Assessment of CHADFDM satellite-based input dataset for the groundwater recharge estimation in arid and data scarce regions

RESEARCH ARTICLE Assessmen o CHADFDM sa elli e-based inpu da ase o he g oundwa e echa ge es ima ion in a id and da a sca ce egions Na iseh Salehi Sia ashani 1,2,3 | Joaquin Jimenez-Ma inez 4,5 | Guille mo Vaque o 2,3 | F ancisco J. Elo za 6 | Jus in She ield 7 | Lucila Candela 2 | Aleix Se a -Capde ila 8 1 Depa men o Ci il and En i onmen al Enginee ing, Technical Uni e si y o Ca alonia, Ba celona, Spain 2 IMDEA Agua, Alcal a de Hena es, Spain 3 Fundaci on G omez Pa do, Mad id, Spain 4 Depa men o Wa e Resou ces and D inking Wa e , Eawag, Dübendo , Swi ze land 5 Depa men o Ci il, En i onmen al and Geoma ic Enginee ing, ETH Zu ich, Zü ich, Swi ze land 6 School o Mining and Ene gy Enginee ing, Technical Uni e si y o Mad id, Mad id, Spain 7 Depa men o Geog aphy, Uni e si y o Sou hamp on, Sou hamp on, UK 8 Wa e Global P ac ice, The Wo ld Bank, Washing on, D.C., USA Co espondence Na iseh Salehi Sia ashani, Depa men o Ci il and En i onmen al Enginee ing, Technical Uni e si y o Ca alonia, Ba celona, Spain. Email: na iseh.salehi.sia asha[email p o ec ed] Abs ac Aqui e na u al echa ge es ima ions a e a p e equisi e o unde s anding hyd ologic sys ems and sus ainable wa e esou ces managemen . As me eo ological da a se ies collec ion is di icul in a id and semia id a eas, sa elli e p oduc s ha e ecen ly become an al e na i e o wa e esou ces s udies. A daily g oundwa e echa ge es ima ion in he NW pa o he Lake Chad Basin, using a soil–plan -a mosphe e model (VisualBALAN), om g ound- and sa elli e-based me eo ological inpu da ase o non-i iga ed and i iga ed land and o he 2005–2014 pe iod is p esen ed. A e - age annual alues we e 284 mm and 30C o p ecipi a ion and empe a u e in g ound-based gauge s a ions. Fo he sa elli e-model-based Lake Chad Basin Flood and D ough Moni o Sys em pla o m (CHADFDM), a e age annual p ecipi a ion and empe a u e we e 417 mm and 29C, espec i ely. Unce ain ies de i ed om sa elli e da a measu emen could accoun o he ain all di e ence. The es ima ed mean annual aqui e echa ge was always highe om sa elli e- han g ound-based da a, wi h di e ences up o 46% o d yland and 23% in i iga ed a eas. Recha ge esponse o ain all e en s was e y a iable and esul s we e e y sensi i e o: wil ing poin , ield capaci y and cu e numbe o uno es ima ion. Ob ained esul s p o ide plausible echa ge alues beyond he unce ain y ela ed o da a inpu and modelling app oach. This wo k p e en s on he impo an de ia ions in echa ge es i- ma ion om weigh ed-ensemble sa elli e-based da a, in o ming in decision making o bo h s akeholde s and policy make s. KEYWORDS CHADFDM da a se , g ound-sa elli e me eo ological da a, g oundwa e echa ge modelling, Lake Chad Basin Recei ed: 29 May 2020 Re ised: 5 May 2021 Accep ed: 14 May 2021 DOI: 10.1002/hyp.14250 This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion-NonComme cial-NoDe i s License, which pe mi s use and dis ibu ion in any medium, p o ided he o iginal wo k is p ope ly ci ed, he use is non-comme cial and no modi ica ions o adap a ions a e made. © 2021 The Au ho s. Hyd ological P ocesses published by John Wiley & Sons L d. Hyd ological P ocesses. 2021;35:e14250. wileyonlinelib a y.com/jou nal/hyp 1o 15 h ps://doi.o g/10.1002/hyp.14250 1|INTRODUCTION Sca ce p ecipi a ion, in e annual d y pe iods, ma ked seasonal and spa ial a iabili y and ex eme ain all e en s a e he main ea u es o a id and semia id zones ha condi ion g oundwa e echa ge (Le ne e al., 1990). The e o e, eliable ain all and empe a u e da a in hese a eas a e ex emely impo an o accu a ely assessing bo h soil wa e balance and aqui e echa ge (Wu e al., 1996). Beyond inhe en unce ain ies o a ailable in si u da a, ha is, om g ound-based wea he s a ions, one o he main di icul ies is ield da a sca ci y (Bhowmik & Cos a, 2014; Dumola d e al., 2007). Me eo ological s a- ions a e gene ally sca ce, une enly dis ibu ed, and may ope a e du - ing sho pe iods o imply majo gaps. This can, in u n, lead o p oblems i long- e m alues a e equi ed when only sho da ase pe iods a e a ailable (Le ne e al., 1990). Da a and in o ma ion a e also o en sca e ed among di e en agencies wi h limi ed access, which makes i di icul o ob ain comple e eco ds. This gene ally esul s in in o ma ion ha is no e y sui able o p ac ical needs. To o e come his challenge, esea che s ha e adi ionally inc eased da a a ailabili y using di e en s a is ical me hods (Wagne e al., 2012), especially by b idging gaps in ime se ies, in e pola ing be ween da a poin s, in oducing unce ain ies gi en ain all in e mi ency and la ge dis ances be ween obse a ion poin s. An al e na i e and complemen a y sou ce o in o ma ion comes om sa elli e p oduc s (e.g., TRMM, CMORPH, TMPA, PERSIANN), which can p o ide me eo ological da a se ies o e la ge a eas, and a e use ul o making mul iple applica ions in hyd ology (Beck, an Dijk, e al., 2017; Beck, Ve gopolan, e al., 2017; Habib e al., 2008; Mzi ai e al., 2005; She ield e al., 2018; Velpu i & Senay, 2013). Fo exam- ple, hese sou ces o in o ma ion ha e been used o de elop la ge- scale d ough moni o ing sys ems wi h also he goal o s akeholde s use in mind, which moni o in nea eal- ime he e es ial wa e cycle based on emo e sensing da a and land su ace hyd ological modelling (She ield e al., 2014). Gi en he gene al g ound da a sca ci y in a id and semia id zones, sa elli e emo e sensing p oduc s a e a po en ially use ul di ec o indi ec da a sou ce o mos hyd ological cycle componen s. Ne e - heless, sa elli e p oduc s use is in luenced/limi ed by unce ain ies esul ing om i s wide space– ime co e age and spa ial esolu ion, which is especially ele an in hese egions due o he high spa ial and empo al a iabili y o su ace me eo ology and ele a ion- dependen biases (Bi ew & Geb emichael, 2010). In some cases, sa el- li e senso s may o e es ima e p ecipi a ion gi en i s abili y o iden i y e en s no eco ded by gauge da a (Milewski e al., 2009). Besides, sa elli e-based physical senso limi a ions (Knoche e al., 2014; P igen , 2010) o changes in senso s ha lead o empo al non- homogenei y a e also challenges as o hei use. To unde s and/ e i y he u ili y o such p oduc s o hyd ological s udies, i is necessa y o compa e sa elli e da a wi h g ound-based da a in o de o indica e he hyd ologic sys em esponse o hese inpu s and a ange o p edic- ions. Much wo k has been done especially on p ecipi a ion, (e.g., Beck, an Dijk, e al., 2017; Beck, Ve gopolan, e al., 2017; Haile e al., 2015; Jiang, Yu, e al., 2016; Jiang, Zhou, e al., 2016; Lu, Sun, e al., 2016; Lu, Wang, e al., 2016; Noguei a e al., 2018). Fo g ound- wa e echa ge quan i ica ions based on he wa e balance equa ion, he use o ield and sa elli e p oduc s has been assessed in simila cli- ma e a eas, anging be ween less han 100 km 2 and 10 4 km 2 , as sum- ma ized by Coelho e al. (2017). Mos app oaches gene ally use ain all and e apo anspi a ion inpu s o s udies conduc ed in he Gua ani aqui e (Lucas e al., 2015), Pakis an (Usman e al., 2015) and in he Wes Bank (Khala & Donoghue, 2012); he la ely also inco po- a es i iga ion da a pa ame e s and su ace wa e inpu . Fo a id No he n B azil, he wo k o Coelho e al. (2017) conside s spa ially a ying uno and soil mois u e in he wa e balance equa ion, along wi h ancilla y land use/land co e da a. Wu e al. (2019) de e mined he aqui e annual and long- e m echa ge end a egional scale wi h GRACE and GLDAS in he semia id egion o he O dos Basin, China. The e was no ob ious long- e m end obse ed, and he annual echa ge can be explained by he a iabili y in p ecipi a ion. Fo modelling pu poses, se e al physically based dis ibu ed nume ical codes, ound in di e en wa e balance app oaches such as SAHYSMOD (ILRI, 2005), VisualBALAN (Sampe e al., 2005), TOPOG (CSIRO, 2008) o SWB2 (Wes enb oek e al., 2018), among o he s, a e cu en ly applied. The gene ally a ailable eques ed inpu da a, es ima ed easonably accu a ely, and he acili y o modi y o subs i- u e di e en inpu da ase s, make VisualBALAN an excellen candi- da e o echa ge calcula ions. VisualBALAN 2.0 o e s an in e media e le el o di icul y. This model code es ima es a sequen ial wa e balance o he soil, he unsa u a ed zone and he aqui e , and has p o en success ul in calcula ing g oundwa e echa ge in a ious hyd ogeological condi ions (Sampe e al., 1999; Espinha-Ma ques e al., 2011; Touhami e al., 2013). The quan i ica ion o na u al g oundwa e echa ge is a basic equi emen o e icien wa e esou ces managemen . The di use echa ge is a complex unc ion ha esul s om he coupling o se - e al ac o s: p ecipi a ion ( olume, in ensi y, du a ion), ai empe a- u e, opog aphy, ege a ion (c opping pa e n, oo ing dep h) and e apo anspi a ion, soil and subsoil ypes, low mechanisms in he unsa u a ed zone, bed ock geology and a ailable g oundwa e s o age (Scanlon e al., 2002). O hese, p ecipi a ion and e apo anspi a ion a e he sys em's d i en o ces. Depending on elie , no concen a ed echa ge om uno and ponding may be also dominan mechanisms in a id en i onmen s. Se e al e iews on aqui e echa ge quan i ica- ion based on di e en me hods ha e been conduc ed in he pas , and ha e ocused p ima ily on a id and semia id egions (de V ies & Simme s, 2002; Moeck e al., 2020; Scanlon e al., 2002; Scanlon e al., 2006). Al hough he spa io- empo al dis ibu ion o p ecipi a- ion is he mos c i ical ac o (Wu e al., 1996), he chosen me hod can make es ima ions highly a iable (Leduc e al., 2000). Fo egional scale s udies, he soil wa e balance me hod is widely used due o i s e sa ili y o es ima e spa ially and empo ally dis ibu ed aqui e echa ge. This analysis calcula es he empo al (e.g., daily) esponse o e a wide a ea based on a physically obus echa ge es ima e p o- cess, aking in o accoun he condi ions o he land co e and he ain- all and i iga ion con ibu ions. Wi h ega d o he me hods, i p esen s mo e complexi y as i also akes in o accoun u he 2o 15 SALEHI SIAVASHANI ET AL. in o ma ion om di e en sou ces on clima ic, soil da a, adose zone/ aqui e pa ame e s and ege a ed a eas. This wo k ocuses on he Lake Chad Basin, an a id egion in which su ace wa e is no enough o ul il u ban and u al popula ion needs, and g oundwa e is he main wa e supply. Na u al echa ge (di use) o he egion, which is widely a iable on spa ial and empo al scales, s ill emains unce ain. In he las 40 yea s, much a en ion has been paid o imp o e echa ge es ima ions in his a ea, gene ally by local esea ch. Me hods o echa ge es ima ion mainly included iso opic s udies (Djo e & T a i, 2001; Edmunds e al., 1998; Gaul ie , 2004; Goni, 2006; Leduc e al., 2000; Ngounou Nga cha, Mud y, A anyossy, e al., 2007; Nji choua & Ngounou Nga cha, 1997; Tewolde e al., 2019) and ma hema ical modelling o di e en hyd ologic objec i es (Babama'aji, 2013; Leblanc, 2002). Main indings indica e maximum alues in he sou he n pa o he Lake Chad Basin (Sou h o 14 h pa allel) and in no he n bounda y pa o he Lake, while i is almos inexis en in he no he n pa o he Basin. Di e en esea ch s udies ha e applied emo e sensing da a (e.g., Me eosa he mal da a), combined wi h hyd ogeological da a, o g oundwa e esea ch pu poses. Leblanc (2002) and Leblanc e al. (2007) wo ked a he basin le el o iden i y su ace indica o s o echa ge and discha ge a eas o g oundwa e modelling. Buma e al. (2016) in e ed he e ec o ain- all on wa e s o age on he basin by applying emo e sensing da ase s om he G a i y Reco e y and Clima e Expe imen (GRACE) and Global Land Da a Assimila ion Sys em (GLDAS) o wa e s o age and soil mois u e, espec i ely. Subsu ace wa e a ia ions we e com- pa ed wi h g oundwa e ou pu s om a global hyd ological model ha showed a simila pa e n. The main objec i e o his wo k was o explo e changes in g oundwa e di use echa ge; ha is, pe o mance e alua ion o cli- ma e da a inpu s in a gene a ed daily wa e budge om a soil–plan - a mosphe e model based on he wa e balance by conside ing wo me eo ological da a sou ces and exis en ag icul u al i iga ion: (i) a ailable g ound-based me eo ological da a om local s a ions s o ed in he T ans-A ican Hyd o-Me eo ological Obse a o y, TAHMO (Van de Giesen e al., 2014); (ii) sa elli e-based ain all om he Mul i-Sou ce Weigh ed-Ensemble P ecipi a ion (MSWEP, Beck, an Dijk, e al., 2017; Beck, Ve gopolan, e al., 2017; Beck e al., 2018; Beck e al., 2019) and ai empe a u e da a om he P ince on Global Fo cing (PGF, She ield e al., 2006) p o ided by he Lake Chad Basin Flood and D ough Moni o Sys em pla o m (CHADFDM; Amani e al., 2021; She ield e al., 2014). The CHADFDM was de eloped o hyd ologic applica ions by P ince on Uni e si y in collabo a ion wi h ICIWaRM (In e na ional Cen e o Wa e Resou ces Managemen ) and UNESCO-IHP. The ul ima e goal is o assess he sui abili y o sa elli e p oduc s (based in a use - iendly pla o m CHADFDM) o de elop spa ially and empo ally es ima es o g oundwa e echa ge o egional man- agemen s udies wi h he goal o s akeholde use in mind. This analy- sis essen ially p o ides ‘a con olled expe imen ’ o assess he e ec s o changes d i en only by unning he model a ying da a inpu om he wo sou ces. The echa ge es ima ions we e ca ied ou o he 2005–2014 pe iod. he eliabili y and unce ain y o he es ima ed echa ge alues in he non-i iga ed a ea om he wo da a sou ces as ega ds clima e and soil pa ame e s we e e alua ed by a sensi i i y analysis o he model pa ame e s. 2|STUDY AREA The s udy a ea is loca ed in he no h-wes e n pa o he Lake Chad basin sys em, which co e s an a ea o abou 155 000 km 2 o he basin in Chad, Nige and Nige ia (Figu e 1). The Lake Chad size is highly a - iable ac oss seasons and yea s explained by ain all a ia ions o e i s basin, which lead o a wide a iabili y in i e lows and lake inpu , pa icula ly o e he Cha i-Logone Ri e Basin, which may accoun o abou 95% o he wa e in lows o he lake. A egional le el, aqui e exchanges wi h su ace wa e and di use echa ge cons i u e he main con ol mechanism o g oundwa e le el. This egion is ca ego- ized as a id o semia id, wi h a main ain all season be ween Ap il and Sep embe . A e age annual p ecipi a ion anges om 20 o 600 mm, daily empe a u es be ween 8 and 45C, and mean annual po en ial e apo a ion exceeds 2000 mm (LCBC-GIZ, 2016; Mahmood & Jia, 2018). A succession o d y pe iods in he las 100 yea s has led o a se e e deple ion o he lake a ea om 22 000 km 2 o 8 000 km 2 oday, which has d as ically educed he ex ension o seasonally inun- da ed i e plains. The associa ed impac on in il a ion and g ound- wa e echa ge emains unknown, bu is likely impo an . The egion is ela i ely la wi h gen le slopes (10%) om he highlands in he NW owa ds he SE. Land co e is g assland (65%), ba e land (20%), spa se ege a ion (5% e.g., acacias) and c oplands ( ain ed and i iga ed, 10%) (LCBC-GIZ, 2016). Non-i iga ed c ops a e mille , so ghum, co n and ice. In he s udied a ea, c op i iga ion (5000 mm/y , mainly peppe s; LCBC-IRD, 2016) is done by combining g oundwa e and su ace wa e om Yobe i e and co e s app oxi- ma ely 21 295 km 2 . On he egional scale, he geology o he s udy a ea consis s o ma e ials om he P ecamb ian, Mesozoic (C e aceous), and Plio- Qua e na y (BRGM, 1994; Bu ke, 1976; Schneide , 1989; Schneide & Wol , 1992). The C e aceous is p edominan ly con inen al and, along wi h Miocene o ma ions, i is known as he Con inen al Te minal. Plio-Qua e na y deposi s comp ise lu io-lacus ine, lu io-del aic and aeolian ma e ial. Ou c ops o igneous ocks comple e he geology in he egion. F om a hyd ogeological poin o iew, h ee aqui e s a e dis inguished in he a ea (Schneide & Wol , 1992), Qua e na y, Lowe Pliocene and Con inen al Te minal, and a e mainly con ined. Only he g oundwa e echa ge o he Qua e na y uppe uncon ined aqui e is he objec i e o his esea ch wo k (Figu e 1). 3|METHODOLOGY A soil wa e balance modelling app oach was applied o es ima e he g oundwa e di use echa ge o he Qua e na y uncon ined aqui e o he 2005–2014 pe iod. Fo modelling pu poses, he model was se up and he alues o he plan –soil-hyd ologic pa ame e s emained SALEHI SIAVASHANI ET AL.3o 15 in a ian as only he clima ological da a inpu s we e modi ied. Two sou ces o p ecipi a ion (P) and ai empe a u e (T) da a we e used: (i) g ound-based da a om he selec ed local me eo ological s a ions; (ii) sa elli e-based ain all and ai empe a u e da a downloaded om he CHADFDM (Amani e al., 2021; She ield e al., 2014) o he same ime pe iod and geog aphical loca ion (coo dina es) o he ield s a ions. The Pand Tda a downloaded om he CHADFDM pla o m a e always e e ed o as sa elli e da a h oughou he ex . The explo a o y inspec ion o he 9-yea se ies o me eo o- logical daily da a o each da a sou ce in ol ed an analysis wi h sim- ple desc ip i e s a is ics (mean, median, mode, s anda d de ia ion). Ou -o -sample es ing was conduc ed o assess i he s a is ical analysis esul s could be gene alized o an independen da ase . The Pand Tda a de i ing om he CHADFDM we e checked agains he g ound da a o e alua e empo al hyd ological pa e ns, o alida e assump ions and o compa e he esul s. A dependence assessmen (co ela ion) be ween he wo da ase s was made wi h a eg ession analysis and goodness-o - i was assessed by he coe - icien o de e mina ion (R 2 ) o iden i y he possible p esence o ends. Annual p ecipi a ion as cumula i e de ia ion om he mean was calcula ed. Finally, a s a is ical da a analysis was ca ied ou wi h he EXCEL Analysis Toolpack 2013 (Mic oso ® ). A sensi i i y analysis was also ca ied ou on he aqui e echa ge es ima ions o he non-i iga ed a ea. 3.1 |Soil wa e balance model desc ip ion Aqui e echa ge was calcula ed using VisualBALAN .2.0 (Sampe e al., 2005), a nume ical model o simula e wa e balance in soil, adose zone and aqui e wi h a use - iendly modula design. I equi es speci ic in o ma ion on clima e, land use, land co e and opog aphy o simula e echa ge p ocess. This nume ical ool has been success ully applied in di e en a eas (e.g., Candela e al., 2016; Jiménez-Ma ínez e al., 2010, among many o he s). The model is di ided in o h ee sub models (Figu e 2), which conside s he p ocesses in: (i) he uppe pa o soil ( oo zone) o sol e he c i ical in e ac ions o he soil–plan –a mosphe e con inuum; (ii) he adose o unsa u a ed zone (below he oo zone); (iii) he sa u a ed zone (aqui e ). Thewa e balance(inwa e dep h) o a ege a ed soil is ep esen ed by: P+I In Es ETaR=Δθ: whe e P(mm) ep esen s p ecipi a ion, I (mm) i iga ion, In (mm) canopy in e cep ion, Es (mm) uno , ETa (mm) ac ual e apo- anspi a ion, R(mm) po en ial echa ge o he adose zone and Δθ a ia ion in soil wa e s o age (mm). P ecipi a ion (P) and i iga ion (I ) a e dis ibu ed be ween su ace uno and in il a ion. The In olume is a ac ion o he p ecipi a ion in e cep ed by ege a ion (lea es, b anches, unks) o deal wi h wa e FIGURE 1 The Lake Chad Basin and loca ion o he s udy a ea. Exis ing g ound-based me eo ological s a ions a e indica ed on he map 4o 15 SALEHI SIAVASHANI ET AL. loss by e apo a ion. The esidual wa e a e e apo anspi a ion, in il- a ion e m I(mm), was included in he wa e balance equa ions as: I(ETa–R)=Δθ P+I In Es =I Pa o in il a ion comes back o he a mosphe e by e apo anspi a- ion (ETa), ano he pa inc eases he wa e con en in soil and he es con ibu es o po en ial echa ge, he wa e inpu in he unsa u a ed zone. In his zone, low di ec ion can p esen a ho izon al componen as in e low, o can e ically pe cola e o he aqui e (R)(Figu e2). Thecodeallows hewa e balance obesol edinamul icellpa e n by conside ing he de ined zones, assumed homogeneous in ela ion o soil pa ame e s land use, aqui e and me eo ological da a. The model simula es he empo al di e ences (be ween ini ial, i , and inal ime, , Δ =  i ) o he ac ual e apo anspi a ion and g oundwa e echa ge. The model ou pu s a e he wa e balance componen s ( uno , in e low, ege al in e cep ion, ETa, g oundwa e echa ge) exp essed as daily a es. The echa ge is es ima ed acco ding o he assump ion o homoge- nei y and iso opy o soil. In he aqui e , he g oundwa e le el is es i- ma ed o each Δ by conside ing he en y o wa e by e ical low. I also allows pa ame e calib a ion by compa ing be ween he measu ed (obse a ions) and es ima ed g oundwa e le els. A comp ehensi e explana ion o he concep ual model and i s co esponding pa ame e s can be ound in Jiménez-Ma ínez e al. (2010). 3.1.1 | Model se up The cu en model equi es some inpu pa ame e s which include: geo- g aphical coo dina es, daily p ecipi a ion and ai empe a u e, ege a ion ypes and spa ial dis ibu ion, c ops and i iga ion p o isions (i.e., ime and olume), maximum oo -dep h (m), e apo anspi a ion (mm), physical soil cha ac e is ics like hickness (m), po osi y (%), sa u a ion (m 3 /m 3 ), ield capaci y (m 3 /m 3 ), wil ing poin (m 3 /m 3 ), and e ical hyd aulic conduc i - i y (m/day) and Cu e Numbe (CN, non-dimensional). The s a e a iable is wa e olume, exp essed as olume pe su ace uni (e.g., L/m 2 o m 3 / m 2 ) o equi alen heigh o wa e (e.g., mm). Fo ins ance, he wa e ol- ume in soil is he p oduc o sa u a ion and soil hickness. Su ace componen s daily i iga ion (I) and p ecipi a ion (P) a es a e he main inpu a iables. The wa e needs o each c op we e ex ac ed om Allen e al. (1998). Runo , Es, is es ima ed om he Cu e Numbe me hod (Soil Conse a ion Se ice, 1975) based on loss and p ecipi a ion a ios; he CN alues, di ec ly selec ed om he VisualBALAN da a base on CNs o mos gene al land-co e and hyd ologic soil g oup, and om exis ing spa ial in o ma ion o opog- aphy and ege a ion co e in he s udy a ea. In e cep ion, In, de i ed om he me hod o Ho on. Daily po en ial e apo anspi a ion (ETp) was compu ed by he Tho n hwai e me hod (Tho n hwai e & Holzman, 1939), sui able o da a sca ci y egions because o i s e y low da a demanding cha ac e (Yang e al., 2017). No e ha po en ial e apo anspi a ion es ima ed wi h he Tho hwai e o mula end o be sho in ege a ed a id en i onmen s (Tukima e al., 2012). Model pa ame e iza ion uses da ase s on opog aphy, land co e and soil ype. Soil-aqui e pa ame e s (po osi y, hyd aulic conduc i i y, s o age coe icien , ield capaci y, wil ing poin , soil dep h) and plan pa ame e s (heigh , in e cep ion coe icien ) comple e he sys em as main inpu s. P ocesses in he sa u a ed zone (aqui e ) we e no included by assuming ha he in il a ed wa e h ough he adose zone eached a dep h beyond he ac ion o oo s and e apo a ion, which becomes di ec in il a ion o he aqui e . Fo he model se up, he land su ace opog aphy (ele a ion and slope) was ob ained om a 30 30 m Digi al Ele a ion Model-DEM (ASTER GDEM 2.0) om NASA Ea hda a (GDEM2). The digi al ele a ion model (DEM) is used o slope and ele a ion. Land use and land physical co e age we e supplied by The Eu opean Space Cen e (ESA, CCI land co e –S2 p o o ype land co e 20 M map o A ica, 2016). Soil- ela ed in o ma ion (e.g., clay and sil con en , hyd ological soil g oup) was acqui ed om he Eu opean Soil Da a Cen e (ESDAC), Eu opean Commission –Join Resea ch Cen e (Jones e al., 2013) and he li e a u e. Homogeneous loamy sand was conside ed he main soil ype, acco ding o he Uni ed Sa es Depa men o Ag icul u e–USDA Soil Tex u al Classi ica ion (Gaul ie , 2004; LCBC-IRD, 2016). Soil-aqui e pa ame e s we e ob ained om he wo ks o Leblanc (2002), Gaul ie (2004) and Zaï i (2008). Finally, six digi al base maps we e p oduced (clima e, slope, land co e , cul i a ed c ops, aqui e s, soil a ibu es) and o e laid using GIS ools o c ea e a inal base map o echa ge calcula ions by sol ing he wa e balance equa ion in a mul icell pa e n. 3.2 |Clima e da ase s The me eo ological da a inpu s o he pe iod 2005–2014 o he model we e he daily ai empe a u e (C) and p ecipi a ion (mm) ime se ies epo ed om i e g ound s a ions in he s udy egion (Di a, Gou e, Maine, Nguigmi, Zinde ) and CHADFDM da a (sa elli e-based) o he same loca ions and pe iod. FIGURE 2 VisualBALAN concep ual model (adap ed a e Sampe e al., 2005), whe e he nomencla u e deno es: Pp ecipi a ion, I i iga ion, In in e cep ion, ETa ac ual e apo anspi a ion, Raqui e echa ge SALEHI SIAVASHANI ET AL.5o 15 3.2.1 | G ound da a G ound-based ime se ies om local me eo ological s a ions we e di ec ly compiled om he TAHMO pla o m (Figu e 1). Fo he his- o ic da a eco d (1973–2018), he s udy pe iod (2005–2014) was chosen as ha p esen ing he ewes gaps ( ≤20%) in he p ecipi a ion and ai empe a u e ime se ies. The 9-yea leng h o he da ase enabled he a iabili y in daily p ecipi a ion and ai empe a u e in he egion o be cap u ed. The ep esen a i eness o he a ailable da ase included: assess- men o missing alues, accu acy o measu emen s (inaccu acy o he amoun o p ecipi a ion), s a iona i y and homogenei y (changes o s a ions, eloca ion o s a ions, among o he s). To comple e he miss- ing ain all da a, In e se Dis ance Weigh ing was applied, based on ou ain-gauge s a ions in he icini y o he analysed s a ion (Lam, 1983). Thiessen Polygon was used o es ima e he a e age ain- all o e he a ea. 3.2.2 | Sa elli e-based ain all and ai empe a u e da a om he CHADFDM The Mul i-Sou ce Weigh ed-Ensemble P ecipi a ion (MSWEP 2.1; Beck e al., 2018), a ully global his o ic p ecipi a ion da ase de el- oped by P ince on Uni e si y, was selec ed as he sou ce o he sa elli e-based da a, based on he da a a ailabili y o he egion o in e es . MSWEP is p oduced as 3-hou ly and mon hly da a, 0.1-deg ee, by me ging a se o abou 12 sa elli e and eanalysis da ase s; 0.1-deg ee is he highes esolu ion ha is p esen ly sup- po able om hese da ase s and is ypical o o he simila sa elli e- based da ase s ha ange om abou 0.05-deg ee (e.g., CHIRPS, PERSIANN-CCS) o 0.25-deg ee (e.g., TMPA). MSWEP p o ides eli- able ain all es ima es by aking ad an age o he complemen a y s eng hs o gauge, sa elli e, and eanalysis-based da a (Beck e al., 2019). The da ase includes daily gauge co ec ions, and sys em- a ic biases (e.g., om he o og aphical enhancemen o p ecipi a ion) a e co ec ed using eadily a ailable gauge da a obse a ions (abou 25 gauges in he Lake Chad Basin) wi h enough quali y (long eco d, il e ed o spu ious alues, adjus ed o epo ing imes, and no s ep changes) and by compa ing o o he sa elli e-based p oduc s (Beck e al., 2018; Beck, an Dijk, e al., 2017; Beck, Ve gopolan, e al., 2017). Ai empe a u e da a is aken om he P ince on Global Fo cing da ase (PGF, She ield e al., 2006), which me ges sa elli e, eanalysis and g idded gauge da a o p oduce a 0.25-deg ee, 3-hou ly, global p oduc . Bo h da ase s we e downloaded om he CHADFDM. 3.3 |Sensi i i y analysis Knowledge o he model pa ame e s, and hei s a is ical a iabili y and co ela ion s uc u e, is a key aspec o quan i y he e ec o unce ain ies on echa ge es ima es acco ding o he used da a sou ce o de ine he bounda y condi ion, ha is, g ound- o sa elli e-based. The impo ance o he pa ame e unce ain ies on echa ge can be e alua ed as he objec i e unc ion by a sensi i i y analysis (Jiménez- Ma ínez e al., 2010). To his end, se e al simula ions we e un on indi idual model pa ame e s modi ied by a gi en amoun o pe u ba- ion (25% o he o iginal alue), and by es ima ing he g oundwa e echa ge o bo h sou ces o p ecipi a ion and ai empe a u e used o de ine he bounda y condi ion. 4|RESULTS 4.1 |G ound- e sus sa elli e-based da ase s Fo he selec ed 2005–2014 ime pe iod, h ee o he i e g ound- based s a ions (Figu e 1; Nguigmi, Zinde and Maine) had comple e daily ai empe a u e and p ecipi a ion eco ds, while Gou e and Di a p esen ed 3.3% and 2.5% missing p ecipi a ion alues, espec i ely. Occasionally, one s a ion can ha e highe p ecipi a ion alues han nea by s a ions o one pa icula yea when mos annual amoun s all du ing one ain all e en o wo. On a daily scale, he ain all measu ed a he i e me eo ological g ound s a ions p esen ed less a iabili y in he eco ded amoun han he sa elli e-based da a a he same loca ions (Ea h coo dina es) (Figu e 3). The maximum ain all occu s du ing he June–Oc obe pe iod, wi h he alues eco ded a gauges being he highes . Fo he 2005–2014 pe iod (Table 1), he sa elli e-based mean annual p ecipi- a ion (417 mm) was 46% highe han he g ound obse a ions (284 mm). Howe e , he ai empe a u e alues o bo h ime se ies (TAHMO and sa elli e) well ag eed, and anged be ween 15C and 40C (a e age 29C). A he mon hly scale, he sa elli e da a gene ally o e es ima ed he amoun o g ound-based p ecipi a ion (Figu e 4), excep o he wo hea y (ex eme) ain e en s eco ded a he Di a (Augus 2008: 440 mm) and Zinde (Augus 2011: 411 mm) wea he s a ions. A simi- la e ec , bu wi h gene ally highe alues o sa elli e eco ds com- pa ed o he g ound da a, ha e been epo ed by some au ho s (McCollum e al., 1999; Meh an & AghaKouchak, 2013; Noguei a e al., 2018; Young e al., 2014). Be ween he g ound- (gauged) and sa elli e-based da a sou ces, annual ain all a ies acco ding o loca ion, and we and d y condi ion dis ibu ion appea s o be di e en (Figu e 5(a),(b)). The we es and d ies yea s o he pe iod ook place in 2008 (600 mm) and 2006 (10 mm) o he g ound da a, and in 2006 (920 mm) and 2011 (200 mm) o he sa elli e da a, espec i ely. An inspec ion o he cumula i e de ia ion o p ecipi a ion om he mean (Figu e 5(c),(d)) e ealed no signi ican end on he annual scale. Fo he mon hs wi h con inuous ain all (July o Sep embe ), he g ound-based ain all e sus sa elli e-based, a e age daily da a and a e age mon hly da a, a e plo ed in Figu e 6. On a daily basis, he g ound and sa elli e ain all da a ela ion analysis (Figu e 6(a)) did no p esen any co ela ion (R 2 =0.06) be ween da ase s; a poo eg es- sion i may a ise om high sca e in da a by conside ing all he eco ded alues h oughou he hyd ologic yea . By a anging da ase s 6o 15 SALEHI SIAVASHANI ET AL. FIGURE 3 Daily p ecipi a ion (P, ed) and empe a u e (T, black) o he (a–e) i e selec ed g ound-based s a ions, and TAHMO and ( –j) CHADFMD ou pu s a he same ea h coo dina e loca ions ( ain all and empe a u e da a) SALEHI SIAVASHANI ET AL.7o 15 TABLE 1 Annual (mm/y ) and o al echa ge (R, mm) es ima es o he s udy pe iod (2005–2014) om he g ound-based s a ions (TAHMO) and CHADFDM (sa elli e-based da a) in he i iga ed and non-i iga ed a eas. Pdeno es p ecipi a ion and ETa e apo anspi a ion Da a sou ce To al P(mm) A e age P(mm/y ) To al ETa (mm) A e age ETa (mm/y ) To al RNon- i iga ed (mm) A e age RNon- i iga ed (mm/y ) To al R i iga ed (mm) A e age R I iga ed (mm/y )) G ound 2555 283.9 2244 250 90.8 10.0 302.1 33.5 Sa elli e 3754 417.1 3363 375 132.7 14.7 374.4 41.6 % o g ound gauges +46 +46 +33 +33 +46 +46 +23 +23 FIGURE 4 Mon hly a e age p ecipi a ion (mm) om bo h g ound-based s a ions and sa elli e-based da a a he same loca ions (ea h coo dina es). Red ba s indica e mon hs wi h highe TAHMO alues han CHADFDM ones. G een ba s depic he opposi e FIGURE 5 Annual p ecipi a ion (le ) and cumula i e de ia ion om he mean ( igh ) o g ound (a–b, TAHMO) and sa elli e-based (c–d, CHADFDM) da a (2005–2014 pe iod) 8o 15 SALEHI SIAVASHANI ET AL. om daily o mon hly scales (change in ime in e al), and by ocusing on mon hs wi h high p ecipi a ion (July–Augus –Sep embe ), sca e educed, while he co ela ion inc eased (R 2 =0.51). The da a agg e- ga ion clea ly smoo hs ou he di e ences be ween bo h da a sou ces. Mo eo e , i was no ed ha he pe o mance o he p oduc s is la gely scale and geog aphic loca ion dependen . Mos o he p es- en ly applied p oduc s showed compa a i ely good skills on he sea- sonal scale (R 2 > 0.90) a he han a annual scale (A iah e al., 2020). This indica es ha sa elli e da a should be imp o ed aking in o accoun he ype o p ecipi a ion (i.e., in ensi y), and o ha a mini- mum o daily esolu ion is equi ed. 4.2 |Recha ge es ima ion Two g oundwa e echa ge ypes we e assessed in he s udy a ea: (i) om p ecipi a ion; (ii) esul ing om he p ecipi a ion and i iga ion combina ion in i iga ed a eas. Calcula ed daily echa ge and eal e apo anspi a ion o i iga ed a eas and he s udy du a ion (2005– 2014) a e shown in Figu e 7 om bo h g ound-based s a ions and sa - elli e esul s. The es ima ed mean annual echa ge om p ecipi a ion equalled 10 mm/y and 14.7 mm/y o he g ound- and sa elli e- based da a, espec i ely (Table 1). Di e ences we e ma ked in ex emely d y and we yea s, such as 2006 (9 and 20 mm/y o he g ound- and sa elli e-based da a, espec i ely) and 2008 (15 and 12 mm/y o he g ound- and sa elli e-based da a, espec i ely). The compu ed mean annual aqui e echa ge in he i iga ed a eas (P+I) was 33.5 mm/y and 41.6 mm/y o he g ound- and sa elli e-based da a, espec i ely. To be e assess he echa ge p ocess, a daily echa ge es ima ion in he a ea o he non-i iga ed and i iga ed a eas, and om he wo Pand Tda ase s, was calcula ed o 2012 (Figu e 8). Yea selec ion was based on da a a ailabili y, smalles gaps and amoun o p ecipi a- ion (282 mm), simila ly o he a e age p ecipi a ion o all whole 2005–2014 pe iod. In bo h he non-i iga ed and i iga ed a eas, echa ge akes place mainly du ing he we season, be ween June and Oc obe . Recha ge esponse o ain all e en s was mos a iable (Figu e 8(a),(b)), and somewha independen o he amoun o ain all and changes o e ime, bu was s ongly condi ioned by an eceden soil mois u e condi ion, and highly dependen on p ecipi a ion dis i- bu ion o e bo h ime and space. Fo he g ound-based calcula ions, echa ge only occu ed a e one impo an ain all episode (mo e han 100 mm on 2 days, Figu e 8(a)) wi h a peak a e 3 days o con- inuous ain all. G oundwa e echa ge di e s spa ially as a esul o clima ic and physical cha ac e is ics including p ecipi a ion pa e n, soil ype, po osi y o adose zone, dep h o g oundwa e and hyd o- geology. In a id and semia id en i onmen s, g oundwa e echa ge appea s o occu du ing in ense p ecipi a ion e en s as compa ed o o he en i onmen s desc ibed as ha ing a mix o cons an - a e and episodic beha iou s and o en condi ioned by p e e en ial low (De V ies & Simme s, 2002). 4.3 |Sensi i i y analysis A se ies o simula ions we e pe o med by pe u bing (we e mul iplied by 10%, 15%, 20% and 25% es - ac o s, om 25% o +25% in magni ude) he model pa ame e s. The es included: ield capaci y, wil ing poin , soil hickness, soil po osi y, soil hyd aulic conduc i i y, cu e numbe and ini ial condi ions (ini ial soil wa e con en ) o e al- ua e he impac on aqui e echa ge es ima es. The model was un epea edly changing each es - ac o o e alua e he impac on aqui- e echa ge es ima es by ela i e change in each inpu pa ame e . Fo he loamy sand soil in he a ea, he ini ially conside ed pa am- e e s we e: ield capaci y, 0.1 (m 3 /m 3 ); wil ing poin 0.05 (m 3 /m 3 ); soil hickness, 6 m; soil o al po osi y, 0.42; soil hyd aulic conduc i i y 2  10 5 (m/s); cu e numbe (CN), 77. The bounda y condi ions ( empe - a u e, p ecipi a ion, i iga ion) we e le a he baseline alues. Then he e ec o pe u ba ions on es ima ed aqui e echa ge in ela ion o he baseline es ima ion was e alua ed (Figu e 9). The esul s showed o bo h he g ound- and sa elli e-based da a ha only he changes o unce ain ies in ield capaci y, wil ing poin , and cu e numbe led o signi ican changes in he aqui e echa ge es ima ions (Figu e 9). A 25% inc ease in wil ing poin led o a 1.6% FIGURE 6 Rain all da a, g ound (TAHMO) e sus sa elli e (CHADFDM): (a) a e age daily da a; (b) a e age mon hly da a o mon hs wi h con inuous ain all (July o Sep embe ) SALEHI SIAVASHANI ET AL.9o 15