climate Article Site Selection for a Network of Weather Stations Using AHP and Near Analysis in a GIS Environment in Amazonas, NW Peru Nilton B. Rojas Briceño 1,* , Rolando Salas López 1 , Jhonsy O. Silva López 1 , Manuel Oliva-Cruz 1, Darwin Gómez Fernández 1 , Renzo E. Terrones Murga 1 , Daniel Iliquín Trigoso 1 , Miguel Barrena Gurbillón 1 and Elgar Barboza 1,2 1 Instituto de Investigación para el Desarrollo Sustentable de Ceja de Selva, Universidad Nacional Toribio Rodríguez de Mendoza de Amazonas, Chachapoyas 01001, Peru; rsalas@indes-ces.edu.pe (R.S.L.); jhonsy.silva@untrm.edu.pe (J.O.S.L.); soliva@indes-ces.edu.pe (M.O.-C.); darwin.gomez@untrm.edu.pe (D.G.F.); renzo.terrones@untrm.edu.pe (R.E.T.M.); diliquin@indes-ces.edu.pe (D.I.T.); miguel.barrena@untrm.edu.pe (M.B.G.); ebarboza@indes-ces.edu.pe (E.B.) 2 Dirección de Desarrollo Tecnológico Agrario, Instituto Nacional de Innovación Agraria (INIA), Av. La Molina 1981, Lima 15024, Peru * Correspondence: nrojas@indes-ces.edu.pe  Abstract: Meteorological observations play a major role in land management; thus, it is vital to  properly plan the monitoring network of weather stations (WS). This study, therefore, selected ‘highly Citation: Rojas Briceño, N.B.; Salas suitable’ sites with the objective of replanning the WS network in Amazonas, NW Peru. A set of López, R.; Silva López, J.O.; 11 selection criteria for WS sites were identified and mapped in a Geographic Information System, Oliva-Cruz, M.; Gómez Fernández, as well as their importance weights were determined using Analytic Hierarchy Process and experts. D.; Terrones Murga, R.E.; Iliquín A map of the suitability of the territory for WS sites was constructed by weighted superimposition of Trigoso, D.; Barrena Gurbillón, M.; the criteria maps. On this map, the suitability status of the 20 existing WS sites was then assessed Barboza, E. Site Selection for a Network of Weather Stations Using and, if necessary, relocated. New ‘highly suitable’ sites were determined by the Near Analysis AHP and Near Analysis in a GIS method using existing WS (some relocated). The territory suitability map for WS showed that 0.3% Environment in Amazonas, NW Peru. (108.55 km2) of Amazonas has ‘highly suitable’ characteristics to establish WS. This ‘highly suitable’ Climate 2021, 9, 169. https://doi.org/ territory corresponds to 26,683 polygons (of ≥30 × 30 m each), from which 100 polygons were 10.3390/cli9120169 selected in 11 possible distributions of new WS networks in Amazonas, with different number and distance of new WS in each distribution. The implementation of this methodology will be a useful Academic Editor: support tool for WS network planning. Nektarios Kourgialas Keywords: analytical hierarchy process; meteorological station; near analysis; suitability mapping Received: 31 October 2021 Accepted: 25 November 2021 Published: 28 November 2021 1. Introduction Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in Meteorological observations are used for real-time weather analysis, severe weather published maps and institutional affil- forecasts and warnings, for weather-sensitive local operations (e.g., airfield flights or con- iations. struction work at land and sea facilities), for hydrology and agricultural meteorology, and for meteorological and climatological research purposes [1–3]. However, previous studies around the world [4–7], including in Peru [8–10], warn that there is (i) lack of weather stations (WS) in certain important areas, (ii) a non-uniform spatial distribution of existing WS, and (iii) variable precision of measurements because current WS sites are Copyright: © 2021 by the authors. often inadequate. Therefore, planning an adequate network of WS constitutes the basic Licensee MDPI, Basel, Switzerland. This article is an open access article and necessary infrastructure for land management [7]. Adequate distribution of WS in- distributed under the terms and creases the effectiveness of observations and provides more accurate analysis results [2,11]. conditions of the Creative Commons In addition, it is essential to restructure the WS network and increase the number of WS in Attribution (CC BY) license (https:// countries with a high frequency of natural disasters [2], such as Peru. creativecommons.org/licenses/by/ Peru is located in the intertropical zone of South America, on the Pacific Coast, and, 4.0/). unlike other equatorial countries, it does not have an exclusively tropical climate [12]. Climate 2021, 9, 169. https://doi.org/10.3390/cli9120169 https://www.mdpi.com/journal/climate Climate 2021, 9, 169 2 of 15 In 2020, the Servicio Nacional de Meteorología e Hidrología del Perú (SENAMHI) defined 38 climate types, 11 more than the previous map, and among other factors, the updated map included a greater number of WS [13]. Additionally, due to its geographical location, topography, land cover, and high, non-uniform rainfall, Peru suffers from a variety of natural disasters, including earthquakes, floods, landslides, frost, and forest fires [14,15]. In this regard, in the Amazonas region (NW Peru), the Universidad Nacional Toribio Rodríguez de Mendoza de Amazonas (UNTRM) operates a network of 12 automatic WS [16]. This network, as in other meteorological networks around the world [2,3], aims to support not only research needs (climate monitoring and analysis, weather and natural disaster forecasting capabilities, etc.) but also the needs of various communities in the production sector (agriculture, construction, leisure activities and tourism, etc.). Currently, UNTRM aims to expand its number of WS in suitable locations. Determining the most suitable locations is, therefore, of utmost importance for the sustainability and success of the WS network [2,17]. The selection of locations for a WS involves various spatial analyses, including the distances between different land use zones, slopes, roads, populations, and proximity to natural hazard boundaries, such as landslide areas [2,3]. In this sense, geographic information systems (GIS), integrated with remote sensing (RS) data and multi-criteria analysis (MCA) techniques, is an important decision- support tool that is able to operate and analyze a wide range of spatial data and criteria [18]. GIS–RS–MCA has been used in the selection of suitable sites for hydro/agro-meteorological stations in Turkey [2], Greece [7], the Philippines [19,20], and Colombia [21]. Two important questions emerge from these studies. First, in Turkey [2], after MCA, a near analysis (NA) was proposed and implemented between suitable sites determined by MCA and existing WS, such that highly suitable sites for WS were selected. Second, these studies did not consider the relative importance of the criteria in theie MCAs, the incorporation of which is important in improving MCA [22]. For this, the analytical hierarchy process (AHP) is the most widely used MCA technique, and it consists of weighing the importance of each criterion using expert opinions [23]. In Peru, the “Protocol for the installation and operation of meteorological and hydro- logical stations” (Section 7.2) of SENAMHI [24], establishes several criteria and guidelines for the selection of a WS site, but it does not provide a methodological framework for WS network planning. Accordingly, this study aims to select highly suitable sites for a WS network by integrating AHP and NA, using the Amazonas Department (NW Peru) as the study area. It is expected that existing WS locations are not highly suitable and need to be relocated. In summary, (i) WS site-selection criteria were identified and mapped in a GIS, and the weighting of their importance was determined by AHP; (ii) a map of the suitability of the territory for WS sites was constructed; (iii) the suitability of the conditions of existing WS sites were assessed and, if necessary, the sites were relocated; (iv) the most suitable new sites for monitoring networks with different number of WS were determined by the NA method, using existing WS. Finally, an integrated methodological framework of AHP and NA in a GIS environment is proposed as a useful support tool for WS network planning, which can be replicable in other regions with the necessary complements. 2. Materials and Methods 2.1. Study Area In the northeastern Peruvian Andes, the Amazonas Department covers approximately 42,050.37 km2 of rugged territory, covered mainly by the Amazon rainforest, along an elevational gradient from 120 m a.s.l., in the north, to 4900 m a.s.l., in the south (Figure 1, 3◦0′–7◦2′ S and 77◦0′–78◦42′ W) [25]. It has contrasting climates (“warm and humid”, “dry warm”, and “warm and slightly humid temperate”), ranging from a maximum temperature of 40 ◦C, in the lowland forest of the north, to a minimum temperature of 2 ◦C, in the mountain ranges at the southern boundary; some areas have a water deficit of 924 mm/year and others have a surplus of up to 3000 mm/year [26]. As part of their high biophysical diversity, four ecosystems can be distinguished [27]: (i) lowland Climate 2021, 9, 169 3 of 15 1, 3°0′–7°2′ S and 77°0′–78°42′ W) [25]. It has contrasting climates (“warm and humid”, “dry warm”, and “warm and slightly humid temperate”), ranging from a maximum tem- perature of 40 °C, in the lowland forest of the north, to a minimum temperature of 2 °C, Climate 2021, 9, 169 in the mountain ranges at the southern boundary; some areas have a water deficit o3fo 9f2145 mm/year and others have a surplus of up to 3000 mm/year [26]. As part of their high bio- physical diversity, four ecosystems can be distinguished [27]: (i) lowland forest, (ii) high ffoorreesstt ,o(ri iy)uhniggha,f (oiirie)s At onrdeyaunn fgoar,e(sitisi) aAnndd geraanssfloanredsst,s aanndd (igvr)a tsrsolapnicdasl ,darnyd fo(irve)stt.r Aopmicaazlodnrays fiso rcehsat.raAcmteraizzoenda bsyi sitcsh aagrraicctuelrtiuzreadl bacytiivtsitayg, rwichuilcthu roaclcaucptiiveist y2,4w.9h%ic ohf othcecu teprireisto2r4y.9 a%ndo fgtehne- teerrartietos r5y1.a2n2d%g oefn tehrea tdeesp5a1r.2tm2%enotaf lt hgerodsesp daormtmeesntitca lpgrorodsuscdt o[2m7e].s Cticurprreondtulyc,t t[h2e7r]e. Caruer 2re0n WtlyS, tinh eAremarzeon2a0sW, vSariyninAg mbya ztyopnea sb,etvwareyeinn agubtoymtyatpice abnedtw coenenveanutitonmaal,t aicdmanidnisctoenrevde nbtyio UnNal-, aTdRmMi n(1is2t)e arendd bSyENUNAMTRHMI ((81)2 ()FaingudrSeE 1N). AInM thHeI c(u8)rr(eFnigt unreetw1)o.rIkn tthheerec uisr rneon tuneiftowromr kspthaetirael idsisntoribuuntifoonr mof stphae tWialSd, bisetcraibuustei oenacohf itnhsetitWutSio, nb eicnasutasleleeda WchSi nfosrti tiutst ipounrpinosstea,l laetd dWiffeSrfeonrt ittismpeus,r wpoitsheo, uat dcoifnfesirdenertitnimg ae sd,ewpiatrhtomuetnctoanl dsiidsterriibnugtiaond enpeatwrtmorekn ptalalnd aisntdri bthueti onlyn ectrwiteorrika pulsaend awnedreth feoro cnolynvcerintieerniaceu. sTehduws,e cruerfroernctloyn tvheenriee anrce .elTehvuasti,ocnu rraenngtelys thaetr earaer eweellelv raetpioren- rsaentgeeds, tbhuatt oatrheewr reallnrgeepsr aerse ntoetd, ,abnudt oofthener twraon gWeSssa frreonmo tb,oatnhd inosftietnuttiwonosW arSes tofroo mclobsoe ttho ienascthit uothioenrs. are too close to each other. Fiigure 1.. Currentt Weattherr Sttattiionss (WS)) iin tthe wattersheds tto whiich tthe Amazonas Departtmentt bellongs (a,,b),, iin NW Peru,, SSoouutthh Ameerriiccaa ((cc)).. 2.2. Methodological Design Figure 2 shows the methodological process developed for the selection of WS sites by integrating the analytic hierarchy process (AHP) and near analysis (NA). This study is the first integration of both methodologies. Climate 2021, 9, 169 4 of 15 2.2. Methodological Design Climate 2021, 9, 169 Figure 2 shows the methodological process developed for the selection of WS4 osfi1te5s by integrating the analytic hierarchy process (AHP) and near analysis (NA). This study is the first integration of both methodologies. Near Analysis (NA) Analytic Hierarchy Process (AHP) Criteria Sub Criteria Alternative First hierarchy Second hierarchy Third hierarchy Elevation Terrain slope Terrain shadow Biophysical Land use/Land cover Reclassification based Goal Subgoal Distance to water sources on suitability: Selected sites for Land suitability for Distance to geological faultsProximity to (4) Highly suitableweather stations installed WS weather stations Landslide susceptibility (3) Moderately suitable(WS) (WS) (2) Marginally suitable Distance to roads (1) Not suitable Distance to populations Administrative Distance to the host institution Protected natural areas Then the maps of subcriteria, criteria and subgoal will have this classification FFigiguurere2 2. . Meetthodollogiical process with hiieerraarrcchhyy ooff oobbjejecctitviveess aanndd crcirtietreirai afofro trhteh seeslecleticotnio onf osfitesist efosrf omremteeotreoolorogliocgali csatal - sttaiotinosn isni nAAmmazaoznoansa, sN, NWW PePreur.u . 22.3.3. .C Crrititeerriaiaa annddC CrritieterriaiaT Thhrreesshhooldldssf oforrS Seleelcetcitningga aS SuuitiatabbleleS Sitiete TThheeA AHHPPs tsrtuructcutureressp prorobblelemmssi nintotol elveevleslso fofs usbu-bp-rporbolbelmems/so/obbjejeccttiivveess aanndd ccrritieterriaia, , wwhheereree eaacchhl elevveel lc caannb beea annaalylyzzeeddi ninddeeppeennddeennttlylya annddi sism moorreee eaassiliylyu unnddeerrssttoooodd[ [2288––3300]].. AAh hieireararcrhchyyw wasasc oconnstsrtuructcetded, ,c oconnsissitsitninggo offt hthees tsutuddyys usubb-g-gooaal l( l(alanndds suuitiatabbiliiltiytyf oforrW WSS),), twtwooc rcirtietreiraia( b(iboipophhyysisciaclala nandda dadmmininisitsrtaratitvivee),),e leelevveenns usubbc rcirtietreiraiaa nanddf ofouurra latletrenrnataitviveess (F(Figiguurere2 2).).T Thheec crirtieteriraiaw wereeree setsatbalbislihsehdedc ocnosnisdiedreinrigngte tcehcnhincaicladl odcoucmumenetns tfsr ofrmomW oWrlodrlMd eMtee-- otreoolorogliocgalicOarl gOarngizaantiizoanti(oWn M(WOM) [O1,)1 7[1,3,117] ,a3n1d] aEnndv iEronnvmiroenntmalePnrtoalt ePcrtiootnecAtigoenn Acyge(EnPcAy )(E[3P2A],) th[3e2P],e trhuev Piaenrunvaitaionn naal tpiorontaol cporlootfoScoElN oAf MSEHNIA[2M4]HaIn [d24s]t uadnide sstcuodniceesr ncoinngcetrhneinsegl ethctei osneloefc- htyiodnro o/f ahgyrdo-rmo/eatgeroor-omloegteicoarlosltoagtiiocanl ssittaetsio[2n, 7s,i1te9s,2 [12],.7T,1o9,d2e1v].e Tloop dtehveetlhoiprd thhei etrhairrcdh hyi,esrhaorwchny, inshFoiwgunr ein2 F(itghuerael t2e r(nthaeti avletse)r,ntahteivseusb),- ctrhiete sruiab-wcreirtersicao wreedreo nscaorsecda loeno faf socuarlel eovfe flosuorf lleavnedls soufi tlabnidli tsyuiftoarbimliteyt eforo mloegtiecoarlosltoagtiicoanl stiatetiso(nT asibteles 1(T):ab(4le) ‘1H):i (g4h) l‘yHsiguhitlayb sleu’i,ta(3b)le‘’M, (o3d) ‘eMraoted- Seuriatateb lSeu’,i(t2a)b‘lMe’,a (r2g)i n‘Mallayrgsuinitaallbyle s’uaintadb(l1e)’ ‘aNnodt (s1u)i t‘aNbolet ’s.uTihtaebseles’.u Tithabesilei tsyulietvaeblisliatyre ledveerilvs eadre frdoemrivtehde ffirovme stuhiet afibvieli tsyuilteavbeilsityo fletvhelsF oAfO th’se “FAOFr’sa m“Ae wFroarmk efowroLrka nfodr ELvaanldu aEtvioanlu”a[t3io3n];” w[3h3e]r;e w, ahserine, oatsh ienr ostuhedri esstu[2d2ie,3s4 [,2352],3, 4w,3i5th], awmitho rae mreosrtrei crteisvteriactpivper oaapcphr,otahcehl,a tshtet wlaost ltewv-o ellesv(eClsu (rCreunrtrleynNtlyo tNSouti Staubitleabalne danPde rPmeramneantelnytlNy oNtoStu Situaibtaleb)lew) wererec ocmombibniendedt otof oformrm( 1(1)) ‘N‘Noot ts usuitiatabblele’.’.T Thhee‘ H‘Higighhlylys suuitiatabblele’ ’l elevveell( i(ninccluluddiningg‘ M‘MooddeerraatteelylyS Suuititaabbllee’’f foorrs soomees suubb-- ccrritieterriaia))a aimimsst otom meeettt htheem moosstts strtirninggeennttr ereqquuirieremmeenntstsf oforrC Clalasss1 1s istietess( a(annddt htheererefoforeret hthee ooththeerrc cllaasssseess)) essttablliished by WMO [[11]] aanndd ththee rereqquuirieremmeenntst sfofor rsysnyonpotpicti,c c,licmlimataotlolgoigcai-l, canl,da nadgraogmroemteeotreoolroogliocgailc astlasttiaotnio snitseiste sesetsatbalbislihsehde dinin tthhee Peruviian national pprroottooccoollo of f SSEENNAAMMHHII[ 2[244].]. The site should be on flat terrain, with no steep slopes nearby, and should not be Table 1. Siunba crhitoelrlioa wsc;ooritnhge frowr isseele,ctthinege wqeuaitphmere sntattiwonil lsirtesc einiv Aemcaoznosnidase, rNabWle Pderaui.l y shading and WS observations will have only locally significant peculiarities [1,17]. Reflective surfaces or Highly Moderately Marginally Not Suitable Adapted Criteria/Sub-Criteria artifiScuiiatlabhleea (t4s 1o) urces (eS.gu.i,tabbuliel d(3in 1)g s, concreteSsuuitrafbalcee s(2o 1r) car parks), a(n1 d1) bodies offwroamte r or moisture (e.g., large riBvieorpsh, ypsoicnadl s, lakes or irrigated areas), distort measurements of Elevation (120–4900) te2m20p0e–r4a9t0u0r me, ah.su.lm. idi1t0y9,0r–a2d2i0a0t imo na,.sw.l. ind, an1d20o–t1h0e9r0 mva ar.isa.lb. les [1]. For –b oth cases ab[7o]v e, Terrain slope the max≤im5%u m distances re5c–o1m5%m ended by WMO15[–12]5a%n d SENAMHI [≥2245]%a re >100 m[1a7]n d Terrain hillshade >30 m, 0re–s4p he ctively. Severa5l–7a rhe as of Amazonas 8a–r1e0s hu sceptible to la1n1d–1s3li hd es [15], co[2n] se- Land Use/Land Cover–LU/LC q2 uently, a4r0e as of high susc2e0p, t3i0b,i l6i0ty were considere>d1u00n suitable for W0,S 5s0i,t 8e0s,[ 920] . In add[1it9i]o n, a buffer zone of 500 m from the geological faults was established [2]. Agriculture is the main economic activity in Amazonas [37]; therefore, the agricultural zone was considered very suitable for WS sites. Climate 2021, 9, 169 5 of 15 Table 1. Sub criteria scoring for selecting weather station sites in Amazonas, NW Peru. Criteria/Sub-Criteria Highly Moderately Marginally Not Suitable AdaptedSuitable (4 1) Suitable (3 1) Suitable (2 1) (1 1) from Biophysical Elevation (120–4900) 2200–4900 m a.s.l. 1090–2200 m a.s.l. 120–1090 m a.s.l. – [7] Terrain slope ≤5% 5–15% 15–25% ≥25% [17] Terrain hillshade 0–4 h 5–7 h 8–10 h 11–13 h [2] Land Use/Land Cover–LU/LC 2 40 20, 30, 60 >100 0, 50, 80, 90 [19] Main ≥1 km 0.5–1 km 0.25–0.5 km ≤0.25 [1,2,24] Distance to water bodies Secondary ≥0.5 km 0.25–5 km 0.1–0.25 km ≤0.1 [1,2,24] Distance to geological faults ≥1.5 km 1–1.5 km 0.5–1 km ≤0.5 km [2] Landslide susceptibility Very low; Low Medium High Very high [2] Administrative National 0.3–0.7 km 0.7–1.2 km 1.2–2.2 km ≤0.3/≥2.2 km [1,2,24] Distance to roads Departmental 0.2–0.6 km 0.6–1.1 km 1.1–2.1 km ≤0.2/≥2.1 km [1,2,24] Local 0.1–0.5 km 0.5–1 km 1–2 km ≤0.1/≥2 km [1,2,24] Distance to populations Urban areas 0.2–0.6 km 0.6–1.1 km 1.1–2.1 km ≤0.2/≥2.1 km Villages 0.1–0.5 km 0.5–1 km 1–2 km ≤0.1/≥2 km Distance to the host institution ≤10 km 10–50 km 50–100 km ≥100 km [19] Protected natural areas—PNA Inside Outside – – [24] 1 Pixel value of the map reclassified according to the four levels of land suitability. 2 CGLS-LC100 [36]: 0—NoData, 20—Shrubs, 30—Herbaceous vegetation, 40—Cropland, 50—Urban/built up, 60—Bare/sparse vegetation, 80—water bodies, 90—Herbaceous wetland, and >100–all the forests. In fact, roads and vehicular traffic do affect the measurements of WS but accessibility is necessary for WS’ sustained operation and they must be taken into account [2,7]. Simi- larly, although urban constructions affect measurements, proximity to them is important to ensure power supply and surveillance of the instruments from theft and/or vandal- ism [1,17]. In addition, the worldwide trend towards urbanization should be considered by avoiding areas planned for urban sprawl (or with a buffer to the current urban area) [2,24]. The lack of host institutions where WS can be installed is a real barrier to achieving an optimal number of WS in most developing countries [19]; ergo, they should be identified and considered. Moreover, Amazonas is the third department in Peru with the highest number of Protected Natural Areas (PNA) [38], and given the ecological importance of these territories, they are a priority for having WS. 2.4. Mapping of Sub-Criteria for Selecting a Suitable Site Elevation, terrain slope, and terrain shadow were derived from the ASTER Global Digital Elevation Model V3 (spatial resolution: 30 m) downloaded from the Japan Space Systems platform [39]. A terrain-shadow map was generated for each hour between 1200 and 0000 UTC [2] (0700 and 1900 UTC–5, Peru). For this, the sun’s elevation and azimuth, averaged for each hour in 2020 and obtained from SunEarthTools [40] was used. In each map, pixels fully shaded by another pixel were given a value of 0 and all other pixels were given integer values between 1 and 255. All values greater than 1 were re-classified to 0, and pixels with 0 or 1 and the 13 maps were summed to count the hours of shade per day for each pixel. The land-use/land-cover (LU/LC) base map was obtained from the Copernicus Global Land Service-Land Cover (CGLS-LC100)-Collection 3-2019 of 100-m spatial resolution [36]. On this map, LU polygons (urban, agricultural and livestock areas) were updated from the National Ecosystems Map of Peru [41,42], the National Map of Agricultural Surface [43], the Amazon Economic Ecological Zoning (ZEE-A) [27], and the province of Rodriguez de Mendoza [44]. Water bodies (rivers and lakes) were obtained from the ZEE-A [27]. Lakes and rivers of order ≥3 (Strahler method [45]) are listed as principal in Table 1 and the remaining ones were considered secondary and without significant influence on WS observations. The susceptibility maps for landslides [15] and geological faults [46] were obtained from the Instituto Geológico Minero y Metalúrgico in Peru. Climate 2021, 9, 169 6 of 15 The road network (national, departmental and local categories) was obtained from the portal of the Ministerio de Transportes y Comunicaciones (MTC) [47]. The urban polygons were extracted from the final LU/LC map and the population centres (points) were obtained from the (MINEDU) [48]. The host institutions were the headquarters, experimental stations and branches of the UNTRM (geo-referenced with GPS receiver) and the public educational institutions of higher technical and university level (obtained from MINEDU [49]). The PNA were obtained from the Servicio Nacional de Áreas Naturales Protegidas por el Estado (SERNANP) [38]. The distances to water bodies, roads, geological faults, possible administration centres and population centres from anywhere in the Amazonas Department were mapped using the Euclidean distance tool. In sum, 11 thematic maps were prepared in raster models, one map for each sub-criterion. These maps were standardized in the WGS 1984 UTM 18 South coordinate system and spatial boundaries restricted to Amazonas and a spatial resolution of 30 m. This resolution was based on the maximum dimensions recommended for meteorological stations by the WMO (25 × 25 m) [1] and the SENAMHI national protocol (15 × 25 m) [24]. Then, the thematic maps for each sub-criterion were reclassified according to the thresholds established in Table 1, assigning scores (between 1 and 4) to each pixel. 2.5. Determination of Importance Weights of Criteria and Sub-Criteria The development of the second and first hierarchy in Figure 2 involves constructing pairwise comparison matrices (PCM), which compare one criterion with respect to the others (pairwise) and establishes the degree of importance between them [50]. The applied AHP is a variation of the traditional Saaty method, because the paired comparison was not applied to the third hierarchy (alternatives) [28]. The comparison was based on Saaty’s nine-level scale (Table 2) [29], and each member of a group of experts assigned a value judgement, from the least important to the most important criterion, based on their experience. The questionnaire with the PCM was sent by email to professionals from SENAMHI, the Autoridad Nacional del Agua (ANA), meteorological instrument companies and meteorological professors/researchers. Each expert completed two PCMs at the sub-criterion level and one PCM at the criterion level (hierarchical groups in Figure 2). The experts’ PCMs were processed (for examples of matrix processing of PCM, see [34,51]) and weighted importance was obtained for each sub-criterion and criterion. The subjective preferences between experts can lead to inconsistencies in the importance weights, because their matrices do not comply with the two simultaneous properties of consistency, which are the transitivity of the preferences and the proportionality of the preferences [52]. Therefore, the consistency ratio (CR) of each PCM was calculated to compare with an acceptable inconsistency (CR < 0.1) [28]. CR was calculated by dividing the PCM’s consistency index (CI) with a random consistency index (RI) [30]. This RI is defined according to the number of criteria (n) (Table 3) [53] and CI depends on the largest or principal eigenvalue of the matrix (λmax) and n (Equation (1)). CI = (λmax − n)/(n − 1), (1) Table 2. Scale established for the allocation of value judgments between two criteria in peer comparison matrices (PCM). 1/9 1/8 1/7 1/6 1/5 1/4 1/3 1/2 1 2 3 4 5 6 7 8 9 Extreme Strong Moderate Moderate Strong Extreme Equal less important more important Table 3. Random index (RI) to determine the consistency ratio (CR) of peer comparison matrices (PCM). n 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 RI 0 0 0.525 0.882 1.115 1.252 1.341 1.404 1.452 1.484 1.513 1.535 1.555 1.570 1.583 1.595 Climate 2021, 9, 169 7 of 15 Climate 2021, 9, 169 CI = (λmax − n)/(n − 1), 7 of 15(1) 2.6. Sub-Model Generation and Suitability Modelling 2.6. SubT-hMe ofdinelaGl deneevrealtoiopnmaenndt Soufi ttahbei lsiteycoMnodd ealnlidng first hierarchy consisted of integrating the reclTahsseiffiienda lthdeemvealtoicp mmaepnst (obfatsheed soenc oTnadblaen 1d), fiacrcstorhdieinragr tcoh hyiecoranrscihsitceadl ogfroinutpe g(Friagtuinrge 2th),e by rewclaeisgshifiteedd ltihneemara toivcemrlaapy s(E(bqausaetdioonn (2T)a)b [l2e21,3),4a,3c5c]o. rTdhineg rteosuhliteirnagr cshuiictaalbgilriotyu p(G(FRiIg-Durrees2u),lt) bydwepeeignhdteedd olinn etahreo rveecrllaasysi(fEieqdu amtiaopn p(2i)x)e[l2 s2c,3o4re,3 5(G].RTIhDei)r easnudlt itnhge ssuuibta-bcriliitteyri(oGnR iIm-Dproerstualnt)ce dewpeeingdhet d(WonEItGheHrTeic)l. aTshsiefi iendtemgraaptiopnix oefl ssucobr-ecr(iGteRriIaD gi)enanerdattehde tshueb b-cioriptehryiosincaiml apnodr taadnmcein- wiesitgrhatti(vWe sEuIGitaHbTilii)t.yT shueb-imntoegdrealsti, oanndo fthsue bin-cterigtrearitaiogne onfe trhaetesed stuhbe-bmioopdheylss giceanlearnadtedad tmhei nte-r- istrriatotirvye ssuuitiatabbiliiltiyty msoudbe-ml foodr eWlsS, .a Wndattehre biondtieegsr aantidon urobfatnh easreeassu (bp-omlyogdoenlss)g weneerrea rteesdtrtihcteed tertori ttohrey ssuuiittaabbiilliittyy mmaopdse,l afnord WthSe. aWreaate (rinb okdmie2s) awnads ucrobuannteadre wasit(hpionl ythgoe n‘Ns)owt esrueitraebstleri’c lteevdel. toTthhies spuriotacbeislsit ywmasa ppse,rfaonrdmtehde warietah (tihnek mAr2c)GwISas 1c0o.5u nWteedigwhittehdi nOtvheer‘lNayo t(Sspuaittaiabll eA’ lneavleyls.t) Thtoisopl.r ocess was performed with the ArcGIS 10.5 Weighted Overlay (Spatial Analyst) tool. GRIDresult = Σ [(GRIDi) (WEIGHTi)], (2) GRIDresult = Σ [(GRIDi) (WEIGHTi)], (2) 2.72..7N. NeaeraAr nAanlyalsyissi(sN (NA)At)o tSo eSleecletcWt WS SS iSteistes ThTeh2e0 2e0x eixstiisntigngW WS S(F (iFgiugruer1e) 1w) wereeroe voevrelarliadidw withithth tehWe WS Sla lnadn-dsu-siutaitbaibliitlyitym modoedletlo tod ed-e- tertemrminienteh tehceu crurerrnetncto cnodnidtiiotnioonf othf tehlea nladnodn own hwichhicthh ethyeayr earloe claotceadte. dIn. Iand additdioitnio, nth, eth‘eh i‘ghhiglhyly susiutaitbalbel’et’e trerrarianin(a (caccocrodridnigngto toth tehed edceicsiisoinonru rluelein inF iFgiugruere3 a3)ac) lcolsoessetstto toe aecahchW WS Sw wasads edteetre-r- mminiendedin ino rodredretro troe lroecloactaetteh tehceu crurerrnetnlytlym mispislpalcaecdedW WS.ST. hTihsisst estpepw wasabs absaesdedo nonth tehne enaerar anaanlaylsyissis(N (NAA) )m metehthoodd. . NNAA ccaallccuullaatteedd tthhee ddiissttaanncceess bbeetwtweeeenn twtwoo ininppuut stpsaptaiatila flefaetautruerse (sin- (inppuut tffeeaattuurree == 2200 eexxiissttiinngg WWSS;; nneeaarr ffeeaattuurree == oonnlyly‘ h‘higighhlylys usuitiatabblele’ ’p poloylgyognons sf rformomth tehWe WS S lalnadn-ds-usiutaitbaibliitlyitym modoedle)l()F (iFgiugruere3 b3,bc),c[)2 [,52,45]4. ]T. hTihsips rporcoecsessws wasaps epreforfromrmededw iwthithth tehAe rAcGrcIGSIS 101.50.N5 Neaera(rA (Ananlaylsyissi)st)o toolo. l. Installed In Highly Near feature Near feature WS suitable Relocate? No In Moderately Relocate to the nearest Highly If d < 5 km Yes suitable suitable territory? If d ≥ 5 km In Marginally suitable No, then Near feature 45º Δx distance (NFD) In Not Relocate to the nearest suitable Moderately suitable territory? Yes Input feature Input feature (a) (b) (c) FigFuigruer3e. 3(.a ()aD) Deceicsiisoinonr urulelef ofor rt htheer ereloloccaatitoionno off iinnssttaalllleedd wweeaatthheerr ssttaattiioonnss ((WSS)).. TThhee mmeetthhoodd oof fnneeaar raannaalylysissi s(N(NAA) [)5[45]4: ](:b) (b)ththe eddisitsatannccee frfroomm ininppuutt ffeeaattuurree ttoo nneeaarr ffeeaattuurree iiss ccaallccuullaatteedd bbaasseeddo onnt thheeP Pyyththaaggoorereaannt htheoeroermema nadndth tehde idffieffreernecnecseisn itnh ethireir cocoordoirndainteast,easn, adn(dc) (ca)n aenx eaxmapmleploef ocfa clcaulclautliantginnge naerafre afetuatruerdei sdtiasntacnecfeo rfocri rcciurclaurlasry smybmoblsolws hwehnedne dfienfeindeads assy smymboblodli damiaemteert.er. ThTehne,nt,h tehes esleelcetcitoionno off‘ h‘higighhlyly ssuuiittaabbllee’’ ssiitteess ffoorr iinnssttaalllliinngg nneeww WSS wwaass aalslsoo bbaasesded on onthteh Ne AN Amemtheotdh,o bdu,t bbuyt sbwyapspwinapg pthineg futnhcetifounn octfi tohne otwf oth sepatwtiaol isnppautti afleaintupruest (fienaptuutr efsea- (intupruet =fe oantulyre ‘h=igohnllyy s‘huiigtahbllye’s puoitlaybgloen’ sp oofl ytghoe nWs So flatnhde WsuSitalbainlidtys umiotadbeill;i tnyeamr ofdeaetlu; rnee =a r20 feraetulorcea=ted20 erxeilsoticnagte Wd eSx).i sFtrionmg WthSe) .fiFrsrto miterthateiofinr,s tthiete fruarttiohnes,tt h‘heigfuhrlyth seusitta‘hbilge’h plyolsyugitoanb lwe’as poclhyogsoennw asa sthceh foisrestn naeswth seitfier (sWt nSe nwusmitbee(rW 2S1)n. uInm tbheer s2e1c)o.nIdn ittheerasteicoonn, dthiete sreactoionnd, ntheews esictoen (Wd S nenwumsibteer( W22S) nwuams bdeerte2r2m)iwneads dase ttehrem fiunrethdeasst ‘thiegfhulyrt hsueistta‘bhlieg’ hploylysugoitna binle ’reploaltyiogno ntoi nthree -21 laWtioSn. Stoucthces2si1vWe itSe.rSauticocness swiveerei tpeerraftoiormnsedw, ecroenpsiedreforirnmg ende,wco snitseisd aetr ienagchn eitwerastiitoens ,a utnetaicl hthe itefruarttihoens,t udnisttilanthce fwuartsh lests dthisatna n9c kemw [a1s9l]e. ss than 9 km [19]. 3.3R. eRseuslutslts 3.13..1I.m Impoprotartnacnecoef oCf rCitreirteiariaan adnSd uSbu-bC-rCitreirtiearia ThTehwe ewigehigtehdtesdc osrceosrefos rfoera cehacshu bs-ucbri-tcerriitoenrio(Tna b(Tleab4l)ew 4e) rwe cearelc ucalalcteudlaftreodm frtohme mtheea nmoefan thoef etvhael euvaatilounastioonf sfi ovfe f(i5v)e m(5e) tmeoertoeloorgoilcoagliceaxlp eexrptse.rTtsh. iTshgisro gurpouopf oexf pexeprtesrtths atht arte srpesopnodnedded anadndap apprporvoevdedw withithC RCR< <0 .01.1w wasams madaedue pupo foafn anex epxepretrftr ofrmomSE SNENAMAMHHI, Io, noenfer ofrmomA NAANA one from Peru Davis Instruments E.I.R.L., and two professors/researchers in meteorology. The sub-criteria terrain slope (22.8%) and distance to water sources (21.4%), followed by terrain hill shade (16.2%) and land use/land cover (15.3%), obtained the highest weight- ing with respect to their biophysical criteria group. In the administrative criteria group, distance to roads (36.2%) and distance to populations (32.8%) were the most important Climate 2021, 9, 169 8 of 15 sub-criteria. Regarding the group of criteria, biophysical (68.3%) was more important than administrative (31.8%). Table 4. Importance weights (%) for the second and third hierarchies of criteria for selecting sites for weather stations in Amazonas, NE Peru. Criteria Weight (%) Rank Sub-Criteria Weight (%) Rank Standardized StandardizedWeight (%) Rank elevation 9.1 5 6.2 7 terrain slope 22.8 1 15.6 1 terrain hill shade 16.2 3 11.1 4 Biophysical 68.3 1 land use/land cover–LU/LC 15.3 4 10.4 5 distance to water sources 21.4 2 14.6 2 distance to geological faults 8.8 6 6.0 9 landslide susceptibility 6.4 7 4.4 10 distance to roads 36.2 1 11.5 3 distance to populations 32.8 2 10.4 6 Administrative 31.8 2 distance to host institution 19.1 3 6.1 8 protected natural areas–PNA 11.9 4 3.8 11 3.2. Maps of Sub-Criteria Based on Levels of Land Suitability Figure 4 shows the reclassified maps, based on suitability thresholds (Table 1), of the biophysical and administrative sub-criteria. The sub-criteria with the largest ‘highly suitable’ area, regarding their criteria group, were: distance to geological faults (56.1%) and protected natural areas (14.6%), while those with the largest ‘not suitable’ area were terrain slope (61.0%) and distance to roads (81.2%) (Table 5). Hence, distance to geological faults is the sub-criterion that most favours the selection of sites for weather stations in Amazonas, while distance to roads is the most restrictive. Table 5. Suitability area for sub-criteria to select weather station sites in Amazonas, NW Peru. Highly Suitable Moderately Suitable Marginally Suitable Not Suitable Criteria Sub-Criteria km2 % km2 % km2 % km2 % elevation 9074.79 21.6 12,653.08 30.1 20,322.50 48.3 0.00 0.0 terrain slope 1401.33 3.3 6746.96 16.0 8253.54 19.6 25,648.55 61.0 terrain hill shade 1273.10 3.0 38,736.11 92.1 1457.69 3.5 583.47 1.4 Biophysical land use/land cover 6370.39 15.1 3338.05 7.9 31,968.04 76.0 373.89 0.9 distance to water sources 22,745.08 54.1 8202.42 19.5 5512.17 13.1 5590.70 13.3 distance to geological faults 23,594.53 56.1 5077.33 12.1 6106.74 14.5 7271.78 17.3 landslide susceptibility 9825.67 23.4 13,935.97 33.1 13,380.99 31.8 4907.74 11.7 distance to roads (km) 2466.98 5.9 2276.15 5.4 3167.66 7.5 34,139.59 81.2 distance to populations (km) 1090.42 2.6 2156.41 5.1 4953.48 11.8 33,850.07 80.5 Administrative distance to the host institution 3821.34 9.1 26,049.86 61.9 7972.20 19.0 4206.97 10.0 protected natural areas 6157.50 14.6 35,892.87 85.4 0.00 0.0 0.00 0.0 Climate 2021, 9, 169 9 of 15 Climate 2021, 9, 169 9 of 15 Figure 4.. Suiittabiilliitty mapss off tthee bbiiophyssiiccall ((aa–g)) aand aadmiiniissttrrattiive ((h–k)) ssub--crriitteriia fforr ssellecttiing llocattiions ffor weatther ssttaattiioonnss iinn Amaazzoonnaass,, NW PPeerruu.. 33..33.. SSuuiittaabbiilliittyy SSuubb--MMooddeell MMaappss WWiitthh tthhee wweeiigghhtteedd oovveerrllaapp ooff ssuubb--ccrriitteerriiaa,, ssuuiittaabbiilliittyy ssuubb--mmooddeellss wweerree ggeenneerraatteedd ffoorr eeaacchh hhiieerraarrcchhiiccaallg groruoupp(F (igFuigruer5ea ,5ba).,bT)h. eTahdem aindimstirnaitsivtreatsiuvbe- msuobd-eml hoaddelt hhealda rgtheest l‘ahriggheslyt ‘shuigithalbyl es’uaitraebal(e6’ 6a3r.e4a5 (k6m632.)45a nkdm‘2n) oatnsdu ‘intaobtl seu’ iatraebale(’1 a0r,2e5a3 (.1307,2k5m3.23)7f okrms2e)l efocrti nseglelcotciantgio lnos- cfaotrioWnsS fionr AWmS ainzo Anmasa(zToanbalse (6T)a. bTleh e6)l.a Tnhde- sluanitdab-siuliittyabmiloitdye ml foodreWl fSor( FWigSu (rFeig5uc)rei n5dc)i ciantdeid- ctahtaetd0 t.h3%at 0(1.30%8. 5(510k8m.525) komf t2)h oefA thmea Azomnarzeogni orenghioans chhaas rcahcateraricstteircissttihcas ttharaet ahrieg hhliyghsluyi tsaubilte- afobrlei nfostra ilnlisntgalWlinSg. TWhSis. ‘Thhigish l‘yhisguhiltyab slue’ittaebrlreit’o treyrrciotorrreys pcoornrdesepdotnod1e1d5 ,t7o0 611350,7×063 030m ×p 3ix0e mls, pwixheiclsh, fwohrmicehd fo2r6m,6e8d3 ‘2h6i,g6h83ly ‘hsuigithalbyl es’upitoalbylge’o pnos lfyogroinsst faollri ningstWalSli.ng WS. Table 6. Areas of suitability of sub-models for selecting sites for weather stations in Amazonas, NW Peru. Highly Suitable Moderately Suitable Marginally Suitable Not Suitable Criteria/Sub Goal km2 % km2 % km2 % km2 % Biophysical 228.01 0.5 25,070.59 59.6 16,646.35 39.6 105.42 0.3 Administrative 663.45 1.6 4135.80 9.8 26,997.76 64.2 10,253.37 24.4 Land suitability for weather stations 108.55 0.3 20,562.52 48.9 21,274.89 50.6 104.41 0.2 Climate 2021, 9, 169 10 of 15 Climate 2021, 9, 169 10 of 15 FiguFreig5u.reS u5.i tSaubitialibtyilimty ampaspfso frosr esleelcetcitningg llooccaattiioonnss ffoorr wweeaaththere srtasttiaotnios nins AinmAamzoanzaos,n NasW, NPeWruP. eru. Table 6. Areas of suitab3i.l4it. yRoelfoscuatbio-mn oofd WelSs faonrds tehlee cSteinlegctsioitne soff oSritwes efaorth Neerwst WatiSo ns in Amazonas, NW Peru. Of the 20 existing WS, 5% (1 WS, WS Chachapoyas-SENAMHI), 70% (14 WS), and 20% (5H WigSh)l yarSeu liotacbalted in ‘hiMgholdye rsautietlaybSleu’i,t a‘mbloederatMelyar sguinitaalblyleS’u aintadb l‘me arginallNy ostuSituaibtalbe’l e Criteria/Sub Goal terraink, mre2spectively% (Table 7).k Tmh2e closest ‘%highly suitkambl2e’ terrain %to the 19 WkmS2 that are % not in ‘highly suitable’ terrain is between 0.9 m (WS Bagua-SENAMHI) and 2387.0 m (WS Biophysical 228.01 0.5 25,070.59 59.6 16,646.35 39.6 105.42 0.3 Congon-UNTRM). One hundred iterations were performed, taking into account the rela- Administrative tive di6s6t3a.n45ces of th1e. 620 existi4n1g3 5W.80S (19 WS9 .8relocate2d6 ,t9o9 7t.h76e neare6s4t .2‘highly1 0s,u25it3a.3b7le’ ter-24.4 Land suitability for weather stationsrain). 1F0ig8.u55re 6 show0.s3 11 poss2i0b,5le6 2d.5i2stributi4o8n.9s of the2 W1,2S7 4n.8e9twork i5n0 .A6 mazon1a0s4, .w41ith var- 0.2 ying numbers of new WS and distances between them depending on each distribution. The northernmost point of Amazonas, in Figure 6, is the furthest (70.79 km) from the ex- 3.4i.sRtinelgo cWatSi oanndof itW wSasa niddetnhteifiSeedl eacst itohne olofcSaittieosnf oorf WNeSw nuWmSber 21. WS numbers 22–27 (Fig- ureO 6fat)h wee2r0e eidxeisnttiinfigedW aSt ,552%.07( k1mW, S48, .W72S kCmh, a4c7h.3a8p komy,a 4s2-S.7E4N kmAM, 42H.0I6), k7m0%, a(n1d4 4W0.3S5) ,kamn d 20% (5 rWelSat)ivaer etol tohcea WtedSs itnha‘th wigehrel yusseudi tfaobr liete’r,a‘tmioon.d Aert aitterlaytiosun i1t0a0b,l teh’ea ‘nhidgh‘mly asurgitianbalell’y poslu- itable’ teryragionn, trheastp wecatsi vfuerltyhe(sTta abwleay7 )f.roTmh ethcel opsreecsetd‘hinigg h99ly WsuS i(tFaibgluer’e t6ekr)r awinast 8o.7t3h ekm19 (