Contents lists available at ScienceDirect

Remote Sensing of Environment

journal homepage: www.elsevier.com/locate/rse

Capturing rapid land surface dynamics with Collection V006 MODIS BRDF/NBAR/Albedo (MCD43) products

Zhuosen Wanga,c,⁎, Crystal B. Schaafb, Qingsong Sunb, Yanmin Shuaib, Miguel O. Románc

a Earth System Science Interdisciplinary Center, University of Maryland, College Park, MD, USAb School for the Environment, University of Massachusetts, Boston, MA, USAc Terrestrial Information Systems Laboratory, NASA Goddard Space Flight Center, Greenbelt, MD, USA

A R T I C L E I N F O

Keywords:BRDFNBARAlbedoCollection V006MODISClimateTemporal dynamicLand product validation

A B S T R A C T

MODIS BRDF/NBAR/Albedo products (MCD43) have been widely used for a variety of land surface studies since2000 (e.g., for climate and biosphere modeling, radiative forcing, phenology, and classification schemes). In thepast, limitations in computing and archive capacity had restricted the MCD43 retrievals to only once every eightdays. The latest reprocessing of the entire archive of Terra and Aqua MODIS BRDF/NBAR/Albedo products(Collection V006) provide retrievals each day and therefore represent a major improvement for phenological andsurface monitoring efforts. Compared to the previous retrieval which occurred every 8-days based on a 16-dayretrieval period for Collection V005 MODIS BRDF/NBAR/Albedo, this study demonstrates that the CollectionV006 MODIS BRDF/NBAR/Albedo product retrieved every day captures significantly more seasonal vegetationdynamics and rapid land surface changes. The Collection V006 product still utilizes the 16-days of directionalreflectances surrounding the day of interest (9th day) to retrieve the model of the surface anisotropy but assignsthe highest temporal weight to the day of interest (the center date of the retrieval period). Furthermore, theCollection V006 retrievals utilize the day of interest snow/snow-free status to capture snow albedo (an im-provement over the previous Collection V005 strategy which based the retrieval on the predominant snow/snow-free status of the majority of the observations over the 16-day retrieval period). All valid clear sky ob-servations from Terra and Aqua are now used for each retrieval, which improves the inversion quality. This isespecially true at high latitudes, where up to nine observations from each sensor can be observed per day. Thus,although the high quality full inversion Collection V006 albedo retrieval values are consistent with theCollection V005 results, the overall number of high quality results retrieved has increased. Furthermore, theaccuracy of the poorer quality magnitude inversions (which rely on a back-up algorithm) has improved due toutilization of the latest high quality full inversion retrievals as pixel-specific a priori knowledge. The ability ofthe latest Collection V006 MODIS BRDF/NBAR/Albedo products to capture land surface dynamics is evaluatedby using a globally-distributed record of spatially representative tower measurements. The RMSE of broadbandshortwave blue-sky albedo at these sites is< 0.0318 and the bias is within± 0.0076 for all quality results.

1. Introduction

Land surface albedo and reflectance anisotropy, recognized asEssential Climate Variables (ECV) by the Global Climate ObservingSystem (GCOS), quantify the radiative interaction between the atmo-sphere and the land surface (Schaaf et al., 2009). Albedo is one of thecrucial parameters used in global land surface climate and biospheremodels (Oleson et al., 2003). The MODIS BRDF/NBAR/Albedo products(Schaaf et al., 2002) have been produced since the launch of Terra in2000 (augmented by Aqua in 2002). The consistent high accuracy ofthese surface anisotropy (as described by the Bidirectional Reflectance

Distribution Function—BRDF), Nadir BRDF-Adjusted Reflectance(NBAR), and albedo values have been validated by comparison withglobally distributed ground albedo measurements and with periodicairborne BRDF measurements over a variety of land cover types duringboth the snow-covered and snow-free periods (Cescatti et al., 2012;Román et al., 2009, 2010; Wang et al., 2012, 2014).

The MODIS BRDF/NBAR/Albedo products have been embraced by anumber of global and regional modeling and monitoring communities(Kvalevåg et al., 2010; Lawrence and Chase, 2007). They provide initialconditions for the MODIS land surface temperature and cloud opticalproperties products (Platnick et al., 2003; Wan, 2014), and have been

https://doi.org/10.1016/j.rse.2018.02.001Received 16 September 2016; Received in revised form 13 January 2018; Accepted 6 February 2018

⁎ Corresponding author at: Earth System Science Interdisciplinary Center, University of Maryland, College Park, MD, USA.E-mail address: zhuosen.wang@nasa.gov (Z. Wang).

Remote Sensing of Environment 207 (2018) 50–64

Available online 09 February 20180034-4257/ © 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).

T

used as a priori and validation data for Advanced Along-Track ScanningRadiometer (AATSR) aerosol retrieval (Sayer et al., 2012), GLOBAlbedoproduction using European Satellites including MERIS (Muller et al.,2007), Landsat data gapfilling (Roy et al., 2008), Landsat NBAR andalbedo (Roy et al., 2016; Shuai et al., 2011), MSG/SEVIRI land surfacealbedo (Carrer et al., 2010), and the Global LAnd Surface Satellites(GLASS) albedo (Liu et al., 2013) products. MODIS albedo productshave been utilized to investigate the impact of land cover change onradiative forcing (Myhre et al., 2005; O'Halloran et al., 2012). While theMODIS albedo measures continue to be used in surface energy budgetcalculations, the BRDF and nadir view angle corrected reflectance(NBAR), which removes the variability due to angular effects, havebeen found to improve the accuracy of change detection in intra- andinter-annual temporal dynamics and therefore have been increasinglyused to detect natural and human disturbance, to monitor vegetationcover/phenology (Zhang et al., 2003), and to provide measures of thesurface variability (Chopping et al., 2011; He et al., 2012).

The BRDF describes the reflectance anisotropy of the surface andcan be retrieved from a sufficiently well distributed series of multi-angular satellites observations of a locality. An appropriate BRDF modelis fit to these available observations, and the land surface albedo canthen be obtained from the integration of the BRDF over the hemisphericview and illumination angles. The MODIS BRDF/NBAR/Albedo pro-ducts rely on the semi-empirical linear RossThick-LiSparse Reciprocal(RTLSR) BRDF model which utilizes the weighted sum of isotropic,volumetric and geometric parameters (Roujean et al., 1992; Schaafet al., 2002) to describe the reflectance anisotropy of each pixel. TheRTLSR has been utilized to describe the reflectance anisotropy of a largevariety of global land covers (Schaaf et al., 2002).

Clear sky, multi-angle, high quality atmospherically corrected sur-face reflectance data over a 16 days period from both Terra(MOD09GA) and Aqua (MYD09GA) (Roy et al., 2006) are utilized toobtain sufficient well distributed angularly observations to fit theRTLSR BRDF model for the retrieval of MODIS BRDF/NBAR/Albedoproducts (MCD43). Each observation is assigned a weight during theretrieval. MCD43 is the MODIS product number assigned to MODISBRDF/NBAR/Albedo products. The number of high quality observa-tions available, the Root Mean Square Error (RMSE) of the fit and theWeight of Determination (WOD) which indicates whether the ob-servations over the viewing and illumination geometry are sufficientlywell distributed to capture the surface anisotropy (Schaaf et al., 2002),are all used to generate a high quality full inversion retrieval. A backupalgorithm (or magnitude inversion), using available observations withan a priori BRDF database, is applied if a high quality full inversionBRDF cannot be retrieved due to poor quality input or insufficientsampling of the viewing hemisphere. While this magnitude inversionoften provides reliable results, it is always flagged as a poor qualityretrieval.

The surface measures of Black-Sky Albedo (BSA) (Directional-Hemispherical Reflectance -DHR) at local solar noon, White-Sky Albedo(WSA) (Bi-Hemispherical Reflectance (BHR) (MCD43A3), and NBAR(MCD43A4) at local solar noon are generated from the BRDF para-meters (MCD43A1). These terms are defined in Schaepman-Strub et al.(2006). Spectral NBAR (MCD43A4) values are derived with the geo-metry of a nadir view zenith angle and local solar noon. Users can alsoselect the solar angles appropriate for their application and derive thosequantities from the BRDF parameters themselves. Actual (blue-sky)albedo for a particular instant in time can be obtained by combining theclear sky black and white-sky albedos with an instantaneous measure ofsolar zenith angle and aerosol optical depth. The effects of multiplescattering and anisotropic diffuse illumination should be consideredmore carefully for fully snow-covered areas due to the high reflectanceof snow and large solar zenith angles of high latitudes, therefore, a fullexpression blue-sky albedo that solves the multi scattering effect shouldbe used for snow-covered areas (Román et al., 2010). For snow-freesurfaces, a simple form of blue-sky albedo can be calculated with an

assumption of an isotropic diffuse radiation (Lewis and Barnsley, 1994).The visible (0.3–0.7 μm), near-infrared (0.7–5.0 μm) and shortwave

(0.3–5.0 μm) broadband values of anisotropy and albedo, which arecommonly used in surface energy budget models, are generated fromspectral values via narrow-to-broadband conversion coefficients forsnow-free and snow-covered surfaces (Schaaf et al., 2002). In CollectionV006, due to the large number of non-functional or noisy detectors inMODIS Aqua Band 6 (1.628–1.652 μm), separate narrow-to-broadbandconversion coefficients without Band 6 are applied for those conditionswere Band 6 has failed to be retrieved but all other bands have validretrievals. Previous research indicates that the accuracy of the MODISshortwave broadband albedo meets the requirements (< 5%) for bothsnow-free and snow-covered surfaces (Román et al., 2009, 2010; Wanget al., 2012, 2014).

The Collection V005 MCD43 BRDF/NBAR/Albedo products (Schaafet al., 2002) were only retrieved every 8 days based on 16 days of ob-servations due to previous limitations in NASA's science production andarchiving capacity. However, an albedo product retrieved on a dailybasis day is required to improve the energy budget computation in landsurface models and to capture surface dynamics (i.e. vegetation phe-nology and land cover disturbance) with high accuracy. Therefore, theMODIS BRDF/NBAR/Albedo products are retrieved each day in theCollection V006 reprocessing, representing a major milestone in thissuite of global land products. This paper demonstrates how the avail-ability of Collection V006 products significantly enhances the detectionof land surface dynamic characteristics, especially during rapidlychanging situations (Wang et al., 2012, 2014). In particular, this studydescribes the retrieval strategy employed for the Collection V006MODIS BRDF/NBAR/Albedo products and evaluates the ability of thenew BRDF/NBAR/Albedo product retrieved on a daily basis using a 16-day retrieval window to capture land surface dynamics.

2. Methodology

2.1. Improvements in input MODIS Collection V006 values

The Collection V006 MODIS Level-1B reprocessing is implementedvia numerous calibration Look Up Tables (LUTs) updates (Lyapustinet al., 2014; Toller et al., 2013) which improve the quality and con-sistency of the downstream Level 2 (L2) and Level 3 (L3) MODIS pro-ducts for the entire Terra/Aqua mission period. The main Level-1Bimprovements include solar diffuser (SD) degradation correction, re-sponse vs. scan angle (RVS) LUTs updates, and polarization correction.The improvements of MODIS Collection V006 L2G-lite surface re-flectance (MO/YD09GA) products include improved aerosol retrievaland correction algorithms; new aerosol retrieval LUTs; and refined in-ternal snow, cloud, and cloud shadow detection algorithms (https://lpdaac.usgs.gov/dataset_discovery/modis/modis_products_table/mod09ga_v006).

The MODIS Collection V006 L2 cloud mask product is improved byincluding Normalized Difference Vegetation Index (NDVI) to betterdefine cloud detection thresholds and enhance the knowledge of ex-pected surface reflectances (Baum et al., 2012).

Using these improved MODIS surface reflectance and cloud maskproducts, the Collection V006 implementation of the MODIS BRDF,NBAR and albedo products represents a major improvement for en-vironmental modeling and monitoring. Key upgrades include: (1) re-trieved every day using a 16-day retrieval period, (2) expanding thequality and uncertainty values, and (3) implementing an improvedbackup database for poorer quality retrievals. Furthermore, theCollection V006 reprocessing algorithm utilizes all the valid clear skyobservations for the retrieval (whereas earlier versions were limited toonly using up to four observations to reduce storage processing volume)(Wolfe et al., 1998). The increased quality and enhanced temporal re-solution afforded by the Collection V006 algorithm has the greatestimpact at high latitudes, where persistent cloud cover, difficulties in

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

51

discriminating cloud versus snow, large solar zenith angles, and rapidchanges in snow and freeze conditions all contribute to difficulties inretrieving frequent high quality surface BRDFs (Wang et al., 2014).Collection V006 obtains more retrievals than Collection V005 at higherlatitudes (Fig. 1), while fewer retrievals can be found at some low la-titudes (e.g. Amazon, Congo basin) due to the reduced number of clear-sky valid surface reflectances. More observations in the Collection V006surface reflectance products (MO/YD09GA) are labeled as high aerosolor cloud contaminated than in Collection V005 at mid and low lati-tudes. The total number of observations can be around 18 per day athigh latitudes by combining both the Terra and Aqua MODIS ob-servations (Fig. 3) — which increases the algorithm's ability to retrievehigh quality full inversions. The retrieval quality is determined by thesolar angles, the temporal distribution of clear sky valid observations,and the RMSE and WOD thresholds. The temporal variation in thetiming of the spring melt, fall freeze, and snow cover over land, as well

as the spatial distribution and character of Arctic water bodies (e.g.,tundra ponds and lakes) all contribute to significant changes in radia-tive forcing and albedo feedback (Chapin et al., 2005). As such, theproduction of albedo values retrieved every day using a 16-day retrievalperiod offers a major improvement in monitoring high latitudes. Fur-thermore, estimates of albedo retrieved every day using a 16-day re-trieval period can improve the ability to quantify how continuedchanges in the timing of melt due to climate change will affect theenergy feedback in the future.

2.1.1. Temporal processing strategyBoth Collection V005 and V006 BRDF/NBAR/Albedo products uti-

lize 16 days' worth of multi-angular clear sky valid observations to re-trieve an appropriate reflectance anisotropy model (BRDF). CollectionV005 products were retrieved every 8 days using 16 days of observa-tions with uniform weighting over the retrieval period. The CollectionV006 products are retrieved on a daily basis. Each observation is as-signed a weight which is a function of the quality, the temporalproximity to the day of interest, and the observation coverage. Thetemporal weight, based on a Laplace distribution, emphasizes the day ofinterest and thus the retrieval also represents the NBAR and albedovalues for the day of interest (Fig. 2). A Laplace distribution assignshigher weights to the data adjacent to the day of interest as compared tolinear and Gaussian distributions. The day of interest used in CollectionV006 is the center date of the 16 day retrieval period (9th day). Whilethe multiday observations provide information about the general sur-face anisotropy, emphasis on the observations from a single day refinesthe estimates to more closely represent the day of interest. Therefore,the daily retrieval greatly increases the ability of the MODIS BRDF/NBAR/Albedo products to provide accurate representations of thetemporal dynamics occurring due to vegetation phenology, ephemeralsnow or snow melt, and rapidly occurring disturbances. The retrievalquality can still be affected due to the missing observations within the16-day period especially during the seasonal transition periods (e.g.green-up grow or senescence). The retrieval will only use observationsfrom the surrounding days if no clear-sky observations are available onthe day of interest due to cloud and aerosol contamination. In this case,the Quality Assurance (QA) of MODIS BRDF/NBAR/Albedo products islowered. Other short-term changes (e.g. ephemeral snow) could also bemissed if no clear sky observations are available during the entireperiod. The valid observation layer in MCD43A2 provides more detailsabout the days with valid clear-sky observations for the retrievals.

2.1.2. Snow/snow-free statusPreviously the MCD43 snow albedo retrieval was determined by

whether or not the majority of observations were snow-covered withineach 16-day period and if the majority was snow-covered then onlysnow observations were used for the retrieval. Therefore, short-termephemeral snow (where snow falls that melted within one or two days)were ignored following this snow/snow-free retrieval strategy, since the

Fig. 1. Collection V006 (a) and V005 (b) global true color NBAR composite for Day ofYear (DOY) 201 in 2003 shown in the MODIS equal area Sinusoidal projection. The NBARranges of the “natural” color composite are red (MODIS red band: band 1): 0–0.3; green(MODIS green band: band 4): 0–0.3; blue (MODIS blue band: band 3): 0–0.3. (For in-terpretation of the references to color in this figure legend, the reader is referred to theweb version of this article.)

Fig. 2. Temporal retrieval of Collection V006 MCD43BRDF/albedo product (red star stands for the day of interestwithin the 16-days period, DOY 009 represents the day ofinterest for period DOY 001 to 016). The upper graph istemporal weights based on Laplace distribution. (For inter-pretation of the references to color in this figure legend, thereader is referred to the web version of this article.)

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

52

number snow-covered days within a 16-day period would be less thansnow-free days. This produced less frequent snow albedo results andthus the overall product tended towards provision of “snow-free” al-bedos. In the Collection V006 BRDF/Albedo/NBAR product im-plementation, the snow or snow-free retrieval is determined by thesnow status of the day of interest, therefore each snow-covered day canbe captured and only snow-covered observations from the surroundingperiod are used for the BRDF/albedo retrieval of that target day. Inaddition to utilizing the snow/ice flag (MOD35) stored in surface re-flectance (MO/YD09GA) product, the Normalized Difference SnowIndex (NDSI) snow cover algorithm (Hall et al., 2002) is also utilized toimprove the quality of the retrieval. This is especially effective incapturing the snow melt period (Wang et al., 2012). MODIS RTLSRBRDF model has been evaluated over snow surface with high accuracy.The RMSEs (Eq. 1) of the MCD43 snow albedo retrievals are< 0.06over Greenland, tundra, forest, cropland, and grassland using in-situmeasurements (Wang et al., 2012, 2014). The difference between Col-lection V006 BRDF model derived NBAR values and Unmanned AerialSystem nadir reflectance measurements has been found to be<0.02(Burkhart et al., 2017) at a location in Greenland.

∑==

−RMSE 1n

(y y )j 1

n

j j2

(1)

where n is the sample size, yj is MODIS albedo, and yj is ground mea-sured albedo.

2.1.3. Backup algorithm based on a priori resultsA backup algorithm (known as a magnitude inversion) is employed

if the anisotropic reflectance model cannot be confidently retrievedwith a full inversion due to either a limited number of clear sky ob-servations (e.g. due to aerosol or cloud contamination) or to an in-sufficient sampling of the viewing hemisphere. This method (Strugnelland Lucht, 2001) relies on an a priori database for an estimate of the

underlying BRDF, and couples this estimate with all available clear-skycloud free observations to constrain that initial model. Collection V005relied on static BRDF-based land cover specific seasonal implementa-tions for the a priori BRDF models (Strugnell and Lucht, 2001). In theCollection V006 algorithm, the a priori database is updated per dayutilizing the most recent high quality snow-free and snow-covered fullinversions for each pixel. Therefore, the a priori BRDF database inCollection V006 is closer to the real BRDF on the day of interest ascompared to the static seasonal land cover BRDFs used in CollectionV005. Although the magnitude inversion method often performs quitewell under normal conditions and over homogenous surfaces (Cescattiet al., 2012; Román et al., 2009), users of the MCD43 V006 product areadvised to use these magnitude (backup) results with caution, as theyare considered lower quality results and are flagged accordingly in theMODIS BRDF/Albedo quality product (MCD43A2) as poor quality.

2.1.4. Quality assurance and uncertaintyThe Collection V006 MODIS BRDF/albedo algorithm also updates

the Quality Assurance (QA) fields and expands the measures of un-certainty provided for modeling applications. The BRDF/NBAR/Albedoretrieval status (full inversion, magnitude inversion, and filled value)can be found from the QA fields. Byte-based quality flags in separatelayers have been introduced into MCD43A2 in Collection V006 to re-place the bit-based quality flags of earlier versions (this in order toencourage users to use this crucial quality information by avoidingadded steps in unpacking bit-level quality information). Although theexplicit quality of the retrieval is still stored in MCD43A2, the rudi-mentary mandatory QA fields for each band are also stored as part ofeach Collection V006 MCD43 product (MCD43A1 (BRDF parameters),MCD43A3 (Albedo quantities), and MCD43A4 (NBAR)) to provide atleast the most basic quality information by describing the status of highquality full inversions, poorer quality magnitude inversions, or filledvalue. The shortwave broadband mandatory QA indicates the retrievalstatus of all the MODIS land narrow bands (Band 1–Band 7). The

Fig. 3. The number of observations from Terra (left) and both Terra and Aqua (middle) with the resulting shortwave WSA (right) for tile H16V01 (north of Greenland, latitude 80°–70°,top figures) and H11V04 (north of USA, latitude 50°–40°, bottom figures) on DOY 201, 2010 shown in the MODIS equal area Sinusoidal projection.

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

53

shortwave broadband mandatory QA is “0” if all the narrowband re-trievals are full inversions, “1” if all bands are not full inversion, and fillvalue if one or more of the bands could not to be retrieved due tolimited number of clear sky observations. The status of the input surfacereflectance (ranging from high quality clear sky to cloud/high aerosolcontaminated) for each day is also provided in Collection V006 to betterunderstand the retrieval quality. The retrieval quality is also lowered ifthe day of interest has no valid observation or the solar zenith angle islarger than 70°. Updated RMSE and WOD thresholds to constrain thefull inversions have been implemented, and are based on the in-vestigation of inversions at typical biomes located at different latitudes(Shuai et al., 2008). The updated thresholds for high quality full in-versions are 0.08, 1.65, and 2.50 for RMSE, WOD-NBAR at 45° solarzenith angle, and WOD-WSA respectively. Surface albedo is a keyparameter for many surface energy budget model simulations and theuncertainty of model input parameters is required. In Collection V006,the albedo uncertainty is provided via the WOD values in MCD43A2 QAproduct for data assimilation and modeling efforts. The details of QAfields can be found from https://ladsweb.modaps.eosdis.nasa.gov/filespec/collection6/.

2.1.5. Increased number of input observationsIn previous implementation of MCD43, a maximum of four ob-

servations per day from each sensor was used for the BRDF/Albedoretrieval for each pixel due to processing constraints. This limit hasbeen relaxed in Collection V006 so that all the available clear sky ob-servations with solar zenith angles of< 80° are considered for the per-pixel retrieval. This greatly improves the retrieval quality, especially inhigh latitude regions. Fig. 3 shows the number of available observationsfrom Terra-MODIS and both Terra and Aqua MODIS for tiles H16V01and H11V04 (https://modis-land.gsfc.nasa.gov/MODLAND_grid.html)for 2010, Day of Year (DOY) 201. The tile H16V01 is one of the mostnortherly MODIS land tiles and therefore has nearly the greatestnumber of overpasses. There are greater than four observations overmost of the tile areas and there can be up to nine observations. Giventhe increased cloud contamination prevalent at higher latitudes, thisgreater number of observations greatly increases the possibility of ob-taining high quality full inversion results. The total number of ob-servations of both Terra and Aqua MODIS for tile H11V04 (latitude50°–40°) is up to 4 on DOY 201.

2.2. Albedo evaluation

Evaluation of a satellite product via comparison with ground basedmeasurements is a complex process involving an understanding of thesatellite product gridded footprint and the appropriateness of the sur-face measure to represent that spatial scale (Campagnolo et al., 2016).Thus, while a point-to-pixel direct comparison method of validation byusing ground measured albedos may be reasonable over homogenous orrepresentative surfaces, this evaluation can lead to large discrepanciesif the surface is heterogeneous or unrepresentative of the satellite ob-servation scale. A spatial representativeness analysis is therefore ne-cessary for the validation of satellite albedo, especially when comparingcoarse spatial resolution datasets against fine scale ground measure-ments (Román et al., 2009, 2010; Wang et al., 2012, 2014). In a spatialrepresentativeness analysis, the ability of tower measures to representthe surface within the 2 km surrounding area is generally established. Inthis study, ground measured albedo from twenty-one homogeneous (orhomogeneously heterogeneous) tower sites, as based on the spatialrepresentativeness analysis, were selected to validate the ability of theMODIS albedo product to capture land surface dynamics (Table 1).Here “homogenously heterogenous” indicates that the surface may beheterogenous at finer scales but is homogenous at the satellite spatialresolution. The measurements were collected by the international WMOBaseline Surface Radiation Network (BSRN) (including the NOAA/SURFRAD (Surface Radiation Budget Network)), as well as the ArcticTa

ble1

Cha

racteristics

ofgrou

ndsitesused

inthean

alysis.

Site

name

Lat/lon

Tower

height

(m)

Tower

footprint(m

)MODIS

tile

Years

Land

type

Flux

netw

orks

Referen

ces

BC-Cam

pbellRiver

1949

Dou

glas-fir(C

A-Ca1

)49

.867

/−12

5.33

445

568.2

H09

V04

2004

,200

5Ev

ergreenne

edleleaf

forest

CCP

Morge

nstern

etal.(20

04)

SK-199

8Fire

(CA-SF3

)54

.092

/−10

6.00

530

378.8

H11

V03

2003

,200

4Ev

ergreenne

edleleaf

forest

CCP

Amiroet

al.(20

06)

Wetzstein

(DE-Wet)

50.454

/11.45

829

.537

2.5

H18

V03

2004

,200

5Ev

ergreenne

edleleaf

forest

Carbo

Europe

IPAntho

niet

al.(20

04)

Griffin-Abe

rfeldy

-Sco

tlan

d(U

K-G

ri)

56.607

/−3.79

815

.519

5.7

H17

V03

2000

,200

1Ev

ergreenne

edleleaf

forest

Carbo

Europe

IPMed

lynet

al.(20

05)

Tumba

rumba

(AU-Tum

)−

35.656

/148

.152

7088

3.9

H30

V12

2005

Evergreenbroa

dleafforest

OzF

lux

Finn

igan

andFinn

igan

andLe

uning(200

0)Hesse

Forest-Sarrebo

urg(FR-H

es)

48.674

/7.066

1822

7.3

H18

V04

2004

,200

5Deciduo

usbroa

dleafforest

Carbo

Europe

IPGranier

etal.(20

00)

Walke

rBran

ch(U

S-WBW

)35

.959

/−84

.287

3847

9.9

H11

V05

2004

,200

5Deciduo

usbroa

dleafforest

AmeriFlux

Vermaet

al.(19

86)

SK-198

9Fire

(CA-SF2

)54

.254

/−10

5.87

830

378.8

H11

V03

2003

,200

4Mixed

forests

CCP

Amiroet

al.(20

06)

Bily

Kriz-

BeskidyMou

ntains

(CZ-BK

1)49

.502

/18.53

711

214

14.3

H19

V04

2004

,200

5Mixed

forests

Carbo

Europe

IPMarek

etal.(20

11)

Hainich

(DE-Hai)

51.079

/10.45

242

530.4

H18

V03

2004

,200

5Mixed

forests

Carbo

Europe

IPKno

hlet

al.(20

03)

Harva

rdFo

rest

(US-Ha1

)42

.538

/−72

.172

3037

8.8

H12

V04

2007

Mixed

forest

Ameriflux

Bakw

inet

al.(20

04)

LasMajad

asde

lTietar

(ES-LM

a)39

.942

/−5.77

315

189.4

H17

V05

2008

,200

9Closedshrublan

dsCarbo

Europe

IPGöc

kede

etal.(20

07)

Skuk

uza(ZA-Kru)

−25

.020

/31.49

716

202.0

H20

V11

2002

,200

3Sa

vann

asCarbo

Africa

Scho

leset

al.(20

01)

ElSa

ler-Su

eca(ES-ES

2)39

.276

/−0.31

51.6

20.2

H17

V05

2008

,200

9Croplan

dsCarbo

Europe

IPMoo

rset

al.(20

10)

Dripsey

(IE-Dri)

51.987

/−8.75

210

126.3

H17

V03

2004

,200

5Grassland

sCarbo

Europe

IPPe

ichl

etal.(20

11)

Cottonw

ood(U

S-Ctn)

43.950

/−10

1.84

75

63.1

H10

V04

2007

,200

8Grassland

sAmeriFlux

Gilm

anov

etal.(20

10)

Fort

Peck

48.308

/−10

5.10

210

126.3

H11

V04

2004

,201

1Grassland

sSU

RFR

AD

–Bo

ulde

r40

.050

/−10

5.01

030

037

88.3

H09

V04

2007

Grassland

sBS

RN

–Ba

rrow

71.323

/−15

6.60

94

50.5

H12

V01

2007

Tund

raBS

RN

–Im

naviat

68.613

/−14

9.31

22.5

31.6

H12

V02

2009

Tund

raAON

–Desertrock

36.620

/−11

6.02

010

126.3

H08

V05

2004

Desert

SURFR

AD

–

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

54

Observatory Network (AON, http://aon.iab.uaf.edu/AON_IC_Site_Location.html), and the FLUXNET community (La Thuile datasets) in-cluding Ameriflux, the Integrated Project “Assessment of the EuropeanTerrestrial Carbon Balance” (CarboEuropeIP, http://www.carboeurope.org), the Australian Ozflux, the Canadian Carbon Program (CCP) andthe African CarboAfrica (http://www.carboafrica.net) locations. Theseglobally distributed sites cover most land surface types, includingevergreen needleleaf forest, evergreen broadleaf forest, mixed forest,deciduous broadleaf forest, croplands, closed shrublands, open shrub-lands, savannas, grasslands, tundra, and desert regions (Table 1, Fig. 4).

The tower-based shortwave albedos were primarily obtained fromKipp & Zonen or Eppley pyranometers models mounted at heightsranging from 1.6m to 300m. The footprints of the ground measure-ments therefore vary from 20.204m to 3788.25m, with most field sitesproviding measures appropriate for regions of< 500m (Table 1). Theground albedos are computed as the ratio of upward shortwave radia-tion to the downward shortwave radiation during the noon hour so as tobe comparable with the MODIS albedo product at local solar noon. Theoverall expected accuracy of the ground measurements is within 7%under clear sky and 4% under overcast conditions (Cescatti et al., 2012;

Pirazzini, 2004). Several sites also monitor the direct normal and dif-fuse shortwave fluxes, as measured by Normal Incidence Pyranometersand shaded pyranometers, which can be used to calculate the snow-freeblue-sky albedo. For those sites without ground measurements, thediffuse irradiation fraction is simulated by the 6S radiative transfermodel (Vermote et al., 1997) with the Aerosol Optical Depth (AOD)from the MODIS Aerosol product (MOD08) (Remer et al., 2005). Aclimatological value for AOD of 0.2 at 550 nm is used if the MOD08aerosol is not available (Wang et al., 2010; Yu et al., 2006). The blue-sky albedo over all snow-covered surfaces can generated with AODvalues and a MODTRAN simulation considering multi-scattering effects(Román et al., 2010). Both full and magnitude inversions are used forthis evaluation.

2.3. Evaluation of the ability of the Collection V006 albedo and NBARproducts to capture land surface dynamics

2.3.1. PhenologyThe enhancement of MCD43's temporal retrieval resolution, from 8-

day to daily, improves the ability of researchers to capture land surface

Fig. 4. The spatial distribution of ground sites (black star) listed in Table 1. The background map is MODIS International Geosphere-Biosphere Programme (IGBP) land cover map(MCD12Q1).

Table 2The statistics of retrieval quality and broadband shortwave White Sky Albedo (WSA) differences within latitude zones are investigated with MODIS albedo data from combining DOY 009and DOY 201, 2010.

Latitude Percentage of fullinversion

Percentage of magnitudeinversion

Percentage of filled value Full inversion and magnitude inversionWSA difference

Full inversion WSA difference

V005 V006 V005 V006 V005 V006 Mean Standard deviation Mean Standard deviation

90°–70° 11.8% 28.0% 33.1% 28.1% 55.1% 43.9% 0.023 0.048 0.005 0.01770°–50° 25.4% 30.6% 28.5% 33.7% 46.1% 35.7% 0.007 0.036 0.000 0.00950°–30° 46.3% 38.5% 31.2% 43.1% 22.5% 18.4% 0.005 0.022 0.003 0.00930°–10° 57.2% 46.3% 15.4% 28.7% 27.4% 25.0% 0.001 0.011 0.002 0.00610° to −10° 23.5% 17.5% 31.5% 43.2% 45.0% 39.3% −0.002 0.014 0.001 0.007−10° to −30° 54.2% 45.3% 22.1% 33.9% 23.7% 20.8% −0.001 0.012 0.001 0.006−30° to −50° 45.2% 37.4% 20.3% 29.4% 34.5% 33.2% 0.001 0.012 0.002 0.005−50° to −70° 7.3% 5.5% 38.0% 42.8% 54.6% 51.7% 0.005 0.023 0.005 0.007−70° to −90° 16.3% 23.7% 34.0% 34.1% 49.7% 42.2% 0.024 0.045 0.016 0.021

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

55

dynamics. Vegetation indices derived from the NBAR product removeeven more of the angular effect (than already reduced by the indicesthemselves) (Gao et al., 2002; Schaaf et al., 2002). The Enhanced Ve-getation Index (EVI) is used in this study as it is also used for the MODISphenology product (Zhang et al., 2003). The improvements offered bythe algorithm retrieved each day (albeit utilizing a 16-day retrievalperiod) are illustrated by analyzing the vegetation phenology derivedfrom Collection V006 MODIS NBAR product at Harvard Forest.

2.3.2. DisturbanceThe ability of the Collection V006 MCD43 BRDF/NBAR/Albedo

product to detect surface disturbance is evaluated at a location in theNorth Slope of Alaska during the Anaktuvuk River Fire (Latitude 69.2°,Longitude −150.7°). This fire, started by a lightning strike in July2007, is the largest fire on record (since 1950) for the tundra biome. Itburned 1039 km2 of Alaska's arctic slope. This arctic fire released ap-proximately 2.1 Tg of carbon into the atmosphere with the maximum

Fig. 5. The albedo retrieval quality over tile H16V01 (top) and H11V04 (bottom) on DOY 201, 2010 for Collection V005 (left) and V006 (right). Red is high quality full inversion, green ispoor quality magnitude inversion, and white represents filled value. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of thisarticle.)

Table 3The percentage of retrievals where no clear sky observations were available on the day of interest for DOY009, 110, 201, and 300, 2010.

Latitude DOY009 DOY110 DOY201 DOY300

Snow free Snow Snow free Snow Snow free Snow Snow free Snow

90°–70° N/A N/A 1.5% 58.3% 33.9% 12.7% N/A N/A70°–50° 9.8% 66.7% 24.0% 31.9% 56.0% 0.6% 55.5% 23.8%50°–30° 36.4% 19.3% 58.8% 1.8% 43.7% 0.0% 48.4% 1.4%30°–10° 25.0% 0.1% 48.0% 0.1% 59.8% 0.0% 34.0% 0.0%10° to −10° 61.0% 0.0% 61.2% 0.0% 68.6% 0.0% 77.3% 0.0%−10° to −30° 65.2% 0.0% 32.4% 0.0% 27.2% 0.0% 43.2% 0.0%−30° to −50° 22.3% 0.0% 39.1% 0.0% 43.2% 1.3% 56.6% 0.1%−50° to −70° 11.0% 27.8% 36.0% 38.6% 32.4% 0.4% 16.5% 27.6%−70° to −90° 0.1% 10.0% N/A N/A N/A N/A 0.0% 39.5%

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

56

age of soil carbon lost estimated at 50 years (Mack et al., 2011).

2.3.3. SnowmeltThe snow/snow-free retrieval of the Collection V006 MODIS BRDF/

NBAR/Albedo products is determined by the snow status of day of in-terest. The snow can melt very rapidly from a completely snow-coveredsurface to one that is snow-free. The ability of the Collection V006MODIS albedo product to monitor rapid snowmelt is demonstrated overthe area around Barrow, Alaska.

3. Results and discussion

3.1. Intercomparison of Collection V005 and V006 albedo

The day of interest for the Collection V006 BRDF/albedo product isthe middle day (9th day) of the 16-day retrieval period (e.g. the DOY009 BRDF is retrieved by using data from DOY 001 to 016) whereas theday of interest of the Collection V005 BRDF/albedo product was thefirst day of the 16-day period (e.g. DOY 001 BRDF is retrieved usingdata from DOY 001 to 016). Therefore, here the DOY of Collection V005products have been adjusted by 8 days so that they are using the sameinput data as the V006 product during this comparison. More highquality albedos are acquired at higher latitudes in Collection V006because all of the available clear sky cloud free observations are nowutilized for the retrieval, rather than just a maximum of four observa-tions per day from each sensor (Terra and Aqua), as was the case forCollection V005 (imposed by a processing constraint). Note however,that the percentage of full inversions for Collection V006 is somewhatlower than Collection V005 at mid and low latitudes (Table 2). OneBand 5 and fifteen of the twenty Aqua Band 6 detectors are currentlynon-functional or noisy. In Collection V005, the values of these non-functional or noisy detectors were spatially interpolated from the valuesof the surrounding detectors. In Collection V006, the non-functional ornoisy detectors are kept as fill values, and therefore the number of clearsky observations available for BRDF retrieval has decreased as com-pared to Collection V005. This is particularly true at mid and low la-titudes as the detection of clouds and aerosols has been improved in theCollection V006 surface reflectance product. The percentage of fill va-lues in the Collection V006 BRDF/Albedo/NBAR product is less than inthe Collection V005. Fig. 5 shows the albedo retrieval quality over theMODIS tiles H16V01 and H11V04 on DOY 201, 2010. The percentageof high quality full inversion retrievals increases from 13.2% in Col-lection V005 to 34.1% in Collection V006 for tile H16V01, while itdecreases from 72.6% in Collection V005 to 63.7% for tile H11V04.Fig. 5 also indicates the quality flag distribution of all full inversion(high quality) and magnitude inversion (poor quality) values with the

strips at the south and east portions of the Collection V006 tile H11V04indicating the effect of non-functional or noisy detectors.

Table 3 shows the percentage of retrievals with no clear sky ob-servations available on the day of interest on DOY009, 110, 201, and300, 2010. The values are high at low latitudes (10° to −10°) due tohigh frequency of clouds and aerosol contamination and low over therest of the southern hemisphere. Relatively high percentages of wintersnow albedo are retrieved for dates with no clear sky observationsavailable on the day of interest in high northern hemisphere latitudes.The Collection V006 retrieval method uses 16-days' worth of observa-tions in the retrieval period and therefore can estimate an albedo valueseven if no clear sky observations are available on the day of interest dueto atmospheric contamination. However, the retrieval QA value islowered for this situation. The ValidObs layer of the accompanying QAenables users to identify whether clear sky observations were availableon the day of interest. Thus the BRDF/albedo values retrieved in si-tuations where there were no clear sky observations on the day of in-terest can be easily filtered out using the ValidObs layer provided in thequality flag (MCD43A2) if the end-users do not want to use them.

The mean values of the shortwave broadband WSA difference be-tween Collection V005 and V006 at each 20° latitude range is within0.024 for all quality retrievals (both full inversion and magnitude in-version) and within 0.016 for high quality full inversion from com-bining DOY 009 and DOY 201, 2010 (Table 2). The standard deviationof the shortwave WSA difference is< 0.048 for all quality and 0.021 forfull inversion. The standard deviation of the difference increases fromlow to high latitude where snow retrievals are performed. The meanvalues of the broadband shortwave WSA difference between CollectionV005 and V006 is within 0.025 for all quality and 0.017 for high qualityover all MODIS IGBP land types (MCD12Q1) (Friedl et al., 2010)(Table 4). The standard deviation of the difference is< 0.046 for allquality and 0.020 for high quality. The snow and ice regions have thelargest mean differences and standard deviations of the difference in-dicating an improved ability to capture snow quantities.

3.2. Collection V006 MODIS albedo product evaluation

The tower site data for both the Collection V006 and the CollectionV005 MODIS blue-sky full and magnitude inversion albedos are pro-vided in Fig. 6. Both the Collection V006 and Collection V005 albedosmatch well with the ground measurements over these spatially re-presentative sites from a variety of land surface types. A large range inalbedo occurs over the Fort Peck grassland site (Fig. 6(a)). The MODISderived albedo values quite closely follow the timing of the springthaw. The albedos are around 0.8 during the snow-covered period anddrop to about 0.2 after snowmelt. The ephemeral snow fall usually

Table 4The statistics of the broadband shortwave WSA difference between Collection V006 and V005 at each land type (MODIS IGBP) by using MODIS albedo data from combining DOY009 andDOY 201, 2010. High quality retrievals indicate full inversion retrievals.

Retrieval quality LC1 LC2 LC3 LC4 LC5 LC6 LC7 LC8 LC9

All retrievals Mean 0.001 0.006 −0.003 0.008 −0.001 0.007 −0.001 0.001 0.000Standard deviation 0.027 0.036 0.014 0.043 0.016 0.037 0.013 0.016 0.019

High quality retrievals Mean 0.004 0.001 0.003 −0.001 0.003 0.001 0.000 0.001 0.001Standard deviation 0.019 0.008 0.006 0.007 0.006 0.008 0.006 0.007 0.007

Retrieval quality LC10 LC11 LC12 LC13 LC14 LC15 LC16 LC17

All retrievals Mean −0.001 0.005 0.002 0.006 0.006 0.005 0.025 0.003Standard deviation 0.016 0.023 0.026 0.020 0.020 0.026 0.046 0.014

High quality retrievals Mean 0.001 0.002 0.001 0.003 0.004 0.001 0.017 0.002Standard deviation 0.006 0.008 0.008 0.007 0.008 0.007 0.020 0.007

LC1: water, LC2: evergreen needleleaf forest; LC3: evergreen broadleaf forest, LC4: deciduous needleleaf forest, LC5: deciduous broadleaf forest, LC6: mixed forest, LC7: closedshrublands, LC8: open shrublands, LC9: woody savannas, LC10: savannas, LC11: grasslands, LC12: permanent wetlands, LC13: croplands, LC14: urban and built-up, LC15: cropland/natural vegetation mosaic, LC16: snow and ice, LC17: barren or sparsely vegetated.

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

57

melts very rapidly, therefore the temporal range with snow-coveredconditions is very short. These short-term snow-covered periods arealso more likely to be cloudy which makes them harder to capture. Forinstance, no clear-sky valid observations were available during theephemeral snow event around DOY 90 at Fort Peck. The tower albedosat the Desert Rock site remain around 0.22 throughout the year, whilethe MODIS albedo values show a little variation during the winterperiods and the difference between MODIS albedo and ground mea-surements is< 0.05. However, it should be noted that this period wasparticularly cloudy and the albedo retrievals are dominated by poorerquality magnitude inversion, demonstrating that the accuracy ofMODIS albedo products is impacted by the detection of cloud and cloudshadow in the surface reflectance product. The albedos from the Har-vard Forest mixed forest site show strong seasonal dynamics, with thealbedo decreasing with canopy snowmelt, increasing during green up,

and then decreasing over the dormant period (Fig. 6(d)). Two poorquality albedo values were retrieved in Collection V005 while no al-bedos were retrieved in Collection V006 during the DOY 041–061period at the Harvard Forest site. The cloud and aerosol status used forthe MCD43 BRDF/NBAR/Albedo products are taken from the MODISsurface reflectance products (MOD09GA and MYD09GA) which havebeen significantly improved in Collection V006. The clear-sky validobservation status (ValidObs) in MCD43A2 indicates that no clear-skyobservations in Collection V006 were available during this period. Thealbedo over the evergreen needleleaf BC-CR1949 forest site is relativelystable (Fig. 6(c)). The Walker Branch deciduous broadleaf forest site hasno snowfall throughout the year and both the tower and MODIS albedosare only slightly higher during the growing season as compared to thedormant period (Fig. 6(f)).

Table 5 shows the accuracy of the MODIS albedo with respect to the

Fig. 6. Temporal comparison of MCD43 blue-sky albedo with ground measurements at sites (a) Fort Peck, (b) Desert Rock, (c) Douglas fir, (d) Harvard Forest, (e) Skukuza, and (f) WalkerBranch.

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

58

Fig. 6. (continued)

Table 5The accuracy of Collection V006 MCD43 albedo products.

Mixed forest Needleleaf forest Broadleaf forest Savanna/shrubland Grass/cropland/tundra Desert

All quality Bias 0.0015 −0.0076 −0.0009 0.0071 0.0069 0.0023RMSE 0.0201 0.0237 0.0196 0.0125 0.0318 0.0111

Full inversion Bias 0.0013 −0.0061 −0.0005 0.0102 0.0017 0.0023RMSE 0.0141 0.0140 0.0196 0.0134 0.0204 0.0108

Magnitude inversion Bias 0.0016 −0.0085 −0.0012 0.0032 0.0130 0.0029RMSE 0.0233 0.0278 0.0197 0.0113 0.0412 0.0152

Snow-free Bias 0.0020 −0.0065 −0.0007 0.0071 0.0054 0.0023RMSE 0.0193 0.0185 0.0194 0.0125 0.0231 0.0111

Snow Bias −0.0195 −0.0155 −0.0449 N/A 0.0188(0.0147)a

N/A

RMSE 0.0431 0.0462 0.0505 N/A 0.0681(0.0487)a

N/A

a RMSE during fully snow-covered period.

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

59

bias (MODIS albedo minus ground albedo) and the RMSE betweenMODIS and the ground albedo during the snow-free and snow-coveredperiods, as based on the 21 sites listed in Table 1. The ground towerfootprints (whether they be a high tower or a short tower) at the spa-tially representative sites capture a similar variability as that sensed bya MODIS pixel. The RMSE is< 0.0318 and the bias is within± 0.0076for all of the retrievals (of all quality). The RMSEs and biases for thehigh quality full inversions over all the land types are< 0.0204 andwithin±0.0102 respectively and are< 0.0412 and within± 0.0130respectively for the poorer quality magnitude inversions. The RMSEduring the snow-free periods (< 0.0231) are lower than those duringthe snow-covered periods (< 0.0505). The RMSEs of the magnitudeinversions have been improved from 0.0414 (V005) to 0.0233 (V006)for mixed forest, 0.0518 to 0.0278 for needleleaf forest, 0.0296 to0.0197 for broadleaf forest, 0.0483 to 0.0412 for grass/cropland/tundra, and 0.0372 to 0.0152 for desert. The Collection V006 magni-tude inversion RMSE for savanna/shrubland is 0.0113 which is close toCollection V005 (0.0103). This is largely due to the new more timely apriori backup database used in the magnitude inversion, as it is con-tinuously updated with the latest full inversion for each MODIS pixel.The accuracy of the MODIS Collection V006 albedo was also recentlyevaluated at a single heterogenous site in China dominated by crop-lands using 17 ground albedo measurements node and high spatialresolution airborne albedo with a RMSE of 0.013 and bias of 0.013 (Wuet al., 2017).

It must be remembered that the spatial representativeness of vali-dation sites may vary with the surface snow status. During the partlysnow-covered periods, some validation sites which are homogenous andspatially representative during snow-free and/or completely snow-covered periods are more heterogenous and not spatially representativeof the MODIS footprint due to snow patchiness (e.g. the days around theend of year at the Fort Peck site in year 2011 (Fig. 6a)). The CollectionV006 RMSE of grassland/cropland/tundra during the snow-coveredperiod drops from 0.0681 to 0.0487 if only fully snow-covered data areconsidered, decreasing the bias to 0.0147.

Differences between the MODIS albedo products and ground mea-surements are mainly caused by the spatial representativeness ofground measurements, the accuracy of cloud/aerosol/snow status inthe surface reflectance product, and the accuracy of MODIS albedoretrieval quality. The Collection V006 MODIS albedo values are nearlythe same as Collection V005 MODIS albedo during long stable surfaceperiods, although the Collection V006 MODIS albedos offer more valuesthan Collection V005 during periods of surface change (e.g. transitionseasons, and during snow fall and snow melt).

3.3. Capturing rapid surface dynamics

3.3.1. PhenologyFig. 6 and 7 indicate how the Collection V006 values provide more

temporal details. This is especially true for mixed and broadleaf forests

Fig. 7. Time series MCD43 Collection V005 and V006 NBAR-EVI at Harvard Forest in 2009 with corresponding PhenoCam images.

Fig. 8. Temporal Collection V005 and V006 shortwave WSA during the Anaktuvuk River Fire located at the north slope of Alaska in 2007.

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

60

Fig. 9. The time series Collection V006 shortwave WSA during the Anaktuvuk River fire (Latitude 69.2°, Longitude −150.7°) located on the north slope of Alaska in 2007. The black colorarea indicates the boundary of the fire scar on each day. The size of the subset is 200 km by 200 km from tile H12V02 shown in the MODIS equal area Sinusoidal projection.

Fig. 10. The Collection V006 shortwave broadband WSAand Collection V005 shortwave broadband WSA in 2007 (InCollection V005, the day of interest is the first day of the16-day period while in Collection V006, the day of interestis the middle day (9th day) of the 16-day period).Therefore, the DOY of the Collection V005 data has beenoffset by 8 days to match with the Collection V006 values.The subset is located in the north of Alaska with a size of50 km by 50 km. The red star is the location of the Barrowfield site (latitude 71.30°, longitude −156.75°). (For in-terpretation of the references to color in this figure legend,the reader is referred to the web version of this article.)

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

61

in which the albedos are higher during the growing season than duringthe dormant periods. Fig. 7 shows the temporal NBAR derived EVI atHarvard Forest in 2009. Collection V006 EVI has more temporal valuesthan Collection V005 during the transition seasons (e.g. green-up andsenescence). PhenoCam images provide illustrations of key points in thevegetation phenology (Richardson et al., 2007). The EVI remained lowduring the dormant period and increased with the vegetation green-upon DOY 122 and reached maturity on DOY 146. The EVI then decreasedsignificantly from DOY 278 during senescence and became dormantstarting with DOY 305. This is consistent with the results of Shuai et al.(2013) which indicated that the vegetation indices from MODIS DirectBroadcast NBAR product retrieved each day captured the evolution ofvegetation phenological trends over crop, orchard, and forest covers.

3.3.2. DisturbanceThe Collection V006 MCD43 BRDF/NBAR/Albedo product provides

the temporal frequency to capture rapid land surface disturbance. Fig. 8shows the temporal shortwave WSA over one MODIS pixel as it ex-perienced the Anaktuvuk River Fire. The shortwave WSA decreasedsignificantly after the fire with the values dropping from around 0.2to< 0.1. Therefore, the burned area absorbed more solar energy,warming the land surface, and reflected less energy. Recollect thatCollection V006 MODIS BRDF/NBAR/Albedo products assign temporalweights on the observations to emphasize the observations on the dayof interest. Unfortunately, no clear-sky valid observations (ValidObs)were available during days around the day of burning on DOY 245 forthe example pixel in Fig. 8 which increases the uncertainty of the re-trievals during that time. The availability of clear sky observations foreach day can be checked with the Valid Observation layer in QA fieldsof MCD43A2. Note that in Collection V006, a few criteria have beenadded to improve the ability to capture disturbance effects. Only thedata from day of interest are used for the retrieval, if the NIR bandstandard deviation of the values over the 16 days is larger than 0.15.The data that are higher than three times or lower than 0.33 of the dayof interest values are excluded for the BRDF/albedo retrieval. Thesecriteria do not guarantee the separation of observations before and afterthe disturbance if the changes are not significant and obviously it isdifficult if no clear sky observations are available on the actual day ofdisturbance. Jin and Roy (2005) used a modified approach to deriveMODIS albedo and radiative forcing before and after the date ofburning independently — thus keeping the pre and post disturbanceBRDF and Albedo information separate. This methodology captures firedisturbance regions better although it can suffer from a scarcity of clearsky observations due to the smoke. With the Collection V006 MCD43BRDF/NBAR/Albedo product, not only the albedo values, but the spa-tial expansion trajectory of the fire scar can now be monitored on adaily basis (Fig. 9) which can improve the accuracy of regional radia-tive forcing estimates (O'Halloran et al., 2012) and efforts to describe itseffects on climate feedback.

3.3.3. SnowmeltSurface snowmelt dynamics are also better captured by the

Collection V006 MCD43 BRDF/NBAR/Albedo product due to the dailysnow/snow-free retrieval strategies. The trajectory of snowmelt wascaptured at the Fort Peck site in Fig. 6(a). The time series of shortwavealbedo during snowmelt in the area around Barrow, Alaska in 2007 isshown in Fig. 10. The surface snow melt occurs very rapidly and thetransition from totally snow-covered to snow-free only takes about oneweek. The Collection V006 albedo captures more temporal details thandid the 8-day Collection V005 albedo during the snowmelt periodsbecause of this improved snow/snow-free daily retrieval strategy.

4. Conclusion

This study investigates the ability of the Collection V006 MCD43BRDF/NBAR/Albedo products to capture rapid surface changes,

including seasonal vegetation phenology, snowmelt, disturbance, andother land surface dynamics. The Collection V006 MCD43 productsimprove the detection of temporal details primarily by applying tem-poral weights emphasizing the day of interest and utilizing a snow/snow-free retrieval strategy associated with that day of interest.Compared to the 8 day retrieval frequency based on a 16-day retrievalperiod used for Collection V005 products, the daily retrieval frequencyof the Collection V006 MCD43 BRDF/NBAR/Albedo products (based ona 16-day retrieval period) provides an enhanced identification andmonitoring of these dynamics. Note that the days around a day ofburning or around a day of ephemeral snow fall are often contaminatedby smoke, high aerosol, or cloud which does increase the uncertainty ofthe MODIS BRDF/NBAR/Albedo products for that day. However, thesesituations are reflected in poorer associated Quality flags. Furthermore,the availability of observations for each day can be checked with a newValid Observation layer in the MCD43A2 Quality Assessment (QA)product to better understand the retrievals.

Overall, more high quality full inversions are produced in CollectionV006, by using all available valid clear sky observations for the re-trieval, especially at high latitudes. However, the percentage of highquality full inversions in Collection V006 is lower than in CollectionV005 at the mid and low latitudes due to the improvements of cloud/aerosol detection in the surface reflectance product (MOD09GA andMYD09GA) and the presence of additional fill values from non-func-tional or noisy detectors in Collection V006. An estimate of albedouncertainty is also now provided in Collection V006 for climate modelsimulations. Note that a snow or snow-free retrieval in Collection V006is determined by the snow status of the day of interest as opposed to theCollection V005 strategy which determined the snow/snow-free statusby utilizing whichever state was observed for the majority of observa-tions over the 16-day retrieval period.

The Collection V006 MCD43 albedos agree well with spatially re-presentative ground measurements over a variety of land types, evenduring periods of rapid change, with RMSEs of< 0.0318 and biaseswithin± 0.0076 for all quality results. The ground sites for albedovalidation in this study are mainly distributed over Mid-Latitude re-gions although they cover most of the land surface types. Further al-bedo accuracy analysis should be performed to include more high-quality ground albedo measurements over high and low latitude re-gions. These enhancements in the Collection V006 MODIS BRDF/NBAR/Albedo products improve the accuracy of estimates of radiativeforcing and the surface energy budget, and improve our understandingof vegetation phenology during land-atmosphere interactions and cli-mate feedbacks. Note that well distributed multi-angular reflectancescan often be obtained over retrievals periods of< 16 days if multi-sa-tellite observations are utilized (e.g. MISR, MODIS, and VIIRS (Jinet al., 2002)), although differences in spectral range and effective re-solution then do need to be accounted for. However, shorter retrievalperiods can obviously further improve the monitoring of rapid changeand therefore this is an active area of future research.

Acknowledgement

The radiometric measurements for validation are acquired by theBSRN, SURFRAD, AON and FLUXNET communities and in particular bythe following networks: AmeriFlux (US Department of Energy,Biological and Environmental Research, Terrestrial Carbon Program(DE-FG02-04ER63917 and DEFG02-04ER63911)), CarboAfrica,CarboEuropeIP, Canadian Carbon program and OzFlux. Much of thiswork was done under NASA grant NNX12AL3 8G.

References

Amiro, B.D., Barr, A.G., Black, T.A., Iwashita, H., Kljun, N., McCaughey, J.H.,Morgenstern, K., Murayama, S., Nesic, Z., Orchansky, A.L., Saigusa, N., 2006. Carbon,energy and water fluxes at mature and disturbed forest sites, Saskatchewan, Canada.

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

62

Agric. For. Meteorol. 136, 237–251. http://dx.doi.org/10.1016/j.agrformet.2004.11.012.

Anthoni, P.M., Knohl, A., Rebmann, C., Freibauer, A., Mund, M., Ziegler, W., Kolle, O.,Schulze, E.D., 2004. Forest and agricultural land-use-dependent CO2 exchange inThuringia, Germany. Glob. Chang. Biol. 10, 2005–2019. http://dx.doi.org/10.1111/j.1365-2486.2004.00863.x.

Bakwin, P.S., Davis, K.J., Yi, C., Wofsy, S.C., Munger, J.W., Haszpra, L., Barcza, Z., 2004.Regional carbon dioxide fluxes from mixing ratio data. Tellus Ser. B Chem. Phys.Meteorol. 56, 301–311. http://dx.doi.org/10.1111/j.1600-0889.2004.00111.x.

Baum, B.A., Menzel, W.P., Frey, R.A., Tobin, D.C., Holz, R.E., Ackerman, S.A., Heidinger,A.K., Yang, P., 2012. MODIS cloud-top property refinements for collection 6. J. Appl.Meteorol. Climatol. 51, 1145–1163. http://dx.doi.org/10.1175/JAMC-D-11-0203.1.

Burkhart, J.F., Kylling, A., Schaaf, C.B., Wang, Z., Bogren, W., Storvold, R., Sølbo, S.,Pedersen, C.A., Gerland, S., 2017. Unmanned Aerial System nadir reflectance andMODIS Nadir BRDF-adjusted surface reflectances intercompared over Greenland.Cryosphere 11, 1575–1589. http://dx.doi.org/10.5194/tc-11-1575-2017.

Campagnolo, M.L., Sun, Q., Liu, Y., Schaaf, C., Wang, Z., Román, M.O., 2016. Estimatingthe effective spatial resolution of the operational BRDF, albedo, and nadir reflectanceproducts from MODIS and VIIRS. Remote Sens. Environ. 175, 52–64. http://dx.doi.org/10.1016/j.rse.2015.12.033.

Carrer, D., Roujean, J.-L., Meurey, C., 2010. Comparing operational MSG/SEVIRI landsurface albedo products from land SAF with ground measurements and MODIS. IEEETrans. Geosci. Remote Sens. 48. http://dx.doi.org/10.1109/TGRS.2009.2034530.

Cescatti, A., Marcolla, B., Santhana Vannan, S.K., Pan, J.Y., Román, M.O., Yang, X., Ciais,P., Cook, R.B., Law, B.E., Matteucci, G., Migliavacca, M., Moors, E., Richardson, A.D.,Seufert, G., Schaaf, C.B., 2012. Intercomparison of MODIS albedo retrievals and insitu measurements across the global FLUXNET network. Remote Sens. Environ. 121,323–334. http://dx.doi.org/10.1016/j.rse.2012.02.019.

Chapin, F.S., Sturm, M., Serreze, M.C., McFadden, J.P., Key, J.R., Lloyd, A.H., McGuire,A.D., Rupp, T.S., Lynch, A.H., Schimel, J.P., Beringer, J., Chapman, W.L., Epstein,H.E., Euskirchen, E.S., Hinzman, L.D., Jia, G., Ping, C.-L., Tape, K.D., Thompson,C.D.C., Walker, D.A., Welker, J.M., 2005. Role of land-surface changes in arcticsummer warming. Science 310, 657–660. http://dx.doi.org/10.1126/science.1117368.

Chopping, M., Schaaf, C.B., Zhao, F., Wang, Z., Nolin, A.W., Moisen, G.G., Martonchik,J.V., Bull, M., 2011. Forest structure and aboveground biomass in the southwesternUnited States from MODIS and MISR. Remote Sens. Environ. 115, 2943–2953. http://dx.doi.org/10.1016/j.rse.2010.08.031.

Friedl, M.A., Sulla-Menashe, D., Tan, B., Schneider, A., Ramankutty, N., Sibley, A., Huang,X., 2010. MODIS collection 5 global land cover: algorithm refinements and char-acterization of new datasets. Remote Sens. Environ. 114, 168–182. http://dx.doi.org/10.1016/j.rse.2009.08.016.

Gao, F., Jin, Y., Li, X., Schaaf, C.B., Strahler, A.H., 2002. Bidirectional NDVI and atmo-spherically resistant BRDF inversion for vegetation canopy. IEEE Trans. Geosci.Remote Sens. 40, 1269–1278. http://dx.doi.org/10.1109/TGRS.2002.800241.

Gilmanov, T.G., Aires, L., Barcza, Z., Baron, V.S., Belelli, L., Beringer, J., Billesbach, D.,Bonal, D., Bradford, J., Ceschia, E., Cook, D., Corradi, C., Frank, a., Gianelle, D.,Gimeno, C., Gruenwald, T., Guo, H., Hanan, N., Haszpra, L., Heilman, J., Jacobs, a.,Jones, M.B., Johnson, D.a., Kiely, G., Li, S., Magliulo, V., Moors, E., Nagy, Z.,Nasyrov, M., Owensby, C., Pinter, K., Pio, C., Reichstein, M., Sanz, M.J., Scott, R.,Soussana, J.F., Stoy, P.C., Svejcar, T., Tuba, Z., Zhou, G., 2010. Productivity, re-spiration, and light-response parameters of world grassland and agroecosystems de-rived from flux-tower measurements. Rangel. Ecol. Manag. 63, 16–39. http://dx.doi.org/10.2111/REM-D-09-00072.1.

Göckede, M., Foken, T., Aubinet, M., Aurela, M., Banza, J., Bernhofer, C., Bonnefond,J.M., Brunet, Y., Carrara, A., Clement, R., 2007. Quality control of CarboEurope fluxdata—part I: footprint analyses to evaluate sites in forest ecosystems. Biogeosci.Discuss. 4, 4025–4066. http://dx.doi.org/10.5194/bg-5-433-2008.

Granier, A., Ceschia, E., Damesin, C., Dufrêne, E., Epron, D., Gross, P., Lebaube, S., LeDantec, V., Le Goff, N., Lemoine, D., Lucot, E., Ottorini, J.M., Pontailler, J.Y., Saugier,B., 2000. The carbon balance of a young Beech forest. Funct. Ecol. 14, 312–325.http://dx.doi.org/10.1046/j.1365-2435.2000.00434.x.

Hall, D.K., Riggs, G.A., Salomonson, V.V., DiGirolamo, N.E., Bayr, K.J., 2002. MODISsnow-cover products. Remote Sens. Environ. 83, 181–194. http://dx.doi.org/10.1016/S0034-4257(02)00095-0.

He, L., Chen, J.M., Pisek, J., Schaaf, C.B., Strahler, A.H., 2012. Global clumping indexmap derived from the MODIS BRDF product. Remote Sens. Environ. 119, 118–130.http://dx.doi.org/10.1016/j.rse.2011.12.008.

Jin, Y., Roy, D.P., 2005. Fire-induced albedo change and its radiative forcing at thesurface in northern Australia. Geophys. Res. Lett. 32, 1–4. http://dx.doi.org/10.1029/2005GL022822.

Jin, Y., Gao, F., Schaaf, C.B., Li, X., Strahler, A.H., Bruegge, C.J., Martonchik, J.V., 2002.Improving MODIS surface BRDF/albedo retrieval with MISR multiangle observations.IEEE Trans. Geosci. Remote Sens. 40, 1593–1604. http://dx.doi.org/10.1109/TGRS.2002.801145.

Knohl, A., Schulze, E.D., Kolle, O., Buchmann, N., 2003. Large carbon uptake by an un-managed 250-year-old deciduous forest in Central Germany. Agric. For. Meteorol.118, 151–167. http://dx.doi.org/10.1016/S0168-1923(03)00115-1.

Kvalevåg, M.M., Myhre, G., Bonan, G., Levis, S., 2010. Anthropogenic land cover changesin a GCM with surface albedo changes based on MODIS data. Int. J. Climatol. 30,2105–2117. http://dx.doi.org/10.1002/joc.2012.

Lawrence, P.J., Chase, T.N., 2007. Representing a new MODIS consistent land surface inthe Community Land Model (CLM 3.0). J. Geophys. Res. http://dx.doi.org/10.1029/2006JG000168.

Finnigan, J.J., Leuning, R., 2000. Long term flux measurements-coordinate systems andaveraging. In: Proceedings of the International Workshop for Advanced Flux Network

and Flux Evaluation. Center for Global Environmental Research, National Institute forEnvironmental Studies, Japan, Hokkaido, Japan, pp. 51–56.

Lewis, P., Barnsley, M., 1994. Influence of the sky radiance distribution on various for-mulations of the earth surface albedo. In: Proc. Conf. Phys. Meas. Signatures RemoteSens.

Liu, N.F., Liu, Q., Wang, L.Z., Liang, S.L., Wen, J.G., Qu, Y., Liu, S.H., 2013. A statistics-based temporal filter algorithm to map spatiotemporally continuous shortwave al-bedo from MODIS data. Hydrol. Earth Syst. Sci. 17, 2121–2129. http://dx.doi.org/10.5194/hess-17-2121-2013.

Lyapustin, A., Wang, Y., Xiong, X., Meister, G., Platnick, S., Levy, R., Franz, B., Korkin, S.,Hilker, T., Tucker, J., Hall, F., Sellers, P., Wu, A., Angal, A., 2014. Scientific impact ofMODIS C5 calibration degradation and C6+ improvements. Atmos. Meas. Tech. 7,4353–4365. http://dx.doi.org/10.5194/amt-7-4353-2014.

Mack, M.C., Bret-Harte, M.S., Hollingsworth, T.N., Jandt, R.R., Schuur, E.A.G., Shaver,G.R., Verbyla, D.L., 2011. Carbon loss from an unprecedented Arctic tundra wildfire.Nature 475, 489–492. http://dx.doi.org/10.1038/nature10283.

Marek, M.V., Janouš, D., Taufarová, K., Havránková, K., Pavelka, M., Kaplan, V.,Marková, I., 2011. Carbon exchange between ecosystems and atmosphere in theCzech Republic is affected by climate factors. Environ. Pollut. 159, 1035–1039.http://dx.doi.org/10.1016/j.envpol.2010.11.025.

Medlyn, B.E., Robinson, A.P., Clement, R., McMurtrie, R.E., 2005. On the validation ofmodels of forest CO2 exchange using eddy covariance data: some perils and pitfalls.Tree Physiol. 25, 839–857. http://dx.doi.org/10.1093/treephys/25.7.839.

Moors, E.J., Jacobs, C., Jans, W., Supit, I., Kutsch, W.L., Bernhofer, C., Béziat, P.,Buchmann, N., Carrara, A., Ceschia, E., Elbers, J., Eugster, W., Kruijt, B., Loubet, B.,Magliulo, E., Moureaux, C., Olioso, A., Saunders, M., Soegaard, H., 2010. Variabilityin carbon exchange of European croplands. Agric. Ecosyst. Environ. 139, 325–335.http://dx.doi.org/10.1016/j.agee.2010.04.013.

Morgenstern, K., Black, T.A., Humphreys, E.R., Griffis, T.J., Drewitt, G.B., Cai, T., Nesic,Z., Spittlehouse, D.L., Livingston, N.J., 2004. Sensitivity and uncertainty of thecarbon balance of a Pacific Northwest Douglas-fir forest during an El Niño/La Niñacycle. Agric. For. Meteorol. 123, 201–219. http://dx.doi.org/10.1016/j.agrformet.2003.12.003.

Muller, J.-P., Preusker, R., Fischer, J., Zuhlke, M., Brockmann, C., Regner, P., 2007.ALBEDOMAP: MERIS land surface albedo retrieval using data fusion with MODISBRDF and its validation using contemporaneous EO and in situ data products. 2007IEEE Int. Geosci. Remote Sens. Symp. http://dx.doi.org/10.1109/IGARSS.2007.4423326.

Myhre, G., Kvalevåg, M.M., Schaaf, C.B., 2005. Radiative forcing due to anthropogenicvegetation change based on MODIS surface albedo data. Geophys. Res. Lett. http://dx.doi.org/10.1029/2005GL024004.

O'Halloran, T.L., Law, B.E., Goulden, M.L., Wang, Z., Barr, J.G., Schaaf, C., Brown, M.,Fuentes, J.D., Göckede, M., Black, A., Engel, V., 2012. Radiative forcing of naturalforest disturbances. Glob. Chang. Biol. 18, 555–565. http://dx.doi.org/10.1111/j.1365-2486.2011.02577.x.

Oleson, K.W., Bonan, G.B., Schaaf, C., Gao, F., Jin, Y., Strahler, A., 2003. Assessment ofglobal climate model land surface albedo using MODIS data. Geophys. Res. Lett. 30,3–6. http://dx.doi.org/10.1029/2002GL016749.

Peichl, M., Leahy, P., Kiely, G., 2011. Six-year stable annual uptake of carbon dioxide inintensively managed humid temperate grassland. Ecosystems 14, 112–126. http://dx.doi.org/10.1007/s10021-010-9398-2.

Pirazzini, R., 2004. Surface albedo measurements over Antarctic sites in summer. J.Geophys. Res. D: Atmos. 109. http://dx.doi.org/10.1029/2004JD004617.

Platnick, S., King, M.D., Ackerman, S.A., Menzel, W.P., Baum, B.A., Riédi, J.C., Frey, R.A.,2003. The MODIS cloud products: algorithms and examples from terra. IEEE Trans.Geosci. Remote Sens. 41, 459–472. http://dx.doi.org/10.1109/TGRS.2002.808301.

Remer, L. a., Kaufman, Y.J., Tanré, D., Mattoo, S., Chu, D. a., Martins, J. V., Li, R.-R.,Ichoku, C., Levy, R.C., Kleidman, R.G., Eck, T.F., Vermote, E., Holben, B.N., 2005.The MODIS aerosol algorithm, products, and validation. J. Atmos. Sci. 62, 947–973.doi:https://doi.org/10.1175/JAS3385.1.

Richardson, A.D., Jenkins, J.P., Braswell, B.H., Hollinger, D.Y., Ollinger, S.V., Smith,M.L., 2007. Use of digital webcam images to track spring green-up in a deciduousbroadleaf forest. Oecologia 152, 323–334. http://dx.doi.org/10.1007/s00442-006-0657-z.

Román, M.O., Schaaf, C.B., Woodcock, C.E., Strahler, A.H., Yang, X., Braswell, R.H.,Curtis, P.S., Davis, K.J., Dragoni, D., Goulden, M.L., 2009. The MODIS (CollectionV005) BRDF/albedo product: assessment of spatial representativeness over forestedlandscapes. Remote Sens. Environ. 113, 2476–2498. http://dx.doi.org/10.1016/j.rse.2009.07.009.

Román, M.O., Schaaf, C.B., Lewis, P., Gao, F., Anderson, G.P., Privette, J.L., Strahler,A.H., Woodcock, C.E., Barnsley, M., 2010. Assessing the coupling between surfacealbedo derived from MODIS and the fraction of diffuse skylight over spatially-char-acterized landscapes. Remote Sens. Environ. 114, 738–760. http://dx.doi.org/10.1016/j.rse.2009.11.014.

Roujean, J.-L.J.-L., Leroy, M., P-Y, D., Deschamps, P.-Y., 1992. A bidirectional reflectancemodel of the Earth's surface for the correction of remote sensing data. J. Geophys.Res. 97, 20455–20468.

Roy, D.P., Lewis, P., Schaaf, C.B., Devadiga, S., Boschetti, L., 2006. The global impact ofclouds on the production of MODIS bidirectional reflectance model-based compositesfor terrestrial monitoring. IEEE Geosci. Remote Sens. Lett. 3. http://dx.doi.org/10.1109/LGRS.2006.875433.

Roy, D.P., Ju, J., Lewis, P., Schaaf, C., Gao, F., Hansen, M., Lindquist, E., 2008. Multi-temporal MODIS–Landsat data fusion for relative radiometric normalization, gapfilling, and prediction of Landsat data. Remote Sens. Environ. 112, 3112–3130.http://dx.doi.org/10.1016/j.rse.2008.03.009.

Roy, D.P., Zhang, H.K., Ju, J., Gomez-Dans, J.L., Lewis, P.E., Schaaf, C.B., Sun, Q., Li, J.,

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

63

Huang, H., Kovalskyy, V., 2016. A general method to normalize Landsat reflectancedata to nadir BRDF adjusted reflectance. Remote Sens. Environ. 176, 255–271.

Sayer, A.M., Thomas, G.E., Grainger, R.G., Carboni, E., Poulsen, C., Siddans, R., 2012. Useof MODIS-derived surface reflectance data in the ORAC-AATSR aerosol retrieval al-gorithm: impact of differences between sensor spectral response functions. RemoteSens. Environ. 116, 177–188. http://dx.doi.org/10.1016/j.rse.2011.02.029.

Schaaf, C.B., Gao, F., Strahler, A.H., Lucht, W., Li, X., Tsang, T., Strugnell, N.C., Zhang, X.,Jin, Y., Muller, J.-P., Lewis, P., Barnsley, M., Hobson, P., Disney, M., Roberts, G.,Dunderdale, M., Doll, C., D'Entremont, R.P., Hu, B., Liang, S., Privette, J.L., Roy, D.,2002. First operational BRDF, albedo nadir reflectance products from MODIS.Remote Sens. Environ. 83, 135–148. http://dx.doi.org/10.1016/S0034-4257(02)00091-3.

Schaaf, C.B., Cihlar, J., Belward, A., Dutton, E., Verstraete, M., 2009. Albedo and re-flectance anisotropy, ECV-T8: GTOS assessment of the status of the development ofstandards for the terrestrial essential climate variables. In: GTOS Essential ClimateVariable Reports. Rome.

Schaepman-Strub, G., Schaepman, M.E., Painter, T.H., Dangel, S., Martonchik, J.V., 2006.Reflectance quantities in optical remote sensing-definitions and case studies. RemoteSens. Environ. 103, 27–42. http://dx.doi.org/10.1016/j.rse.2006.03.002.

Scholes, R.J., Gureja, N., Giannecchinni, M., Dovie, D., Wilson, B., Davidson, N., Piggott,K., McLoughlin, C., Van der Velde, K., Freeman, A., Bradley, S., Smart, R., Ndala, S.,2001. The environment and vegetation of the flux measurement site near Skukuza,Kruger National Park. Koedoe 44, 73–84. http://dx.doi.org/10.4102/koedoe.v44i1.187.

Shuai, Y., Schaaf, C.B., Strahler, A.H., Liu, J., Jiao, Z., 2008. Quality assessment of BRDF/albedo retrievals in MODIS operational system. Geophys. Res. Lett. http://dx.doi.org/10.1029/2007GL032568.

Shuai, Y., Masek, J.G., Gao, F., Schaaf, C.B., 2011. An algorithm for the retrieval of 30-msnow-free albedo from Landsat surface reflectance and MODIS BRDF. Remote Sens.Environ. 115, 2204–2216. http://dx.doi.org/10.1016/j.rse.2011.04.019.

Shuai, Y., Schaaf, C., Zhang, X., Strahler, A., Roy, D., Morisette, J., Wang, Z., Nightingale,J., Nickeson, J., Richardson, A.D., Xie, D., Wang, J., Li, X., Strabala, K., Davies, J.E.,2013. Daily MODIS 500 m reflectance anisotropy direct broadcast (DB) products formonitoring vegetation phenology dynamics. Int. J. Remote Sens. 34, 5997–6016.http://dx.doi.org/10.1080/01431161.2013.803169.

Strugnell, N.C., Lucht, W., 2001. An algorithm to infer continental-scale albedo fromAVHRR data, land cover class, and field observations of typical BRDFs. J. Clim.http://dx.doi.org/10.1175/1520-0442(2001)014<1360:AATICS>2.0.CO;2.

Toller, G., Xiong, X., Sun, J., Wenny, B.N., Geng, X., Kuyper, J., Angal, A., Chen, H.,Madhavan, S., Wu, A., 2013. Terra and Aqua moderate-resolution imaging

spectroradiometer collection 6 level 1B algorithm. J. Appl. Remote. Sens. 7, 73557.http://dx.doi.org/10.1117/1.JRS.7.073557.

Verma, S.B., Baldocchi, D.D., Anderson, D.E., Matt, D.R., Clement, R.J., 1986. Eddy fluxesof CO2, water vapor, and sensible heat over a deciduous forest. Bound.-LayerMeteorol. 36, 71–91. http://dx.doi.org/10.1007/BF00117459.

Vermote, E.F., El Saleous, N., Justice, C.O., Kaufman, Y.J., Privette, J.L., Remer, L., Roger,J.C., Tanr, D., 1997. Atmospheric correction of visible to middle-infrared EOS-MODISdata over land surfaces: background, operational algorithm and validation. J.Geophys. Res. 102, 17131–17141.

Wan, Z., 2014. New refinements and validation of the collection-6 MODIS land-surfacetemperature/emissivity product. Remote Sens. Environ. 140, 36–45. http://dx.doi.org/10.1016/j.rse.2013.08.027.

Wang, K., Liang, S., Schaaf, C.L., Strahler, A.H., 2010. Evaluation of Moderate ResolutionImaging Spectroradiometer land surface visible and shortwave albedo products atFLUXNET sites. J. Geophys. Res. Atmos. 115. http://dx.doi.org/10.1029/2009JD013101.

Wang, Z., Schaaf, C.B., Chopping, M.J., Strahler, A.H., Wang, J., Román, M.O., Rocha,A.V., Woodcock, C.E., Shuai, Y., 2012. Evaluation of Moderate-resolution ImagingSpectroradiometer (MODIS) snow albedo product (MCD43A) over tundra. RemoteSens. Environ. 117, 264–280. http://dx.doi.org/10.1016/j.rse.2011.10.002.

Wang, Z., Schaaf, C.B., Strahler, A.H., Chopping, M.J., Román, M.O., Shuai, Y.,Woodcock, C.E., Hollinger, D.Y., Fitzjarrald, D.R., 2014. Evaluation of MODIS albedoproduct (MCD43A) over grassland, agriculture and forest surface types during dor-mant and snow-covered periods. Remote Sens. Environ. 140, 60–77. http://dx.doi.org/10.1016/j.rse.2013.08.025.

Wolfe, R.E., Roy, D.P., Vermote, E., 1998. MODIS land data storage, gridding, andcompositing methodology: level 2 grid. IEEE Trans. Geosci. Remote Sens. 36,1324–1338. http://dx.doi.org/10.1109/36.701082.

Wu, X., Wen, J., Xiao, Q., You, D., Liu, Q., Lin, X., 2017. Forward a spatio-temporal trendsurface for long-term ground-measured albedo upscaling over heterogeneous landsurface. Int. J. Digital Earth 0, 1–15. http://dx.doi.org/10.1080/17538947.2017.1334097.

Yu, H., Kaufman, Y.J., Chin, M., Feingold, G., Remer, L.A., Anderson, T.L., Balkanski, Y.,Bellouin, N., Boucher, O., Christopher, S., DeCola, P., Kahn, R., Koch, D., Loeb, N.,Reddy, M.S., Schulz, M., Takemura, T., Zhou, M., 2006. A review of measurement-based assessments of the aerosol direct radiative effect and forcing. Atmos. Chem.Phys. 6, 613–666. http://dx.doi.org/10.5194/acp-6-613-2006.

Zhang, X., Friedl, M. a., Schaaf, C.B., Strahler, A.H., Hodges, J.C.F., Gao, F., Reed, B.C.,Huete, A., 2003. Monitoring vegetation phenology using MODIS. Remote Sens.Environ. 84, 471–475. doi:https://doi.org/10.1016/S0034-4257(02)00135-9.

Z. Wang et al. Remote Sensing of Environment 207 (2018) 50–64

64