LETTER OPEN ACCESS 7KHOLJKW … · reference rainfall dataset because its 3-hourly time resolution enables to document the diurnal cycle (Huffmanetal2007).Dailylandsurfacetemperatures

Environmental Research Letters

LETTER • OPEN ACCESS

The light-deficient climates of western Central African evergreen forestsTo cite this article: Nathalie Philippon et al 2019 Environ. Res. Lett. 14 034007

View the article online for updates and enhancements.

This content was downloaded from IP address 193.51.119.14 on 12/03/2019 at 11:17

https://doi.org/10.1088/1748-9326/aaf5d8

Environ. Res. Lett. 14 (2019) 034007 https://doi.org/10.1088/1748-9326/aaf5d8

LETTER

The light-deficient climates of western Central African evergreenforests

Nathalie Philippon1,10 , GuillaumeCornu2 , LouMonteil3, ValeryGond2, VincentMoron4,5,Julien Pergaud6, Geneviève Sèze7, Sylvain Bigot1, Pierre Camberlin6, CharlesDoumenge2,Adeline Fayolle8 andAlfredNgomanda9

1 IGE,University of Grenoble Alpes, CNRS, IRD,Grenoble INP, 70 rue de la Physique, F-38400 SaintMartin d’Hères, France2 F&S, CIRAD, F-34000Montpellier, France3 Département deGéographie, Université de Laval, 2045 rue de LaTerrasse, Québec, Canada4 AixMarseille Univ, CNRS, IRD, INRA,Coll France, CEREGE, F-13000Aix-en-Provence, France5 IRI, ColumbiaUniversity, Palisades, United States of America6 Biogeosciences, UBFC, 6 bvldGabriel, F-21000Dijon, France7 LMD, IPSL, UPMC,CNRS, EP, ENS, 4 place Jussieu, F-75000 Paris, France8 BioseDepartment andTERRA researchUnit, GemblouxAgro-Bio Tech, Université de Liège, Passage desDéportés 2, B-5030Gembloux,

Belgium9 Institut de Recherche en Ecologie Tropicale (IRET), BP 13354 Libreville, Gabon10 Author towhomany correspondence should be addressed.

E-mail: [email protected]

Keywords: tropical forests, central Africa, irradiance, diurnal cycles, cloudiness

Supplementarymaterial for this article is available online

AbstractRainfall thresholds underwhich forests grow inCentral Africa are lower than those of Amazonia andsoutheast Asia. Attention is thus regularly paid to rainfall whose seasonality and interannual variabilityhas been shown to control Central African forests’water balance andphotosynthetic activity.Nonetheless, light availability is also recognized as a key factor to tropical forests. Therefore this studyaims to explore the light conditions prevailing acrossCentral Africa, and their potential impact onforests’ traits. Using satellite estimates of hourly irradiance, wefindfirst that the fourmain types ofdiurnal cycles of irradiance extracted translate into different levels of rainfall, evapotranspiration, directand diffuse light. Then accounting for scale interactions between the diurnal and annual cycles, weshow that the daily quantity andquality of light considerably vary acrossCentral African forests duringthe annual cycle: the uniqueness ofwesternCentral Africa andGabon in particular, with strongly light-deficient climates especially during themain dry season, points out. Lastly, using anoriginalmapofterrafirme forests, we also show thatmost of the evergreen forests are located inwesternCentral AfricaandGabon.Wepostulate that despitemean annual precipitation below2000mmyr−1, the light-deficient climates ofwesternCentral Africa canharbour evergreen forests because of an extensive low-level cloudiness developing during the June–Septembermain dry season,which strongly reduces thewater demand and enhances the quality of light available for tree photosynthesis. Thesefindings pavetheway for further analyses of the past and future changes in the light-deficient climates ofwesternCentral Africa and the vulnerability of evergreen forests to these changes.

1. Introduction

The climate conditions under which forests grow inCentral Africa are still not well known, as it is one of themost under-studied climatic areas of the world. Thisrelates mainly to the dramatic lack of in situ climaticmeasurements (Washington et al2013, Bigot et al 2016).

As compared to Amazonian and southeast Asianforests, Central African forests grow under lower rain-fall thresholds: mainly below 2000mm yr−1 althoughnoteworthy spatial variations exist between wetter westand east margins (>2000mm), and drier north andsouth ones (∼1500mm, figures S1(a), (e) is availableonline at stacks.iop.org/ERL/14/034007/mmedia).

OPEN ACCESS

RECEIVED

23 September 2018

REVISED

22November 2018

ACCEPTED FOR PUBLICATION

4December 2018

PUBLISHED

11March 2019

Original content from thisworkmay be used underthe terms of the CreativeCommonsAttribution 3.0licence.

Any further distribution ofthis workmustmaintainattribution to theauthor(s) and the title ofthework, journal citationandDOI.

© 2019TheAuthor(s). Published by IOPPublishing Ltd

https://doi.org/10.1088/1748-9326/aaf5d8https://orcid.org/0000-0003-3519-539Xhttps://orcid.org/0000-0003-3519-539Xhttps://orcid.org/0000-0002-7523-5176https://orcid.org/0000-0002-7523-5176mailto:[email protected]://doi.org/10.1088/1748-9326/aaf5d8http://stacks.iop.org/ERL/14/034007/mmediahttps://crossmark.crossref.org/dialog/?doi=10.1088/1748-9326/aaf5d8&domain=pdf&date_stamp=2019-03-11https://crossmark.crossref.org/dialog/?doi=10.1088/1748-9326/aaf5d8&domain=pdf&date_stamp=2019-03-11http://creativecommons.org/licenses/by/3.0http://creativecommons.org/licenses/by/3.0http://creativecommons.org/licenses/by/3.0

This explains that rainfall is acknowledged as themain environmental factor controlling forest photo-synthesis and functioning in Central Africa (Gond et al2013, Guan et al 2015, Cherrington et al 2016).Contrarily to Amazonian forests where the mean sea-sonality in photosynthetic activity is mostly driven bylight availability, i.e. the highest photosynthesislevels are recorded during the luminous dry season(Huete et al 2006, Myneni et al 2007, Bi et al 2015,Wagner et al 2017), in Central Africa, the two seasonsof highest (lowest) photosynthesis, i.e. March–Mayand September–November (December–February andJune–August), are concomitant with the two rainy(dry) seasons (Gond et al 2013, Philippon et al 2016).Indeed, below 2000 mm yr−1 on average, the waterstored during both rainy seasons is not expected tomeet the main dry season water demand and to main-tain evergreen forest.

However, recent studies also show that the westernpart of Central Africa is one of the most cloudy regionacross the tropics (Wilson and Jetz 2016, Dommo et al2018), which accounts for its low mean incomingsolar radiation (irradiance hereafter) at the surface(figures S1(b), (f)). Moreover an analysis of the diurnalcycles for northern Congo (Philippon et al 2016)shows that neither the two rainy seasons nor the twodry seasons resemble each other in terms of cloudcover, irradiance and rainfall, which translates intodifferent levels of photosynthetic activity. In spite ofthat, a clear picture of the light conditions prevailingacross Central Africa is still lacking.

It is also noteworthy that forests are usually seen asa uniform green block across the whole Central Africa(Hansen et al 2008, Verhegghen et al 2012). Accordingto authors ‘lowland rainforests’ or ‘dense moist for-ests’ might encompass a huge variety of forest typesfrom wet Atlantic forests toward moist semi-decid-uous forests further inland (Caballé 1978,White 1983,Letouzey 1985, Fayolle et al 2014b). Further divisionshave also been observed locally (Viennois et al 2013)according to the geological substrate and disturbancehistory (Fayolle et al 2012, 2014a). A forest typology,i.e. evergreenness versus deciduousness, at the regio-nal scale is currently lacking despite its importance forthe improvement of the modelling of land–atmos-phere interactions.

Here, we examine for the first time light availablefor trees across Central Africa and its potential controlon forest traits, particularly forest evergreenness.

To that aim and expanding on a previous study(Philippon et al 2016), we extract first the main typesof diurnal cycles of irradiance. Then we explore theirimplication on variables important for forests suchas potential evapotranspiration (PET) (i.e. waterdemand) and diffuse irradiance (i.e. quality of light).Lastly, accounting for scale interactions between thediurnal and annual cycles, we develop a novel irra-diance-based climatic regionalization for CentralAfrica, that we cross with an original map of forests

types. Themain hypothesis of our study is that the lowirradiance recorded in western Central Africa by redu-cing the water demand but associated with a betterquality of light, enables the existence of evergreenforests under conditions which are drier than inAmazonia and southeast Asia.

2.Data andmethods

Central Africa is defined here as the region encompass-ing latitudes 8°S–7°N and longitudes 8°E–30.5°E.Because we focus on the climate conditions underwhich forests grow, the non-forested land pixels andoceanic pixels are masked out. Only a rapid overviewof data used is given hereafter as details are provided inthe supplementary information. The study period is2005–2013, dictated by the availability of irradi-ance data.

2.1.DataThe diurnal and seasonal cycles of light received at thesurface are documented using mainly direct normal-ized irradiance data (DNI) produced by the CMSAF(Müller et al 2015) (Satellite Application Facility forClimateMonitoring).

To understand variations observed in DNI andtheir implications in terms of water budget and lightquality, information on cloudiness, rainfall, land sur-face temperature, relative humidity, PET and photo-synthetically active radiation (PAR) are analysed usingvarious datasets. This enables the extraction of robustand coherent results across datasets, despite the uncer-tainties associated with the data. Cloudiness data comefrom the Satellite Application Facility for Support toNowcasting and Very Short Range Forecasting CloudType (SAFNWC) product (Derrien and Le Gléau2005). The TRMM 3B42v7 rainfall data is used as thereference rainfall dataset because its 3-hourly timeresolution enables to document the diurnal cycle(Huffman et al 2007). Daily land surface temperaturescome from the Berkeley Earth Temperature Study(Rohde et al 2013) while monthly relative humiditydata originate from the Vmerge project database(Jones and Wint 2015). We also used daily PET pro-vided by the (Senay et al 2008) National Oceanic andAtmospheric Administration’s Global Data Assimila-tion System (GDAS) and monthly PET from World-Clim (Fick and Hijmans 2017). Total and diffuse PAR(400–700 nm) come from the Breathing Earth SystemSimulator (BESS) daily products (Ryu et al 2018).

Lastly, expanding on a forest typology proposedfor northern Congo (Gond et al 2013), forests traits aredocumented using the enhanced vegetation index(EVI) at a 16-day temporal resolution issued fromMODIS (Huete et al 2002), and crossed with forestinventory data and existing vegetation maps (see sup-plementary information).

2

Environ. Res. Lett. 14 (2019) 034007

2.2.Methods2.2.1. Extraction of the four main types of DNI diurnalcyclesFollowing a methodology previously developed (Phi-lippon et al 2016) and extending it for the whole ofCentral Africa, the detection of the four main types ofdiurnal cycles recorded over Central Africa for DNIrelies on a classification of the whole set of 788 400diurnal cycles of hourly DNI (365 days * 9 years from2005 to 2013 and 2400 pixels), and not as commonlyfound on mean diurnal cycles computed on prede-fined seasons (Malhi et al 2002).

The extraction is done thanks to a k-means clus-tering algorithmwhich agglomerates data around ran-domly chosen seeds and iteratively finds the partitionwhich minimizes the variance within clusters, for agiven number of clusters (Michelangeli et al 1995). Aspreliminary steps, we (1) filter the diurnal cycles (low-pass band at 6 h), and (2) normalize them (by subtract-ing the annual hourly mean). Given the large numberof observations available (N=2400 pixels * 365 days *

9 years for 24 variables), the k-means is first performedon a subset of 2000 randomly chosen observations.The number of iterations, i.e. the number of randomseeds is set to 200. The number of clusters varies from3–5 and the 4-cluster solution is retained as the parti-tion for which the four types seem, as expected, mostlydriven by thefirst and second harmonics of the diurnalcycle, that is (1) the opposition between anomalouslydark and bright conditions all along the day (∼ampl-itude of the 1st harmonic) which is (2) superimposedon the temporal phase of the second harmonic (i.e.maximumnear either themorning or the afternoon).

2.2.2. Regionalization of Central Africa according toDNI diurnal cyclesTo delineate irradiance-based regions across CentralAfrica, a k-means clustering is applied to the dailyfrequency of the four types of DNI diurnal cyclesacross the 2400 forest pixels. First, the daily meanfrequency over the 9 years of the study period, for thefour types of diurnal cycles, is computed for each pixel.Then a 30-day low-pass filter is applied to remove thehigh frequency variability. We obtain four meanannual cycles, which are concatenated so that we attaina matrix of 365 days * 4 types as variables and 2400forest pixels as observations to be clustered. Westandardize the variables to zero mean and unitvariance, then reduce the matrix dimensions and filterout the noise through principal component analysis,retaining ∼75% of the total variance, i.e. eightprincipal components. The clustering model is there-fore built in the EOF space with only eight variablestaken into account. The number of clusters is variedfrom 2–9 and a classifiability index is applied todetermine the appropriate number of clusters to beretained. This index rules out partitions into 2–3clusters as not significant at the 95% level and weretain the partition into six clusters.

2.2.3. Compositing of other climatic parametersaccording to the four types of diurnal cycle and/or the sixregionsThe mean diurnal evolution of rainfall, and cloudi-ness, and mean daily values of PET, temperature andrelative humidity, associated with each type of diurnalcycle of DNI are obtained by averaging values over thecorresponding days (obscure, obscure AM K). Therelationships between daily rainfall (fromTRMM) andPET (from GDAS) as well as the one between dailydirect and diffuse irradiance (either fromDNI or PAR)have also been analysed for the 4 types of days. Theconditional and marginal frequencies have beencomputed for deciles of daily rainfall, PET, direct anddiffuse irradiance and PAR computed from all 2400forest grid-points and 3285 days. For daily rainfall, weadded a class for zero rainfall, which accounts forroughly 47% of all forest grid-points and days. Thecounts of observations are normalized by the grandtotal of forest grid-points and days. Then, we com-puted the conditional andmarginal frequencies for thefour daily types of irradiance and computed therelative anomalies versus the ones computed on alldays. If the four types of days would not discriminatein terms of rainfall/PET and direct/diffuse irradiancedeciles, anomalies would be close to zero and notsignificant.

The mean annual cycles of the parameters relativeto climate (rainfallK) observed within each of the sixirradiance-based regions are obtained by averagingvalues over the corresponding pixels (southwestGabon, Cameroon K). For variables provided at adaily resolution, a 30-day low-pass filter is applied onthe computed mean annual cycles to remove the highfrequency variability (e.g.figures S6, S7, S8).

2.2.4. Identifying and mapping forest types according toEVI mean seasonal cycles plus elevation, soils, inventorydata and vegetationmapsTo identify and map forest types with similar trait andphotosynthetic activity, we extended to the whole ofCentral Africa an approach previously developed for arestricted area in the Sangha River Interval, north ofthe Republic of Congo (Gond et al 2013).

The EVI value attributed to each 16-day periodcorresponds the best normalized difference vegetationindex and the lowest zenith angle. Using the ‘good’pixels only (i.e. pixels which are flagged ‘no clouds’) inthe 12 years for each 16-day period, a mean seasonalcycle was computed from the 12-year database. Anunsupervised iterative self-organizing data analysistechnique (ISODATA) classification was applied onthe EVI mean seasonal cycles. The ISODATA classifi-cation is a k-means algorithm which allows selectingclusters by splitting andmerging the initial pixels data-set and has been previously applied for forests inMadagascar (Mayaux et al 2000).

Nine classes were finally retained which were cros-sed with elevation, soils and inventory data, and

3

Environ. Res. Lett. 14 (2019) 034007

existing vegetation maps to corroborate their spatialextend, interpret them in terms of deciduousness anddensity, and label them. In particular we detected sig-nificant and consistent differences in forest structureand composition among the nine remotely-sensedforest types (see supplementary information for adescription of datasets and of the nine types of forests).

3. Results

3.1.Daily variations in direct irradiance inCentralAfricaDNI strongly varies within a day as pictured by thefour main types of diurnal cycles extracted (seeMethods). Obscure days which are the most frequent(29%, figure 1(a)) display maximum DNI as low as

shows that forests in ‘Cameroon’ and ‘SWGabon’ growunder much more days with low PAR and medium tohigh PARdiff than forests further inland (note also infigure 4(b), theDRI/DFI ratio below1during part of theMarch–May andSeptember–November rainy seasons).

3.3. Linkswith forest traitThe six irradiance-based regions identified (figure 3(a))are crossed with a regional map of terra firme forests

(figure 5(a)) developed independently based on EVIseasonality and forest inventory data (seeMethods).Ninedifferent types of forests are identified across CentralAfrica which differ in terms of photosynthetic activity(i.e. showing mean annual cycles of EVI with differentlevels, amplitudes and phases, figure S11), structure (i.e.dense/degraded/secondary) and composition (i.e. ever-green/semi-deciduous/deciduous, table S3).

We find that the spatial distribution of these ninetypes of forests is in good agreement with the six

Figure 1.The fourmain types of diurnal cycles of theDNI. (a)Obscure, (b) obscure in themorning (‘Obscure AM’), (c) obscure in theafternoon (‘Obscure PM’), and (d) bright days. Coloured curves give themean value (average over days and pixels), error bars the 0.5std (withN=1837 830/1422 368/1895 203/2128 900 respectively), and grey bars the overall average (all days and pixels available).In bold, the frequency of occurrence (in%) of each type over the study period (2005–2013) and themean daily values of global (GI),direct (DRI) and diffuse (DFI) irradiances.

5

Environ. Res. Lett. 14 (2019) 034007

irradiance-based regions (figure 5(b)). The ‘SWGabon’, with its dark long dry season with a low waterdemand, andwhere diffuse radiation dominates during

a large part of the year, harbours most of the evergreenanddense forests ofCentral Africa: the ‘dense evergreenforests’, ‘dense evergreen and semi-deciduous forests’

Figure 2.Water demand versus rain, and light quantity versus quality during the four types of diurnal cycles of irradiance. Conditionalfrequencies (expressed in relative anomalies versus all days) for deciles of PET and rainfall (in tenth ofmmper day, panels (a)–(d)) andof diffuse and direct irradiance (inW m−2, panels (e)–(h)). Themarginal frequencies are given for reference in the rightmost columnfor PET or direct irradiance and upper row for rain or diffuse irradiance. Blank squares indicate non-significant anomalies at the two-sided 95% level (seeMethods).

6

Environ. Res. Lett. 14 (2019) 034007

and ‘dense semi-deciduous forests’ of western CentralAfrica (i.e. forest types 1, 3 and 4) are the most repre-sented. But it is also remarkable that the ‘SW Gabon’region closely matches the range of the Okoumé(Aucoumea Klaineana Pierre, figure 3(a)), a pioneerevergreen tree encountered as mono-dominant standsafter shifting cultivation or savannahs, and one ofthe most important timber-producing native tree ofwesternCentral Africa (White et al2000).

4.Discussion

In this study, we propose an original analysis of lightconditions prevailing in Central African forests, andtheir implication on variables relevant for forests’functioning and traits. This analysis relies on the

diurnal cycles of irradiance, the associated levels ofrainfall, ETP and direct versus diffuse light, and theevolution of their frequency along the annual cycle. Itleads to the first irradiance-based regionalization ofCentral African forests.

Our irradiance-based regionalization highlightswestern Central Africa, and more specifically thesouthwestern Gabon, as an area standing apart fromthe rest of Central Africa in terms of mean climatefunctioning: indeed, an important finding is that the‘SW Gabon’ is much darker than the other regions,especially during its main dry season which is also thedriest and longest at the regional scale. This contrastswith most previous climatic regionalizations based onthe mean annual cycles of rainfall only (Dezfuli 2011,Badr et al 2016): they mainly picked-up zonal patternsdriven by the gradual lengthening, on each side of the

Figure 3. Irradiance-based regionalization of Central Africa according to the frequency of the four types of diurnal cycles ofDNI.(a)The six regions are labelled ‘SWGabon’, ‘Cameroon’, ‘CentreDRC’, ‘SouthDRC’, ‘W&Emargins’ and ‘NorthDRC’ plus the rangeofOkoumé as black crosses. (b), (c) Frequency of the four types of diurnal cycles ofDNI across the annual cycle for the ‘SWGabon’and ‘Cameroon’ regions (results for the four other regions are provided in figure S5). Into brackets the number of 0.25° * 0.25° pixelscomprised into each region. Percentages give themean annual frequency of the four types (‘Obscure’ to ‘Bright’ frombottom to top).

7

Environ. Res. Lett. 14 (2019) 034007

Figure 4.Relationships between irradiance, rainfall, potential evapotranspiration, temperature and relative humidity in the sixregions. Scatter-plots of themeanmonthly values of (a)DNI and rainfall, (b)GI andDRI/DFI, (c)PET and rainfall, (d)RHmax andTmax. The vertical and horizontal black dashed lines denote themean annual value across the six regions. The two dry (DJF and JJA)and twowet (MAMand SON) seasons are picturedwith differentmarkers.

Figure 5.The nine types of terrafirme forests of Central Africa and their distributionwithin the six climatic regions. (a) Location of thenine types of terrafirme forests of Central Africa detected based on cross-analyses of inventories data, vegetationmaps andmeanannual cycles of EVI. (b) Frequency (in%) of the nine types of forests within the six climatic regions. The forests types’ colour is thesame as in (a). Empty grey bars give the frequency of the nine forests types within the six regions as obtained from a randompermutation of forests pixels across Central Africa. Chi square equals 2.98 × 107 against a theoretical value of 25.2 for 10° of freedomand 99.5% level of significance.

8

Environ. Res. Lett. 14 (2019) 034007

equator, of the main dry season (and the corresp-onding shortening of the secondary dry season) as aresult of the latitudinal shift of the ITCZ, but fail tomention the darkness of the long dry season inwesternCentral Africa, thus its low water demand and highlevels of diffuse light.

A second key finding is that the light-deficient cli-mates of western Central Africa harbor most of theevergreen forests of Central Africa. Despite the rela-tively low mean annual rainfall (

using HPC resources fromDNUM-CCUB (Universityof Bourgogne FrancheComté).

Authors contribution

NP has designed the study and performed climateanalyses; GC, VG and LM have produced the foresttraits map; VM has analysed the rainfall/PET anddirect/diffuse irradiance relationships; JP andGS havehelped with the retrieval and analysis of cloud data; PChas analysed PET data; AN has provided data for theOkoumé range; SB, CD and AF have helped with theinterpretation of the results with regards to their fieldknowledge. All authors have contributed to thewritingof the paper. The authors declare no conflicts ofinterest.

ORCID iDs

Nathalie Philippon https://orcid.org/0000-0003-3519-539XGuillaumeCornu https://orcid.org/0000-0002-7523-5176

References

Aerts R 1995The advantages of being evergreenTrends Ecol. Evol. 10402–7

BadrHS,Dezfuli AK, Zaitchik B F and Peters-LidardCD2016Regionalizing Africa: patterns of precipitation variability inobservations and global climatemodels J. Clim. 29 9027–43

Bi J, Knyazikhin Y, Choi S, Park T, Barichivich J, Ciais P and FuR2015 Sunlightmediated seasonality in canopy structure andphotosynthetic activity of Amazonian rainforests Environ.Res. Lett. 10 064014

Bigot S et al 2016 État actuel des réseaux demesures éco-climatiquesenAfrique centrale: Les ambitions du projet de rechercheinternational FORGREENE29ème colloque de l’AssociationInternationale de Climatologie ed PCamberlin andYRichard (Dijon) pp 101–6

Bruijnzeel L A,MulliganMand Scatena FN2011Hydrometeorology of tropicalmontane cloud forests:emerging patternsHydrol. Process. 25 465–98

CaballéG 1978 Essai sur la géographie forestière duGabonAdansonia Ser. 2 425–40

Cherrington EA, BarbierN, Ploton P, Vincent G, SabatierD,BergerU and Pélissier R 2016 Equatorial forests displaydistinct trends in phenological variation: a time-seriesanalysis of vegetation index data from three continents IEEEJ. Sel. Top. Appl. EarthObs. Remote Sens. 9 3505–11

DerrienMand LeGléauH2005MSG/SEVIRI cloudmask and typefromSAFNWC Int. J. Remote Sens. 26 4707–32

Dezfuli AK 2011 Spatio-temporal variability of seasonal rainfall inwestern equatorial AfricaTheor. Appl. Climatol. 104 57–69

DommoA, PhilipponN, SezeG andVondouDA2018The June–September low cloud cover inwesternCentral Africa:meandiurnal and spatial patterns and associated atmosphericdynamics J. Clim. 31 9585–603

DoughtyCE, FlannerMGandGouldenML 2010 Effect of smokeon subcanopy shaded light, canopy temperature, and carbondioxide uptake in anAmazon rainforestGlob. Biogeochem.Cycles 24 1–10

Fayolle A, Engelbrecht B, FreyconV,Mortier F, SwaineM,Réjou-MéchainM,Doucet J L, FauvetN, CornuG andGourlet-Fleury S 2012Geological substrates shape tree

species and trait distributions in Africanmoist forests PLoSOne 7 12–4

Fayolle A, PicardN,Doucet J, SwaineM, BayolN, Bénédet F andGourlet-fleury S 2014aAnew insight in the structure,composition and functioning of central Africanmoist forestsFor. Ecol.Manage. 329 195–205

Fayolle A et al 2014b Patterns of tree species composition acrosstropical African forests J. Biogeogr. 41 1–12

Fick S E andHijmans R J 2017WorldClim 2: new 1-km spatialresolution climate surfaces for global land areas Int. J.Climatol. 37 4302–15

Givnish T J 2002Adaptive significance of evergreen versusdeciduous leaves: Solving the triple paradox Silva Fenn. 36703–43

GondV, Fayolle A, Pennec A, CornuG,Mayaux P, Camberlin P,DoumengeC, FauvetN andGourlet-Fleury S 2013Vegetation structure and greenness inCentral Africa fromModismulti-temporal dataPhil. Trans. R. Soc.B 36820120309

GuanK et al 2015 Photosynthetic seasonality of global tropicalforests constrained by hydroclimateNat. Geosci. 8 284–9

HansenMC,RoyDP, Lindquist E, Adusei B, Justice COandAltstatt A 2008Amethod for integratingMODIS and Landsatdata for systematicmonitoring of forest cover and change intheCongoBasinRemote Sens. Drought Innov.Monit.Approaches 112 2495–513

Huete A,DidanK,Miura T, Rodriguez E P, GaoX and Ferreira LG2002Overview of the radiometric and biophysicalperformance of theMODIS vegetation indicesRemote Sens.Environ. 83 195–213

Huete AR,DidanK, Shimabukuro YE, Ratana P, Saleska S R,Hutyra L R, YangW,Nemani RR andMyneni R 2006Amazon rainforests green-upwith sunlight in dry seasonGeophys. Res. Lett. 33 L06405

HuffmanG J, Adler R F, BolvinDT,GuG,Nelkin E J, BowmanKP,HongY, Stocker E F andWolff DB 2007The TRMMmulti-satellite precipitation analysis: quasi-global,multi-year,combined-sensor precipitation estimates atfine scaleJ. Hydrometeorol. 8 38–55

Jones P andWintW2015TheVmerge ProjectDatabase (Oxford:Environmental ResearchGroup)

Letouzey R 1985Notice de la carte phytogeographique duCameroun au 1:500 000 (Toulouse: IRA)

Malhi Y, Pegoraro E,Nobre AD, PereiraMGP,Grace J,Culf AD andClement R 2002 Energy andwater dynamics of acentral Amazonian rain forest J. Geophys. Res. Atmos. 1071–17

Mayaux P, GondV andBartholomé E 2000Anear real-time forestcovermap ofMadagascar derived fromSPOTVEGETATIONdata Int. J. Remote Sens. 21 3139–44

Mercado LM,BellouinN, Sitch S, BoucherO,Huntingford C,WildMandCoxPM2009 Impact of changes in diffuseradiation on the global land carbon sinkNature 458 1014–7

Michelangeli P-A, Vautard R and Legras B 1995Weather regimes:recurrence and quasi stationarity J. Atmos. Sci. 52 1237–56

Müller R, PfeifrothU, Träger-Chatterjee C, Trentmann J andCremer R 2015Digging theMETEOSAT treasure—3 decadesof solar surface radiationRemote Sens. 7 8067–101

Myneni RB et al 2007 Large seasonal swings in leaf area of Amazonrainforests Proc. Natl Acad. Sci. USA 104 4820–3

NamC, Bony S,Dufresne J L andChepferH 2012The too few, toobright tropical low-cloud problem inCMIP5modelsGeophys. Res. Lett. 39 1–7

Oliveira R S, Eller CB, Bittencourt PR L andMulliganM2014Thehydroclimatic and ecophysiological basis of cloud forestdistributions under current and projected climatesAnn. Bot.113 909–20

OuédraogoDY,Mortier F,Gourlet-Fleury S, FreyconV andPicardN2013 Slow-growing species cope best with drought:evidence from long-termmeasurements in a tropical semi-deciduousmoist forest of central Africa J. Ecol. 1011459–70

10

Environ. Res. Lett. 14 (2019) 034007

https://orcid.org/0000-0003-3519-539Xhttps://orcid.org/0000-0003-3519-539Xhttps://orcid.org/0000-0003-3519-539Xhttps://orcid.org/0000-0003-3519-539Xhttps://orcid.org/0000-0003-3519-539Xhttps://orcid.org/0000-0002-7523-5176https://orcid.org/0000-0002-7523-5176https://orcid.org/0000-0002-7523-5176https://orcid.org/0000-0002-7523-5176https://orcid.org/0000-0002-7523-5176https://doi.org/10.1016/S0169-5347(00)89156-9https://doi.org/10.1016/S0169-5347(00)89156-9https://doi.org/10.1016/S0169-5347(00)89156-9https://doi.org/10.1016/S0169-5347(00)89156-9https://doi.org/10.1175/JCLI-D-16-0182.1https://doi.org/10.1175/JCLI-D-16-0182.1https://doi.org/10.1175/JCLI-D-16-0182.1https://doi.org/10.1088/1748-9326/10/6/064014https://doi.org/10.1002/hyp.7974https://doi.org/10.1002/hyp.7974https://doi.org/10.1002/hyp.7974https://doi.org/10.1109/JSTARS.2016.2566670https://doi.org/10.1109/JSTARS.2016.2566670https://doi.org/10.1109/JSTARS.2016.2566670https://doi.org/10.1080/01431160500166128https://doi.org/10.1080/01431160500166128https://doi.org/10.1080/01431160500166128https://doi.org/10.1007/s00704-010-0321-8https://doi.org/10.1007/s00704-010-0321-8https://doi.org/10.1007/s00704-010-0321-8https://doi.org/10.1175/JCLI-D-17-0082.1https://doi.org/10.1175/JCLI-D-17-0082.1https://doi.org/10.1175/JCLI-D-17-0082.1https://doi.org/10.1029/2009GB003670https://doi.org/10.1029/2009GB003670https://doi.org/10.1029/2009GB003670https://doi.org/10.1371/journal.pone.0042381https://doi.org/10.1371/journal.pone.0042381https://doi.org/10.1371/journal.pone.0042381https://doi.org/10.1016/j.foreco.2014.06.014https://doi.org/10.1016/j.foreco.2014.06.014https://doi.org/10.1016/j.foreco.2014.06.014https://doi.org/10.1111/jbi.12382https://doi.org/10.1111/jbi.12382https://doi.org/10.1111/jbi.12382https://doi.org/10.1002/joc.5086https://doi.org/10.1002/joc.5086https://doi.org/10.1002/joc.5086https://doi.org/10.14214/sf.535https://doi.org/10.14214/sf.535https://doi.org/10.14214/sf.535https://doi.org/10.14214/sf.535https://doi.org/10.1098/rstb.2012.0309https://doi.org/10.1098/rstb.2012.0309https://doi.org/10.1038/ngeo2382https://doi.org/10.1038/ngeo2382https://doi.org/10.1038/ngeo2382https://doi.org/10.1016/j.rse.2007.11.012https://doi.org/10.1016/j.rse.2007.11.012https://doi.org/10.1016/j.rse.2007.11.012https://doi.org/10.1016/S0034-4257(02)00096-2https://doi.org/10.1016/S0034-4257(02)00096-2https://doi.org/10.1016/S0034-4257(02)00096-2https://doi.org/10.1029/2005GL025583https://doi.org/10.1175/JHM560.1https://doi.org/10.1175/JHM560.1https://doi.org/10.1175/JHM560.1https://doi.org/10.1029/2001JD000623https://doi.org/10.1029/2001JD000623https://doi.org/10.1029/2001JD000623https://doi.org/10.1029/2001JD000623https://doi.org/10.1080/01431160050145018https://doi.org/10.1080/01431160050145018https://doi.org/10.1080/01431160050145018https://doi.org/10.1038/nature07949https://doi.org/10.1038/nature07949https://doi.org/10.1038/nature07949https://doi.org/10.1175/1520-0469(1995)0522.0.CO;2https://doi.org/10.1175/1520-0469(1995)0522.0.CO;2https://doi.org/10.1175/1520-0469(1995)0522.0.CO;2https://doi.org/10.3390/rs70608067https://doi.org/10.3390/rs70608067https://doi.org/10.3390/rs70608067https://doi.org/10.1073/pnas.0611338104https://doi.org/10.1073/pnas.0611338104https://doi.org/10.1073/pnas.0611338104https://doi.org/10.1029/2012GL053421https://doi.org/10.1029/2012GL053421https://doi.org/10.1029/2012GL053421https://doi.org/10.1093/aob/mcu060https://doi.org/10.1093/aob/mcu060https://doi.org/10.1093/aob/mcu060https://doi.org/10.1111/1365-2745.12165https://doi.org/10.1111/1365-2745.12165https://doi.org/10.1111/1365-2745.12165https://doi.org/10.1111/1365-2745.12165

PhilipponN et al 2016Analysis of the diurnal cycles for a betterunderstanding of themean annual cycle of forests greennessinCentral AfricaAgric. For.Meteorol. 223 81–94

RapA et al 2015 Fires increase Amazon forest productivity throughincreases in diffuse radiationGeophys. Res. Lett. 42 4654–62

RohdeR,Muller R, JacobsenR, Perlmutter S andMosher S 2013Berkeley earth temperature averaging processGeoinform.Geostat. 1 1–13

RyuY, JiangC, KobayashiH andDettoM2018MODIS-derivedglobal land products of shortwave radiation and diffuse andtotal photosynthetically active radiation at 5 km resolutionfrom2000Remote Sens. Environ. 204 812–25

SchwartzD 1993 Les retombées de litière en tant que source ducarbone et de l’azote du sol : quantification et périodicité desapports en relation avec les caractéristiques climatiques etédaphiques dans deux parcelles de forêt dense àDimonika(Mayombe, Congo). Echanges forêt-Atmosphère enMilieuTropical Humide : Recueil de Travaux Effectués dans leMayombe ed BCros, J Diamouangana andMKabala (Paris:UNESCO)pp 141–58

SenayGB,Verdin J P, LietzowR andMelesse AM2008Global dailyreference evapotranspirationmodeling and evaluation J. Am.Water Resour. Assoc. 44 969–79

VerhegghenA,Mayaux P,DeWCandDefourny P 2012MappingCongo basin forest types from 300 mand 1 kmmulti-sensor

time series for carbon stocks and forest areas estimationBiogeosci. Discuss. 9 7499–553

Viennois G, BarbierN, Fabre I andCouteron P 2013Multiresolution quantification of deciduousness inwest-central African forestsBiogeosciences 10 6957–67

Wagner FH,Hérault B, Rossi V,Hilker T,Maeda E E, Sanchez A,Lyapustin A I, Galvão L S,Wang Y andAragão L EOC2017Climate drivers of the Amazon forest greeningPLoSOne 121–15

WashingtonR, James R, PearceH, PokamWMandMoufouma-OkiaW2013Congo Basin rainfall climatology:canwe believe the climatemodels?Phil. Trans. R. Soc.B 36820120296

White F 1983TheVegetation of Africa: ADescriptiveMemoir toAccompany theUNESCO/AETFAT/UNSOVegetationMap ofAfrica (Paris: UNESCO)

White L, Oslisly R, Abernethy K andMaley J 2000 L’Okoumé(AucoumeaKlaineana): expansion et déclin d’un arbrepionnier enAfriqueCentrale Atlantique au cours del’HolocèneDynamique à Long Terme desÉcosystèmesForestiers Intertropicaux edMServant and S Servant-Vildary(Paris: IRD) pp 399–411

WilsonAMand JetzW2016Remotely sensed high-resolutionglobal cloud dynamics for predicting ecosystem andbiodiversity distributions PLoS Biol. 14 1–20

11

Environ. Res. Lett. 14 (2019) 034007

https://doi.org/10.1016/j.agrformet.2016.04.005https://doi.org/10.1016/j.agrformet.2016.04.005https://doi.org/10.1016/j.agrformet.2016.04.005https://doi.org/10.1002/2015GL063719https://doi.org/10.1002/2015GL063719https://doi.org/10.1002/2015GL063719https://doi.org/10.4172/2327-4581.1000103https://doi.org/10.4172/2327-4581.1000103https://doi.org/10.4172/2327-4581.1000103https://doi.org/10.1016/j.rse.2017.09.021https://doi.org/10.1016/j.rse.2017.09.021https://doi.org/10.1016/j.rse.2017.09.021https://doi.org/10.1111/j.1752-1688.2008.00195.xhttps://doi.org/10.1111/j.1752-1688.2008.00195.xhttps://doi.org/10.1111/j.1752-1688.2008.00195.xhttps://doi.org/10.5194/bgd-9-7499-2012https://doi.org/10.5194/bgd-9-7499-2012https://doi.org/10.5194/bgd-9-7499-2012https://doi.org/10.5194/bg-10-6957-2013https://doi.org/10.5194/bg-10-6957-2013https://doi.org/10.5194/bg-10-6957-2013https://doi.org/10.1371/journal.pone.0180932https://doi.org/10.1371/journal.pone.0180932https://doi.org/10.1371/journal.pone.0180932https://doi.org/10.1371/journal.pone.0180932https://doi.org/10.1098/rstb.2012.0296https://doi.org/10.1098/rstb.2012.0296https://doi.org/10.1371/journal.pbio.1002415https://doi.org/10.1371/journal.pbio.1002415https://doi.org/10.1371/journal.pbio.1002415

1. Introduction2. Data and methods2.1. Data2.2. Methods2.2.1. Extraction of the four main types of DNI diurnal cycles2.2.2. Regionalization of Central Africa according to DNI diurnal cycles2.2.3. Compositing of other climatic parameters according to the four types of diurnal cycle and/or the six regions2.2.4. Identifying and mapping forest types according to EVI mean seasonal cycles plus elevation, soils, inventory data and vegetation maps

3. Results3.1. Daily variations in direct irradiance in Central Africa3.2. Seasonal and spatial variations across Central Africa3.3. Links with forest trait

4. Discussion5. ConclusionAcknowledgmentsAuthors contributionReferences