and FVC LAI and CI fromVegeNet have been obtained and shared online The WSNmeasurements with multiple nodes can capture the spatialvariation of parameters in heterogeneous low-resolutionpixel Based on optimized layout and precalculated weightsthe parameters measured with WSN can be upscaled to thescale of low-resolution pixels So it provides a more accuratereference value for validating the remote sensing productsthan the traditional single-point value

The dataset has been shared and many researchers havestarted to use it to assess a variety of remote sensing


Bias = 090

STD = 141

N = 25

Bias = minus122

STD = 094

N = 27

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

VII

RS L

ST (K

)

Daytime Nighttime

(a) NPP VIIRS

STD = 149

N = 31

Bias = minus164Bias = minus233

STD = 085

N = 23

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Terr

a MO

DIS

LST

(K)

Daytime Nighttime

(b) Terra MODIS

STD = 124

N = 45


STD = 103

N = 18

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Aqua

MO

DIS

LST

(K)

Daytime Nighttime

(c) Aqua MODIS

Figure 9 Validation of three land surface temperature products

products The preliminary results show good consistencybetween remote sensing products and the reference valuesacquired with CPP-WSN indicating that the system is run-ning smoothly and fulfills its scientific goal However moresites and longer time series of good quality observations

are needed for comprehensive validation of remote sensingproducts and strict calibration for the WSN sensors andimprovement to the upscaling scheme are also needed tosupport the high quality of the reference values in coarse pixelscale These show our further research direction


MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2

Figure 10 Validation of three MODIS LAI products [9]

Competing Interests

The authors declare that there is no conflict of interestsregarding the publication of this paper

Acknowledgments

This research was jointly supported by National BasicResearch Program of China (2013CB733401) Chinese Nat-ural Science Foundation Project (41271368) and NationalHigh Technology Research and Development Program ofChina (2013AA12A301) Additionally the authors would liketo thank all of the scientists engineers and students whoparticipated in the maintenance of CPP-WSN especially DrBai Junhua at the Institute of Remote Sensing and DigitalEarth of the Chinese Academy of Sciences

References

[1] C Justice D Starr DWickland J Privette and T Suttles ldquoEOSland validation coordination an updaterdquo Earth Observer vol10 no 3 pp 55ndash60 1998

[2] C Justice A Belward J Morisette P Lewis J Privette andF Baret ldquoDevelopments in the ldquovalidationrdquo of satellite sensorproducts for the study of the land surfacerdquo International Journalof Remote Sensing vol 21 no 17 pp 3383ndash3390 2000

[3] S W Running D D Baldocchi D P Turner S T Gower PS Bakwin and K A Hibbard ldquoA global terrestrial monitoringnetwork integrating tower fluxes flask sampling ecosystemmodeling and EOS satellite datardquo Remote Sensing of Environ-ment vol 70 no 1 pp 108ndash127 1999

[4] D Wu H Wu X Zhao et al ldquoEvaluation of spatiotempo-ral variations of global fractional vegetation cover based on


GIMMS NDVI data from 1982 to 2011rdquo Remote Sensing vol 6no 5 pp 4217ndash4239 2014

[5] J T Morisette F Baret J L Privette et al ldquoValidation of globalmoderate-resolution LAI products a framework proposedwithin the CEOS land product validation subgrouprdquo IEEETransactions on Geoscience and Remote Sensing vol 44 no 7pp 1804ndash1814 2006

[6] R H Zhang J Tian Z L Li H B Su S H Chen and X ZTang ldquoPrinciples andmethods for the validation of quantitativeremote sensing productsrdquo Science China Earth Sciences vol 53no 5 pp 741ndash751 2010

[7] M Ma T Che X Li Q Xiao K Zhao and X Xin ldquoAprototype network for remote sensing validation in ChinardquoRemote Sensing vol 7 no 5 pp 5187ndash5202 2015

[8] L Li X Xin H Zhang et al ldquoA method for estimatinghourly photosynthetically active radiation (PAR) in Chinaby combining geostationary and polar-orbiting satellite datardquoRemote Sensing of Environment vol 165 pp 14ndash26 2015

[9] Y Zeng J Li Q Liu et al ldquoAn optimal sampling designfor observing and validating long-term leaf area index withtemporal variations in spatial heterogeneitiesrdquo Remote Sensingvol 7 no 2 pp 1300ndash1319 2015

[10] S L Collins L M A Bettencourt A Hagberg et al ldquoNewopportunities in ecological sensing using wireless sensor net-worksrdquo Frontiers in Ecology and the Environment vol 4 no 8pp 402ndash407 2006

[11] J K Hart and K Martinez ldquoEnvironmental sensor networks arevolution in the earth system sciencerdquo Earth-Science Reviewsvol 78 no 3-4 pp 177ndash191 2006

[12] R N Handcock D L Swain G J Bishop-Hurley et alldquoMonitoring animal behaviour and environmental interactionsusing wireless sensor networks GPS collars and satellite remotesensingrdquo Sensors vol 9 no 5 pp 3586ndash3603 2009

[13] B Krishnamachari Networking Wireless Sensors CambridgeUniversity Press 2005

[14] G Peng ldquoWireless sensor network as a new ground remotesensing technology for environmental monitoringrdquo Journal ofRemote Sensing vol 11 no 4 pp 545ndash551 2007

[15] X Li G Cheng S Liu et al ldquoHeihe watershed allied teleme-try experimental research (HiWater) scientific objectives andexperimental designrdquo Bulletin of the American MeteorologicalSociety vol 94 no 8 pp 1145ndash1160 2013

[16] L Fan Q Xiao J Wen et al ldquoEvaluation of the airborneCASITASI Ts-VI spacemethod for estimating near-surface soilmoisturerdquo Remote Sensing vol 7 no 3 pp 3114ndash3137 2015

[17] D You JWen Q Xiao et al ldquoDevelopment of a high resolutionBRDFalbedo product by fusing airborne CASI reflectance withMODIS daily reflectance in the oasis area of the Heihe RiverBasin Chinardquo Remote Sensing vol 7 no 6 pp 6784ndash6807 2015

[18] A R G Lang and X Yueqin ldquoEstimation of leaf area indexfrom transmission of direct sunlight in discontinuous canopiesrdquoAgricultural and Forest Meteorology vol 37 no 3 pp 229ndash2431986

[19] X Li Q Liu R Yang H Zhang J Zhang and E Cai ldquoThedesign and implementation of the leaf area index sensorrdquoSensors vol 15 no 3 pp 6250ndash6269 2015

[20] D Nie ldquoAn intercomparison of surface energy flux measure-ment systems used during FIFE 1987rdquo Journal of GeophysicalResearch vol 97 no 17 pp 715ndash724 1992

[21] W Kohsiek C Liebethal T Foken et al ldquoThe Energy BalanceExperiment EBEX-2000 Part III behaviour and quality of

the radiationmeasurementsrdquo Boundary-LayerMeteorology vol123 no 1 pp 55ndash75 2007

[22] Z W Xu S M Liu X Li et al ldquoIntercomparison of surfaceenergy flux measurement systems used during the HiWATER-MUSOEXErdquo Journal of Geophysical Research Atmospheres vol118 no 23 pp 13-140ndash13-157 2013

[23] J Peng Validation of pixel-scale land surface albedo products[PhD thesis] University of Chinese Academy of Sciences 2014

[24] General Packet Radio Service Wikipedia 2013 June 2015httpsenwikipediaorgwikiGeneral Packet Radio Service

[25] X Wu J Wen Q Xiao et al ldquoRemote sensing albedo productvalidation over heterogenicity surface based on WSN prelimi-nary results and its uncertaintyrdquo in Proceedings of the SPIE LandSurface Remote Sensing II International Society for Optics andPhotonics Beijing China 2014

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



Study area1km SIN pixel

(km)

(km)

(km)

01250 025 05 075 1

115∘48998400E

115∘48998400E

20∘

30∘

40∘

30∘

40∘

50∘

South China Sea Islands200 0 200 400

N

N

Heihe S Station nameForest Typical land surface typesLand cover type

Evergreen needleleaf forestEvergreen broadleaf forestDeciduous needleleaf forestDeciduous broadleaf forestMixed forestClosed shrublandsOpen shrublandsWoody savannasSavannas

GrasslandsPermanent wetlandsCroplandsUrban and built-up landsCroplandnatural vegetation mosaicSnow and iceBarren or sparsely vegetated landsBodies of water

20∘

15∘

10∘

5∘

120∘115∘110∘

140∘130∘120∘110∘100∘90∘80∘70∘

130∘120∘110∘100∘90∘

8004000400

Figure 1 Location and underlying characteristics of research area (revised from [7])

Wireless sensor network (WSN) which is typically com-prised of a collection of sensors with their own power supplywireless communication data storage and data processingcapability collects and transmits data from remote field sitesback to base station [10ndash12] Networks of embedded devicesthat work together to provide enhanced monitoring acrossspatial and temporal scales are growing in popularity [13]Wireless sensor networks are increasingly applied in researchfields including earth observation environmental monitor-ing agriculture resource management public health publicsecurity transportation and military [14] With the ability oflarge-scale coverage and long time series running WSN hasthe potential to contribute to the solution of the scale effectchallenge in remote sensing research

Recently new remote sensing products of energy bud-get vegetation characteristics and soil characteristics havebeen generated to meet the needs in the climate changeenvironment monitoring and agricultural yield estimationWSN is an effective way for simultaneously monitoring theseparameters at ground With its flexibility automation andrelatively low cost to synchronously acquire data in multiplesites the WSN is recently adopted in validationcalibrationof remote sensing products [15ndash17] For this purpose it isdesirable that each node could contain multiple sensors tomeasure different kinds of parameters In this paper weexplore the establishment of the wireless sensor networkswhich monitor the heterogeneity of coarse spatial resolutionpixel with consideration of different categories of landsurface parameters (hereafter referred to as CPP-WSN)

This paper firstly presents the China validation networkand Huailai remote sensing station where the CPP-WSNlies Then the parameter selection and sensor specificationsof CPP-WSN are discussed Construction of CPP-WSNincluding design of CPP-WSN nodes optimized layout atpixel scale and automatic data transmission and sharing

are described in the third part of this paper And finallythe validations of a variety of remote sensing products areintroduced as preliminary applications of the CPP-WSN

2 Huailai Remote Sensing Station

There are many ecohydrology and flux sites around China(Figure 1) and these sites have provided various data forenvironmental monitoring and remote sensing products val-idation With advanced observational techniques abundantdata accumulation and ability of carrying on multiscaleexperiment the Huailai Remote Sensing Station and around(for short HuailaiS) located in Huailai Hebei provinceChina (40349∘N 115785∘E) becomes one of the ideal studyareas for remote sensing algorithmsproducts validation andscale effect research The HuailaiS is mainly covered by irri-gated corn and unirrigated cornThe differences in irrigationcondition and soil type distribution make the single landcover area heterogeneous but continuously changing whichcharacterize the problem of scale transformation from point-measured parameters to pixel scale reference value

The sinusoidal grid is one of the most widely usedgrid schemes for global remote sensing products For 1 kmsinusoidal pixel covering HuailaiS (the green parallelogramshown on the right of Figure 1) a 2 km lowast 15 km layout areafor CPP-WSN (the red square shown on the right of Figure 1)is chosen in consideration of the adjacent pixelsrsquo influence

3 CPP-WSN Observations and Data Collection

31 Specifications of CPP-WSN and Sensor IntercomparisonConsidering the needs of validating various remote sensingproducts and the different heterogeneity conditions of differ-ent parameters three categories of parameters are observed










Camera





119870 =

sum119899

1

(VV)119899

(1)








ortw

ave i

rrad

ianc

e (W

lowastm

minus2 )

200

400

600

800

1000

151

400

162

900

152

900

154

400

155

900

161

400

145

900

144

400

Time

CNR4_1CNR4_2CNR4_3

CNR4_4CNR4_5CNR4_6



WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node




Data standardizing

GPRS

Firewall





Solar panel

Data collector

Soil sensors

Radiation sensors

Camera sensor

(a) (b)

Bracket = 25 m

Bracket = 4m





093

094

095

096

097

098

099

1

Cum

ulat

ive r

epre

sent

ativ

enes

s

2 3 4 5 6 7 8 9 101Point




(km)

(km)

N

N

N


Spot 5 2012-08-29RGB

Red band 3

Green band 4

Blue band 2

115∘44998400E 115∘46998400E 115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘50998400E115∘42998400E

115∘42998400E 115∘44998400E 115∘46998400E 115∘48998400E 115∘50998400E

(km)

40∘24998400 N

40∘26998400 N

40∘22998400 N

40∘20998400 N

40∘16998400 N

40∘18998400 N

40∘18998400 N

40∘20998400 N

40∘22998400 N

40∘24998400 N

40∘26998400 N

40∘16998400 N

4321050

10750502501250

(km)10750502501250







Full inversion

Magnitude inversion

Overall accuracy


RMSE = 0025 R2 = 071

RMSE = 0020 R2 = 071

RMSE = 0023

R2 = 072

0

01

02

03

04

05M

CD43

B3


(a) MCD43B3

Strict clear sky

Clear sky

Overall accuracy



0

01

02

03

04

05

VII

RS in

vers

ion

RMSE = 0014

R2 = 0802Bias = 0001

RMSE = 0021

R2 = 0553Bias = 0005

(b) VIIRS

Overall accuracy

GLASS inversion


0

01

02

03

04

GLA

SS in

vers

ion


RMSE = 0021

(c) GLASS







y = 09629x + 24821

R2 = 08014

MBE = minus1921397

RMSE = 6378553

y = 09993x minus 13494

R2 = 0958

MBE = minus13536

RMSE = 3262103

0

50

100

150

200

250

300

350

400

450

500

Retr

ieva

l PA

R (W

m2 )

500100 150 200 250 300 350 400 450500

Measured PAR (Wm2)

(a)

y = 10333x + 11627

R2 = 09344

RMSE = 2513 Wm2

350100 150 200 250 300500

Measured PAR (Wm2)

0

50

100

150

200

250

300

350

Retr

ieve

d PA

R (W

m2 )

(b)







5 Conclusions




Bias = 090

STD = 141

N = 25

Bias = minus122

STD = 094

N = 27

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

VII

RS L

ST (K

)

Daytime Nighttime

(a) NPP VIIRS

STD = 149

N = 31


STD = 085

N = 23

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Terr

a MO

DIS

LST

(K)

Daytime Nighttime

(b) Terra MODIS

STD = 124

N = 45


STD = 103

N = 18

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Aqua

MO

DIS

LST

(K)

Daytime Nighttime

(c) Aqua MODIS





MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2


Competing Interests


Acknowledgments


References































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation


















Camera





119870 =

sum119899

1

(VV)119899

(1)








ortw

ave i

rrad

ianc

e (W

lowastm

minus2 )

200

400

600

800

1000

151

400

162

900

152

900

154

400

155

900

161

400

145

900

144

400

Time

CNR4_1CNR4_2CNR4_3

CNR4_4CNR4_5CNR4_6



WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node




Data standardizing

GPRS

Firewall





Solar panel

Data collector

Soil sensors

Radiation sensors

Camera sensor

(a) (b)

Bracket = 25 m

Bracket = 4m





093

094

095

096

097

098

099

1

Cum

ulat

ive r

epre

sent

ativ

enes

s

2 3 4 5 6 7 8 9 101Point




(km)

(km)

N

N

N


Spot 5 2012-08-29RGB

Red band 3

Green band 4

Blue band 2

115∘44998400E 115∘46998400E 115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘50998400E115∘42998400E

115∘42998400E 115∘44998400E 115∘46998400E 115∘48998400E 115∘50998400E

(km)

40∘24998400 N

40∘26998400 N

40∘22998400 N

40∘20998400 N

40∘16998400 N

40∘18998400 N

40∘18998400 N

40∘20998400 N

40∘22998400 N

40∘24998400 N

40∘26998400 N

40∘16998400 N

4321050

10750502501250

(km)10750502501250







Full inversion

Magnitude inversion

Overall accuracy


RMSE = 0025 R2 = 071

RMSE = 0020 R2 = 071

RMSE = 0023

R2 = 072

0

01

02

03

04

05M

CD43

B3


(a) MCD43B3

Strict clear sky

Clear sky

Overall accuracy



0

01

02

03

04

05

VII

RS in

vers

ion

RMSE = 0014

R2 = 0802Bias = 0001

RMSE = 0021

R2 = 0553Bias = 0005

(b) VIIRS

Overall accuracy

GLASS inversion


0

01

02

03

04

GLA

SS in

vers

ion


RMSE = 0021

(c) GLASS







y = 09629x + 24821

R2 = 08014

MBE = minus1921397

RMSE = 6378553

y = 09993x minus 13494

R2 = 0958

MBE = minus13536

RMSE = 3262103

0

50

100

150

200

250

300

350

400

450

500

Retr

ieva

l PA

R (W

m2 )

500100 150 200 250 300 350 400 450500

Measured PAR (Wm2)

(a)

y = 10333x + 11627

R2 = 09344

RMSE = 2513 Wm2

350100 150 200 250 300500

Measured PAR (Wm2)

0

50

100

150

200

250

300

350

Retr

ieve

d PA

R (W

m2 )

(b)







5 Conclusions




Bias = 090

STD = 141

N = 25

Bias = minus122

STD = 094

N = 27

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

VII

RS L

ST (K

)

Daytime Nighttime

(a) NPP VIIRS

STD = 149

N = 31


STD = 085

N = 23

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Terr

a MO

DIS

LST

(K)

Daytime Nighttime

(b) Terra MODIS

STD = 124

N = 45


STD = 103

N = 18

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Aqua

MO

DIS

LST

(K)

Daytime Nighttime

(c) Aqua MODIS





MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2


Competing Interests


Acknowledgments


References































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











ortw

ave i

rrad

ianc

e (W

lowastm

minus2 )

200

400

600

800

1000

151

400

162

900

152

900

154

400

155

900

161

400

145

900

144

400

Time

CNR4_1CNR4_2CNR4_3

CNR4_4CNR4_5CNR4_6



WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node

WSN node




Data standardizing

GPRS

Firewall





Solar panel

Data collector

Soil sensors

Radiation sensors

Camera sensor

(a) (b)

Bracket = 25 m

Bracket = 4m





093

094

095

096

097

098

099

1

Cum

ulat

ive r

epre

sent

ativ

enes

s

2 3 4 5 6 7 8 9 101Point




(km)

(km)

N

N

N


Spot 5 2012-08-29RGB

Red band 3

Green band 4

Blue band 2

115∘44998400E 115∘46998400E 115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘50998400E115∘42998400E

115∘42998400E 115∘44998400E 115∘46998400E 115∘48998400E 115∘50998400E

(km)

40∘24998400 N

40∘26998400 N

40∘22998400 N

40∘20998400 N

40∘16998400 N

40∘18998400 N

40∘18998400 N

40∘20998400 N

40∘22998400 N

40∘24998400 N

40∘26998400 N

40∘16998400 N

4321050

10750502501250

(km)10750502501250







Full inversion

Magnitude inversion

Overall accuracy


RMSE = 0025 R2 = 071

RMSE = 0020 R2 = 071

RMSE = 0023

R2 = 072

0

01

02

03

04

05M

CD43

B3


(a) MCD43B3

Strict clear sky

Clear sky

Overall accuracy



0

01

02

03

04

05

VII

RS in

vers

ion

RMSE = 0014

R2 = 0802Bias = 0001

RMSE = 0021

R2 = 0553Bias = 0005

(b) VIIRS

Overall accuracy

GLASS inversion


0

01

02

03

04

GLA

SS in

vers

ion


RMSE = 0021

(c) GLASS







y = 09629x + 24821

R2 = 08014

MBE = minus1921397

RMSE = 6378553

y = 09993x minus 13494

R2 = 0958

MBE = minus13536

RMSE = 3262103

0

50

100

150

200

250

300

350

400

450

500

Retr

ieva

l PA

R (W

m2 )

500100 150 200 250 300 350 400 450500

Measured PAR (Wm2)

(a)

y = 10333x + 11627

R2 = 09344

RMSE = 2513 Wm2

350100 150 200 250 300500

Measured PAR (Wm2)

0

50

100

150

200

250

300

350

Retr

ieve

d PA

R (W

m2 )

(b)







5 Conclusions




Bias = 090

STD = 141

N = 25

Bias = minus122

STD = 094

N = 27

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

VII

RS L

ST (K

)

Daytime Nighttime

(a) NPP VIIRS

STD = 149

N = 31


STD = 085

N = 23

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Terr

a MO

DIS

LST

(K)

Daytime Nighttime

(b) Terra MODIS

STD = 124

N = 45


STD = 103

N = 18

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Aqua

MO

DIS

LST

(K)

Daytime Nighttime

(c) Aqua MODIS





MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2


Competing Interests


Acknowledgments


References































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Solar panel

Data collector

Soil sensors

Radiation sensors

Camera sensor

(a) (b)

Bracket = 25 m

Bracket = 4m





093

094

095

096

097

098

099

1

Cum

ulat

ive r

epre

sent

ativ

enes

s

2 3 4 5 6 7 8 9 101Point




(km)

(km)

N

N

N


Spot 5 2012-08-29RGB

Red band 3

Green band 4

Blue band 2

115∘44998400E 115∘46998400E 115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘50998400E115∘42998400E

115∘42998400E 115∘44998400E 115∘46998400E 115∘48998400E 115∘50998400E

(km)

40∘24998400 N

40∘26998400 N

40∘22998400 N

40∘20998400 N

40∘16998400 N

40∘18998400 N

40∘18998400 N

40∘20998400 N

40∘22998400 N

40∘24998400 N

40∘26998400 N

40∘16998400 N

4321050

10750502501250

(km)10750502501250







Full inversion

Magnitude inversion

Overall accuracy


RMSE = 0025 R2 = 071

RMSE = 0020 R2 = 071

RMSE = 0023

R2 = 072

0

01

02

03

04

05M

CD43

B3


(a) MCD43B3

Strict clear sky

Clear sky

Overall accuracy



0

01

02

03

04

05

VII

RS in

vers

ion

RMSE = 0014

R2 = 0802Bias = 0001

RMSE = 0021

R2 = 0553Bias = 0005

(b) VIIRS

Overall accuracy

GLASS inversion


0

01

02

03

04

GLA

SS in

vers

ion


RMSE = 0021

(c) GLASS







y = 09629x + 24821

R2 = 08014

MBE = minus1921397

RMSE = 6378553

y = 09993x minus 13494

R2 = 0958

MBE = minus13536

RMSE = 3262103

0

50

100

150

200

250

300

350

400

450

500

Retr

ieva

l PA

R (W

m2 )

500100 150 200 250 300 350 400 450500

Measured PAR (Wm2)

(a)

y = 10333x + 11627

R2 = 09344

RMSE = 2513 Wm2

350100 150 200 250 300500

Measured PAR (Wm2)

0

50

100

150

200

250

300

350

Retr

ieve

d PA

R (W

m2 )

(b)







5 Conclusions




Bias = 090

STD = 141

N = 25

Bias = minus122

STD = 094

N = 27

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

VII

RS L

ST (K

)

Daytime Nighttime

(a) NPP VIIRS

STD = 149

N = 31


STD = 085

N = 23

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Terr

a MO

DIS

LST

(K)

Daytime Nighttime

(b) Terra MODIS

STD = 124

N = 45


STD = 103

N = 18

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Aqua

MO

DIS

LST

(K)

Daytime Nighttime

(c) Aqua MODIS





MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2


Competing Interests


Acknowledgments


References































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










(km)

(km)

N

N

N


Spot 5 2012-08-29RGB

Red band 3

Green band 4

Blue band 2

115∘44998400E 115∘46998400E 115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘48998400E

115∘50998400E115∘42998400E

115∘42998400E 115∘44998400E 115∘46998400E 115∘48998400E 115∘50998400E

(km)

40∘24998400 N

40∘26998400 N

40∘22998400 N

40∘20998400 N

40∘16998400 N

40∘18998400 N

40∘18998400 N

40∘20998400 N

40∘22998400 N

40∘24998400 N

40∘26998400 N

40∘16998400 N

4321050

10750502501250

(km)10750502501250







Full inversion

Magnitude inversion

Overall accuracy


RMSE = 0025 R2 = 071

RMSE = 0020 R2 = 071

RMSE = 0023

R2 = 072

0

01

02

03

04

05M

CD43

B3


(a) MCD43B3

Strict clear sky

Clear sky

Overall accuracy



0

01

02

03

04

05

VII

RS in

vers

ion

RMSE = 0014

R2 = 0802Bias = 0001

RMSE = 0021

R2 = 0553Bias = 0005

(b) VIIRS

Overall accuracy

GLASS inversion


0

01

02

03

04

GLA

SS in

vers

ion


RMSE = 0021

(c) GLASS







y = 09629x + 24821

R2 = 08014

MBE = minus1921397

RMSE = 6378553

y = 09993x minus 13494

R2 = 0958

MBE = minus13536

RMSE = 3262103

0

50

100

150

200

250

300

350

400

450

500

Retr

ieva

l PA

R (W

m2 )

500100 150 200 250 300 350 400 450500

Measured PAR (Wm2)

(a)

y = 10333x + 11627

R2 = 09344

RMSE = 2513 Wm2

350100 150 200 250 300500

Measured PAR (Wm2)

0

50

100

150

200

250

300

350

Retr

ieve

d PA

R (W

m2 )

(b)







5 Conclusions




Bias = 090

STD = 141

N = 25

Bias = minus122

STD = 094

N = 27

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

VII

RS L

ST (K

)

Daytime Nighttime

(a) NPP VIIRS

STD = 149

N = 31


STD = 085

N = 23

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Terr

a MO

DIS

LST

(K)

Daytime Nighttime

(b) Terra MODIS

STD = 124

N = 45


STD = 103

N = 18

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Aqua

MO

DIS

LST

(K)

Daytime Nighttime

(c) Aqua MODIS





MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2


Competing Interests


Acknowledgments


References































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Full inversion

Magnitude inversion

Overall accuracy


RMSE = 0025 R2 = 071

RMSE = 0020 R2 = 071

RMSE = 0023

R2 = 072

0

01

02

03

04

05M

CD43

B3


(a) MCD43B3

Strict clear sky

Clear sky

Overall accuracy



0

01

02

03

04

05

VII

RS in

vers

ion

RMSE = 0014

R2 = 0802Bias = 0001

RMSE = 0021

R2 = 0553Bias = 0005

(b) VIIRS

Overall accuracy

GLASS inversion


0

01

02

03

04

GLA

SS in

vers

ion


RMSE = 0021

(c) GLASS







y = 09629x + 24821

R2 = 08014

MBE = minus1921397

RMSE = 6378553

y = 09993x minus 13494

R2 = 0958

MBE = minus13536

RMSE = 3262103

0

50

100

150

200

250

300

350

400

450

500

Retr

ieva

l PA

R (W

m2 )

500100 150 200 250 300 350 400 450500

Measured PAR (Wm2)

(a)

y = 10333x + 11627

R2 = 09344

RMSE = 2513 Wm2

350100 150 200 250 300500

Measured PAR (Wm2)

0

50

100

150

200

250

300

350

Retr

ieve

d PA

R (W

m2 )

(b)







5 Conclusions




Bias = 090

STD = 141

N = 25

Bias = minus122

STD = 094

N = 27

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

VII

RS L

ST (K

)

Daytime Nighttime

(a) NPP VIIRS

STD = 149

N = 31


STD = 085

N = 23

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Terr

a MO

DIS

LST

(K)

Daytime Nighttime

(b) Terra MODIS

STD = 124

N = 45


STD = 103

N = 18

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Aqua

MO

DIS

LST

(K)

Daytime Nighttime

(c) Aqua MODIS





MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2


Competing Interests


Acknowledgments


References































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











y = 09629x + 24821

R2 = 08014

MBE = minus1921397

RMSE = 6378553

y = 09993x minus 13494

R2 = 0958

MBE = minus13536

RMSE = 3262103

0

50

100

150

200

250

300

350

400

450

500

Retr

ieva

l PA

R (W

m2 )

500100 150 200 250 300 350 400 450500

Measured PAR (Wm2)

(a)

y = 10333x + 11627

R2 = 09344

RMSE = 2513 Wm2

350100 150 200 250 300500

Measured PAR (Wm2)

0

50

100

150

200

250

300

350

Retr

ieve

d PA

R (W

m2 )

(b)







5 Conclusions




Bias = 090

STD = 141

N = 25

Bias = minus122

STD = 094

N = 27

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

VII

RS L

ST (K

)

Daytime Nighttime

(a) NPP VIIRS

STD = 149

N = 31


STD = 085

N = 23

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Terr

a MO

DIS

LST

(K)

Daytime Nighttime

(b) Terra MODIS

STD = 124

N = 45


STD = 103

N = 18

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Aqua

MO

DIS

LST

(K)

Daytime Nighttime

(c) Aqua MODIS





MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2


Competing Interests


Acknowledgments


References































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Bias = 090

STD = 141

N = 25

Bias = minus122

STD = 094

N = 27

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

VII

RS L

ST (K

)

Daytime Nighttime

(a) NPP VIIRS

STD = 149

N = 31


STD = 085

N = 23

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Terr

a MO

DIS

LST

(K)

Daytime Nighttime

(b) Terra MODIS

STD = 124

N = 45


STD = 103

N = 18

280 285 290 295 300 305 310 315 320275

Ground LST (K)

275

280

285

290

295

300

305

310

315

320

Aqua

MO

DIS

LST

(K)

Daytime Nighttime

(c) Aqua MODIS





MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2


Competing Interests


Acknowledgments


References































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










MOD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50LA

I

(a) MOD15A2

MYD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(b) MYD15A2

MCD15A2

Reference data

190 200 210 220 230 240180

DOY

10

15

20

25

30

35

40

45

50

LAI

(c) MCD15A2


Competing Interests


Acknowledgments


References































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation



































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









research article wireless sensor network of typical land...

Documents