Using remote sensing energy balance and evapotranspiration to characterize montane landscape vegetation with focus on grass and pasture lands

Using remote sensing energy balance and evapotranspiration to characterize montane landscape vegetation with focus on grass and pasture lands

International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172 Contents lists available at SciVerse ScienceDirect Internati...

2MB Sizes 21 Downloads 62 Views

International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172

Contents lists available at SciVerse ScienceDirect

International Journal of Applied Earth Observation and Geoinformation journal homepage:

Using remote sensing energy balance and evapotranspiration to characterize montane landscape vegetation with focus on grass and pasture lands Isabel Pôc¸as a,∗ , Mário Cunha b , Luís S. Pereira c , Richard G. Allen d a

Faculdade de Ciências, Universidade do Porto and CEER – Biosystems Engineering, Campus Agrário de Vairão, 4485-661 Vairão, Portugal Faculdade de Ciências, Universidade do Porto and Centro de Investigac¸ão em Ciências Geo espaciais, Campus Agrário de Vairão, 4485-661 Vairão, Portugal c CEER – Biosystems Engineering, Institute of Agronomy, Technical University of Lisbon, Portugal d University of Idaho, Kimberly R&E Center, USA b

a r t i c l e

i n f o

Article history: Received 29 January 2012 Received in revised form 25 August 2012 Accepted 28 August 2012 Keywords: Landsat METRICtm Irrigated meadows Lameiros Deciduous and evergreen forests PCA

a b s t r a c t Water and energy balance interactions with vegetation in mountainous terrain are influenced by topographic effects, spatial variation in vegetation type and density, and water availability. This is the case for the mountainous areas of northern Portugal, where ancestral irrigated meadows (lameiros) are a main component of a complex vegetation mosaic. The widely used surface energy balance model METRIC was applied to four Landsat images to determine the spatial and temporal distribution of the energy balance terms in the identified land cover types (LCT). A discussion on the variability of evapotranspiration (ET) through the various vegetation types was supported by a comparison between the respective crop coefficients and those available in the literature corresponding to the LCT, which has shown the appropriateness of METRIC estimates of ET. METRIC products derived from images of May and June – NDVI, surface temperature, net radiation, soil heat flux, sensible heat flux, and ET – were used to characterize the LCTs, through application of principal component analysis. Three principal components explained the variance of observed variables and their varimax rotated loadings allowed a good explanation of the behaviour of the explanatory variables in association with the LCTs. Information gained contributes to improve the characterization of the study area and may further support conservation and management of these mountain landscapes. © 2012 Elsevier B.V. All rights reserved.

1. Introduction The montane landscape of Northern Portugal has been moulded across the centuries by the interactions between man and environment, which resulted in a landscape where ancestral semi-natural irrigated meadows – locally called lameiros – play a main role. Lameiros are interconnected with common natural pastures (“baldios”), evergreen and deciduous forests and rainfed cropped fields (mainly rye and potato) to constitute a diversified landscape mosaic (Fig. 1). Since remote times, lameiros are a key element in this traditional landscape because, together with extensive common pasture lands, they are the main source of forage for livestock production, which is the main source of income of local rural populations. Presently, lameiros are also recognized as a protected habitat – Natura 2000 habitat, code 6510 – and for their multi-functional services in montane environments (Pôc¸as et al., 2011b).

∗ Corresponding author. Tel.: +351 220 402 489; fax: +351 220 402 490. E-mail addresses: [email protected] (I. Pôc¸as), [email protected] (M. Cunha), [email protected] (L.S. Pereira), [email protected] (R.G. Allen). 0303-2434/$ – see front matter © 2012 Elsevier B.V. All rights reserved.

Lameiros are irrigated by water spreading onto sloping fields. Water is applied all the year round. The irrigation water is derived from small mountain streams and springs and conveyed to fields through a network of earthen channels to be spread over the pastures through a cascade of small contour ditches. Water in excess is reused downstream, or returns to the water course, or percolates to groundwater with generally non-degraded quality (Pôc¸as et al., 2012). The intersectoral competition for water is expected to impact local water use in lameiros, thus making it essential to develop integrated strategies that support lameiros and landscape conservation. However, water use data are generally unknown. On the one hand, lameiros have centennial traditional management, that corresponds well to the existing fragmented structure and that does not require water measurement. On the other hand, medium to steep slopes and frequent variation in aspect make it difficult to estimate ET using ground measurements, such as eddy covariance or Bowen ratio methods, due to their requirements for accuracy (Allen et al., 2011b; Kalma et al., 2008; Rana et al., 2007). In addition, these ET measurement systems provide only local/point information that are not easily extended to the surrounding landscape because in a hilly area, due to variation in orientations of air-streamlines and fragmentation of the landscape, there is a large


I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172

CP DEM DF dT EF ET ETinst ETo ETr ETrF G H IM Kc Kc act Kc mid Kratio

common pastures digital elevation model deciduous forests near-surface temperature gradient evergreen forests evapotranspiration instantaneous evapotranspiration grass reference evapotranspiration alfalfa reference evapotranspiration fraction of evapotranspiration (alfalfa reference) soil heat flux sensible heat flux irrigated meadows crop coefficient (grass-based) actual crop coefficient (grass-based) mid season crop coefficient ratio between alfalfa and grass reference evapotranspiration adjustment factor to minimize soil effect L LAI leaf area index LCM land cover map LCT land cover type METRICtm Mapping Evapotranspiration at high Resolution using Inverse Calibration NDVI normalized difference vegetation index NIM non-irrigated meadows principal component PC PCA principal component analysis PRR PaRedes do Rio R rotated principal component rainfed crops RC Rn net radiation soil adjusted vegetation index SAVI SLT SaLTo surface temperature Ts TsDEM surface temperature according to DEM latent heat of vaporization  E latent heat flux NIR reflectance at the near infrared wavelength band red reflectance at the red wavelength band

variability in the spatial distribution of energy fluxes, making that measurements in a reference site do not relate well with energy fluxes measured on different slopes (Mutiga et al., 2010; Rana et al., 2007). Furthermore, the quantification of the water balance components is complex in hilly and fragmented areas, and is difficult to be performed with acceptable accuracy due to the large spatial variability of vegetation, runoff, subsurface flow, infiltration and deep percolation. Considering the aforementioned limitations and the importance of lameiros in the context of the mountain landscapes of Northern Portugal, remote sensing has been considered to quantify the spatial and temporal distribution of ET and energy balance terms, and, further, to support the characterization of the main land cover types (LCTs). These LCTs were previously classified using remote sensing information (Fig. 1, Pôc¸as et al., 2011a). Satellite-based surface energy balance models have been successfully applied to estimate agricultural water use and ET and to map the spatial distribution of energy fluxes from general landscapes. The METRICtm model, Mapping EvapoTranspiration at high Resolution using Internalized Calibration (Allen et al., 2005a, 2007b), originated from SEBAL, Surface Energy Balance Algorithm for Land (Bastiaanssen et al., 1998), is an example of ET

estimation and mapping models that utilize satellite sensors having a thermal band. The METRIC model has been widely and successfully used to estimate ET over a wide range of vegetation types and various spatial scales (Allen et al., 2005a, 2007a, 2011a; Anderson et al., 2012; Chávez et al., 2009; Gowda et al., 2008; Santos et al., 2008; Tasumi and Allen, 2007), including mountainous terrain. The METRIC “mountain model” integrates corrections for slope, aspect and elevation, which are parameters influencing the solar radiation received on a slope, and contains algorithms to estimate terrain roughness and wind acceleration in mountainous terrain that impact convective heat transfer (Allen et al., 2007b). Thus it is a valuable tool for assessing spatial and temporal dynamics of energy fluxes in mountain landscapes such as that of Northern Portugal. METRIC has been broadly applied for water management, water balance estimation in hydrologic modelling, water use estimation for irrigated agriculture, including the development of crop coefficients curves, and assessing historical water use aimed at water rights transfers (Allen et al., 2005a, 2007a, 2011a; Gowda et al., 2008; Santos et al., 2008; Tasumi and Allen, 2007). Following referred applications, energy balance and ET data obtained through METRIC, coupled with the normalized difference vegetation index (NDVI) and surface temperature data, could be useful for characterizing and mapping biophysical patterns of LCTs, and to further assess the impacts of land and water management on landscape health. In fact, through vegetation characteristics and phenology, LCTs influence processes like solar radiation absorption and thermal emission, as well as latent and sensible heat fluxes and photosynthesis (Maselli et al., 1998; Pettorelli et al., 2005; Xiao et al., 2004). Thus, using METRIC derived parameters to characterize LCTs may provide for an improved definition of patterns and variation of biophysical parameters and for the characterization of landscape elements, thereby supporting future remote sensing applications for monitoring and management. Vegetation indices, which highlight specific plant properties or quantities via spectral transformation of two or more bands, have been widely used for monitoring vegetation dynamics and biophysical vegetation parameters (Cunha et al., 2010; Glen et al., 2008; Nagler et al., 2005; Pettorelli et al., 2005). However, vegetation indices alone may not provide an effective differentiation in vegetation phenology or type that can occur with seasonal and land cover change (Julien et al., 2006). Thus, due to its potential interest, a few attempts have been made to integrate NDVI with additional information to improve vegetation characterization. For example, Melesse et al. (2008) considered energy fluxes and land surface parameters such as surface temperature, NDVI, albedo and emissivity to compare pasture fields where different grazing management were practiced. The coupled use of land surface temperature and NDVI was described by Julien and Sobrino (2009) and Julien et al. (2006) aimed at improving vegetation description at the regional scale in terms of vegetation type and climate influence, and by Nemani and Running (1997) to differentiate seasonal vegetation changes. In this context, the integrated use of vegetation indices such as NDVI with METRIC derived surface energy fluxes and surface roughness shows potential for improved LCTs characterization, thereby providing means to further predict impacts of land-cover changes and land management on the overall landscape. The characterization of LCTs through variables that are somehow interdependent can be supported by principal component analysis (PCA) because it transforms a dataset of original variables, interrelated or correlated to various degrees, into a new dataset of fewer and new orthogonal, uncorrelated variables while keeping information loss to a minimum (Bastianoni et al., 2008; Hair et al., 1998; Raziei et al., 2008). Considering the aspects analysed before, the main goals of this research, focusing a complex montane landscape of Northern Portugal, are (i) to assess spatial and temporal distribution of the

I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172


Fig. 1. Location of the study area (Montalegre) and main land cover types (adapted from Pôc¸as et al., 2011a). The area where test sites (Paredes do Rio – PRR and Salto – SLT) were selected is identified and a sample area (grey box), later considered in Fig. 4, is also highlighted.

energy balance terms, including ET, derived from remote sensing through the METRICtm algorithm, in relation to the primary land cover types and (ii) to test the relevance of using METRICtm output data, mainly NDVI, sensible heat flux, soil heat flux, net radiation, surface temperature, and ET to better characterize and improve the explanation and understanding of primary vegetation types. This characterization study is expected to ultimately contribute to an improved biophysical characterization of LCTs in relation to their distinct ET and energy balance behaviour across time and space, to complement the existing information of landscape classification, and to support further implementation of integrated strategies for water and land use management.

The landscape of Montalegre is typical of traditional mountain agrarian systems, combining a mixture of land cover types, mainly lameiros, common natural pastures (“baldios”), rainfed cropped fields, and evergreen and deciduous forests (Fig. 1). A more detailed description of main LCTs in the landscape of Montalegre is documented in Pôc¸as et al. (2011a,b). Permanent meadows and

2. Data and methods 2.1. Study area Study areas were selected within the Montalegre municipality (806 km2 ), Northern Portugal, between latitudes 41◦ 34 37 and 41◦ 56 34 N and longitudes 7◦ 33 23 and 8◦ 08 03 W (Fig. 1). The Atlantic Ocean influence is prevalent in this region. Precipitation is high, averaging 1530 mm per year and occurring mainly during autumn and winter. Temperature in winter is low and a large number of frost days occur; mean monthly temperatures range from 3.5 ◦ C to 17.2 ◦ C (Fig. 2).

Fig. 2. Average monthly precipitation (R, mm), number of frost days, mean air temperature (Tmean , ◦ C), and minimum air temperature (Tmin , ◦ C) at Montalegre (1951–1980, INMG, 1991).


I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172

grasslands, i.e. lameiros and baldios, comprise about 70% of the agricultural area (INE, 2001), thus being a key element. Although lameiros fields are usually small, they are often grouped into larger contiguous areas covering the montane valleys, with mean patch size of 3.8 ha (Pôc¸as et al., 2011b). Lameiros are irrigated all the year round, in winter and early spring to control frost impacts on vegetation, and in late spring and summer for satisfying crop water requirements. The irrigation frequency is variable among fields depending on location, water availability and water rights. However, a number of lameiros are not irrigated in summer when water is insufficient and they are located far from the water sources. An orthophoto map having 0.5 m ground resolution from March 2003 and information gathered from field surveys were used to select two areas inside the Montalegre municipality – Paredes do Rio (PRR) and Salto (SLT) for focused study. The total number of pixels for each studied area was 23,633 in PRR and 20,490 in SLT. A variety of test sites representative of the most important regional LCTs were selected: (i) irrigated meadows (IM); (ii) non irrigated meadows (NIM); (iii) common pastures or “baldios” (CP); (iv) rainfed crops (RC); (v) deciduous forests (DF); and (vi) evergreen forests (EF). The rainfed crops consist primarily of rye (Secale cereale) and potato (Solanum tuberosum). For each study area, nine test sites were selected for each LCT (Fig. 1 and Table 1). For the non-irrigated lameiros and rainfed crops less test sites were defined because in the characterization process these two classes were considered together. A digital elevation model (DEM), derived from altimetry and planimetric data (scale 1/10,000), was used to correct the surface temperature according to the differences in elevation and to produce slope and aspect maps required in METRIC to estimate solar radiation (Allen et al., 2006). A land cover map, obtained from CORINE Land Cover 2000 (scale 1/100,000), was used to support estimation of the momentum roughness length used in calculating convective heat transfer.

procedure (Allen et al., 2005b). Alfalfa reference ETr is used to calibrate METRIC algorithm for sensible heat flux because ETr values tend to correspond to the upper limit on ET over a wide range of vegetation types (Allen et al., 2007c, 2011b). Air temperature, wind speed and precipitation data were collected at the Montalegre weather station (Latitude 41◦ 48 N, longitude 7◦ 47 W and elevation 1005 m). This weather station is considered to be a reference station based on assessment procedures described by Allen et al. (1998). All weather data were subjected to quality control following the procedures recommended by Allen et al. (2005b). Vapour pressure was estimated using daily minimum air temperature following recommendations in FAO-56 (Allen et al., 1998). Solar radiation was estimated on satellite overpass days, which were clear days, using theoretical solar radiation corresponding to clear sky conditions. Precipitation data were used to perform a soil water balance of the soil upper layer for days preceding satellite overpass days for assessing the ET conditions for bare soil. No precipitation occurred on any of the Landsat image dates. However there was precipitation occurrence within 10-day periods prior to two image capture dates: 40 mm before May 29th distributed over seven days, with larger amounts on May 22 (24 mm) and 23 (10 mm); 14 mm before September 2nd but prior to August 26th. General information on phenological stages of vegetation and general conditions during the satellite image periods is summarized in Table 2. The LCTs of the studied area corresponded to classes identified in both the Land Cover Map retrieved from the CORINE Land Cover 2000 (Table 3) and the LCTs classification performed by Pôc¸as et al. (2011b). The classes of the land cover map were established using standard CORINE designations for Europe and adapted to Portuguese conditions (Painho and Caetano, 2006).

2.2. Satellite data

The METRIC ET algorithm is based upon the energy balance at the land surface. The latent heat flux (E), which is readily converted to ET, is estimated by subtracting soil heat flux (G) and sensible heat flux (H) from net radiation (Rn):

One Landsat5 TM and three Landsat7 ETM+ images (path 204/row 31) for spring and summer periods of 2002: June 22nd, May 29th, June 30th, and September 2nd were used. Images had L1T processing level (geometric and terrain correction). Cloud cover was absent over the selected study area for all satellite images, but some clouds were present over the Salto study area in the image of June 30th. For all test sites, NDVI and soil adjusted vegetation index (SAVI; Huete, 1998) were computed from satellite image reflectance data by combining top of atmosphere reflectance for the red (3) and near infrared (4) wavelength bands: NDVI =

NIR − red NIR + red



(NIR − red ) (1 + L) (NIR + red + L)


where NIR and red are the reflectances respectively at the near infrared and the red wavelength bands, and L is an adjustment factor to minimize the soil effect, assumed to equal 0.1 in METRIC applications (Allen et al., 2007b, 2010). SAVI was used to estimate leaf area index (LAI), which is used in METRIC to estimate aerodynamic roughness. 2.3. Meteorological data and vegetation dynamics Meteorological data, including wind speed, air temperature, solar radiation, and vapour pressure were used to estimate the alfalfa reference evapotranspiration (ETr) computed using the ASCE

2.4. Application of METRICtm algorithms

E = Rn − G − H


METRIC utilizes information on the thermal and shortwave bands from Landsat5 TM and Landsat7 ETM+ satellite images (Allen et al., 2010). In the current study, METRIC was applied to the four Landsat images to produce spatial distributions of ET, expressed as a fraction of alfalfa reference evapotranspiration (ETrF) and expressed as daily ET. Because of the complex mountainous terrain conditions, the mountain version of METRIC model was applied, which considers adjustments for the influence of slope, aspect and elevation on solar and thermal radiation estimates (Allen et al., 2007b). The following variables were used in estimating energy balance components (Allen et al., 2007b, 2010): (i) short and longwave radiation, albedo and surface emissivity for estimating net radiation (Rn); (ii) surface temperature, albedo, and NDVI for the estimation of G; (iii) surface temperature for estimating the temperature gradients above the surface (dT), and aerodynamic resistance and wind speed for the estimation of H (Fig. 3). The well-known approach of a scaled reference ET is used to translate instantaneous ET to longer periods (Allen et al., 1998, 2007b). The scaling coefficient is the reference ET fraction (ETrF), which is the ratio of the instantaneous ET derived for each pixel to the alfalfa reference ET (ETr) (Fig. 3). ETrF corresponds to the wellknown crop coefficient (Kc ) used with the grass reference ET (Allen et al., 1998, 2007c). ETr expresses the evaporative demand of the atmosphere relative to the alfalfa reference crop, thus representing

I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172


Table 1 Number of pixels per test site selected in both study areas, Paredes do Rio (PRR) and Salto (SLT). Land-cover type

No. pixelsa Test site



Test site



Irrigated lameiros (IM)


35 23 25 22 20

41 61 46 36 28


27 21 45 77

15 40 30 24

Non-irrigated lameiros (NIM)


18 18 –

38 47 35


– – –

– – –

Rainfed crops (RC)


35 23 46

43 59 45


14 17 24

15 22 26

Common pastures, baldios (CP)


60 70 42 24 32

81 46 72 30 18


18 24 30 40

16 36 27 65

Deciduous forest (DF)


112 29 65 22 20

42 18 33 20 12


66 35 40 70

10 8 8 8

Evergreen forests (EF)


31 24 42 17 20

349 63 70 16 20


14 14 18 42

20 60 21 42

a The number of pixels correspond to Landsat satellite images with 30 m pixel size. EF, DF, CP, RC, IM and NIM are the abbreviations for, respectively, evergreen forest, deciduous forest, common pastures, rainfed crops, irrigated meadows (lameiros) and non-irrigated meadows.

the influence of weather conditions on maximum rates of ET for the alfalfa reference crop (Allen et al., 2005b). ETrF describes the relative amount of ET from specific vegetation when compared with ETr and is primarily influenced by the vegetation type and its phenological characteristics (e.g. growth stages, stomatal controls) that differentiate the vegetation from the reference crop. Considering

the concepts described and the relatively small area of the study, ETr was assumed spatially constant over the study area. The instantaneous ET was computed for each pixel at the instant of satellite image using the following relation (Allen et al., 2007b): ETinst = 3600



Table 2 Vegetation characterization and phenology of the studied land cover types (source: Humphries et al., 1996; Pôc¸as et al., 2011b; Teles, 1970). Land cover types


Vegetation dynamics, phenology and agronomic practices

“lameiros” (semi-natural meadows)

Permanent herbaceous species. Biodiverse pastures Mainly Holcus lanatus, Plantago lanceolata, Dactylis glomerata, Anthoxanthum odoratum, Festuca spp., Cynosurus cristatus, Trifolium dubium.

May: Vegetation is growing quickly. Grasses (Gramineae) are mainly at the leaf and heading phenological stage and Legumes at flowering. June: Vegetation reaches its maximum development (heading/flowering) before the cut for hay (in July); before the end of June water application is cut-off to allow vegetation to dry, aiming at improving the quality of the hay and its preservation. July: Grass cut for hay, followed by the restart of the vegetation re-growth. August to September: Re-growth of the vegetation.

“baldios” (common pastures)

Shrubs and herbaceous permanent species. e.g. Gytisius spp., Pterospartum tridentatum, Ulex europaeus and Erica spp. and Festuca rubra, Agrostis spp., Nardus stricta, Holcus mollis

May: Herbaceous vegetation development slightly delayed relative to lameiros vegetation. June to October: Pastures are used for livestock grazing all summer presenting at the end sparse and dry vegetation (also because of low precipitation values in this period).

Rainfed crops

Mainly rye (Secale cereale) and potato (Solanum tuberosum)

Rye: Flowering usually occurs in late May; ear development by June; and harvest at the beginning of July; September: stubble-land. Potato: The plantation usually occurs between late April and May and harvest in August.

Deciduous forests

Mainly oaks (Q. robur and Q. pyrennaica) and riparian species (Populus sp. and Betula celtiberica)

Tree canopies were totally leaved during all satellite image periods considered. The oak flowering period occurs in April–May and fruit maturation in October. Betula flowering between April and May and fruiting in July–August.

Evergreen forests

Mainly Pinus pinaster

Flowering occurred in June and pine cone maturation in the Autumn of the following year.


I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172

Table 3 Classes defined in the land cover map for the study areas and their correspondence with the considered land cover types. Land cover types (LCT) studieda

Land cover map (LCM) classes Designation




Meadows Agriculture with natural spaces Annual crops associated to permanent crops Non irrigated crops

231 243 241

“Lameiros” “Lameiros” –




Deteriorated forestry spaces, cuts and new plantations Deciduous forests Evergreen forests Natural meadows Shrublands Sparse vegetation


Non-irrigated meadows and rainfed crops –

311 312 321 322 333

Deciduous forests Evergreen forests Common pastures Common pastures Common pastures


a In this study only the most important regional land cover types were considered. Thus, LCTs for the areas corresponding to the classes 241 and 324 of the LCM were not defined.

where ETinst is the instantaneous ET (mm h−1 ), 3600 is time conversion from seconds to hours, E is the latent heat flux, and  is the latent heat of vaporization (J kg−1 ). The METRIC model uses the CIMEC process (Calibration using Inverse Modelling at Extreme Conditions) for internal calibration of sensible heat flux (Allen et al., 2007b). In this model, two “anchor

points for calibration” – the “cold pixel” and “hot pixel” – are selected to define the limit conditions for the energy balance over the study area. The two anchor points allow for determination of calibration coefficients (a and b defined in Fig. 3). The cold pixel is usually selected and defined over a well irrigated and non-stressed cropped field, with ET set to 1.05 of ETr (i.e. ETrFcold pixel = 1.05)

Fig. 3. Simplified scheme of METRIC algorithms (adapted from Allen et al., 2007b, 2010).

I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172

thus representing maximum ET (Allen et al., 2007b). In this application, irrigated lameiros selected within those showing the best conditions were used to define the cold pixels. However, because lameiros are not managed to achieve maximum ET, it results that ET from lameiros do not attain the maximum rate corresponding to ETrFcold pixel = 1.05. Therefore, it was necessary to adjust the ETrFcold pixel to represent the actual vegetation condition, i.e. adopt ETrF < 1.05. Knowing that the grass reference crop coefficient (Kc ) for the mid season period in grazing pastures ranges from 0.7, in extensive grazing, to 0.8–1.0 for rotated grazing (Allen et al., 1998), it was assumed that the Kc for lameiros would not exceed 1.0, even for locations having healthy and abundant green vegetation and water availability. Any grass based Kc can be converted into the alfalfa based ETrF as

ETrFalfalfa-based =

Kc(grass-based) Kratio


where Kratio is the ratio between the alfalfa and grass reference ET (Kratio = ETr/ETo). Hence, the maximum Kc = 1.0 referred above corresponds to ETrF = 0.83, which could be assumed to characterize the cold pixel of METRIC in this application. In all images, the hot pixel was selected and defined for a bare agricultural soil. The surface soil moisture was found, via soil water balance, to be dry enough for three image dates, then corresponding to a very small or null ET, thus allowing to assume H = Rn − G for these hot pixels. Significant precipitation occurred four days prior to the May 29th image, so ETbare soil was then estimated through a daily soil water balance of the surface layer, which was used to estimate H = Rn − G − ETbare soil for the hot pixel (Allen et al., 2007b). METRIC uses an indexed function of radiometric surface temperature for estimating the near-surface temperature gradient, dT. This variable is used to compute the sensible heat flux as discussed by Allen et al. (2011a). A major assumption of METRIC refers to the linearity of the dT vs surface temperature (Ts) function, which is assumed in former studies (Allen et al., 2010; Bastiaanssen et al., 1998; Tasumi et al., 2005). To prevent for the cooling impacts on Ts due to increasing elevation within an image but that are not related to dT and sensible heat flux, a Ts adjustment for each image pixel was performed using the above referred DEM and a customized lapse rate (Allen et al., 2011a). The software ERDAS IMAGINE v.9.1 (Leica Geosystems) was used for the METRIC algorithm application. The complex mountain conditions relative to vegetation cover, fragmentation and relief limit the accuracy of ET and energy balance estimation using ground observations as referred before (e.g. Kalma et al., 2008; Rana et al., 2007). Moreover, a value of each energy balance component, ETrF or ET is assigned to each pixel and these values show a great spatial variability due to the variability of characteristics of vegetation and respective management by farmers. A good example of that variability is shown by Allen et al. (2005a) for values of potato actual crop coefficients in a large area of Idaho. Therefore, the average ETrF values estimated through METRIC for each LCT were compared with reference values in the literature. With this purpose, because information on Kc is more easily available, the ETrF values were first transformed into actual Kc (Kc act ) with Eq. (5). This comparison is possible because Kc values are transferable as discussed by Allen et al. (1998, 2007c, 2011b) and Pereira et al. (1999). The comparison is also easy because climatic conditions in the study area at time of images are close to the standard ones, thus not requiring a climatic adjustment of the Kc . Interpretation of results is also made easier due to existing knowledge resulting from previous LCTs’ studies in the area (Pôc¸as et al., 2011a,b).


2.5. Land cover types characterization A biophysical characterization of Montalegre’s montane LCTs was implemented by employing a number of combinations of METRIC intermediate and final outputs (Fig. 2), mainly NDVI, ET, TsDEM, Rn, G and H, for the four Landsat image dates. The purpose of this characterization was to analyse the behaviour of energy fluxes and land surface parameters across the LCTs aiming to improve the explanation and understanding of each primary LCT and help to explain differences among LCTs. References to the combined use of ET and NDVI are reported by several authors (e.g. Duchemin et al., 2006; Gonzalez-Dugo et al., 2009). That combination may provide insights on the occurrence of stress related with root depth or vegetation height for similar NDVI conditions. ET and surface temperature were used in combination to detect differences in root depth, stomatal control and albedo (e.g. Sandholt et al., 2002). The combined use of energy fluxes and surface parameters is referred by Melesse et al. (2008) to assess impacts of heavy grazing, and by Melesse (2004) for improving land-cover and surface microclimate mapping. Principal component analysis (PCA) was used for the LCTs characterization. This method reduces inter-correlated variables to some new linearly uncorrelated ones called principal components (PCs), which account for as much as possible of the variation in the original variables (Bagayoko et al., 2007; Bastianoni et al., 2008; Hair et al., 1998; Raziei et al., 2008). PCA is commonly used with climate variables for characterizing vegetation types (Bagayoko et al., 2007), or for analysing the spatial variability of climate variables and identifying related homogeneous regions (Lolis et al., 1999; Raziei et al., 2008). PCA is based on the computation of the eigenvalues and eigenvectors of either the correlation matrix or the covariance matrix of the observed variables. As the variables considered herein for the characterization of LCTs have different units, they were standardized prior to further analysis. An orthogonal rotation is widely used to obtain more spatially localized and uncorrelated PCs (Von Storch and Zwiers, 1999). The decision regarding how many PCs to retain for rotation can be based on scree plot of the eigenvalues (Hair et al., 1998). This scree plot rule is based on the estimation of the sampling errors of the eigenvalues associated to the principal components. In this study, PCA was first applied to the combination of variables computed for the four image dates. However, it was found that the variables relative to September 2nd show very different behaviour in relation to the remaining variables, which is physically significant as it relates to a different date and to different phenological stages of the vegetation. Therefore, the variables corresponding to September 2nd were excluded from the analysis and the PCA was applied only on three dates: May 29th and June 22nd and 30th. Based on the scree plot (not shown) the first three leading PCs, explaining approximately 85% of the total variance of the original dataset, were retained and then rotated using varimax orthogonal rotation (Von Storch and Zwiers, 1999).

3. Results and discussion 3.1. Evapotranspiration Fig. 4 shows the spatial and temporal variability of ETrF as indicator of the variability of vegetation ET for May 29th, June 22nd and September 2nd over the sample areas in the PRR and SLT sites identified in Fig. 1. Statistics for some intermediate and final outputs from METRIC are shown in Table 4 for the studied Montalegre LCTs and for the four satellite images used.


I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172

Fig. 4. Maps of the fraction of reference evapotranspiration (ETrF) with identification of the test sites of irrigated ( ) and non-irrigated lameiros ( ) for sample areas of Paredes do Rio (PRR) and Salto (SLT): (a) May 29th; (b) June 22nd; (c) September 2nd. Lines delineate the areas represented in the land cover map corresponding to: IM – lameiros; CP – common pastures; NIM/RC – non-irrigated meadows and rainfed crops; DF – deciduous forest.

The highest ET values were observed by June 22nd for all vegetation types (Table 4); comparing LCTs, the highest ET values were observed in May and June for irrigated lameiros (IM); higher ET values were observed in September for deciduous and evergreen forests (DF and EF) and common pastures (CP) as shown in Table 4. These results are also observable in the sample areas illustrated in Fig. 4, where higher ETrF values refer mainly to areas corresponding to irrigated lameiros. The peak of ET for lameiros was observed for the image of June 22nd (Table 4) which was a period of high vegetation development prior to cutting for hay, which generally occurs in July (Pôc¸as et al.,

2012). However, the values for ETrF were nearly constant (0.73) for the period referring to the images of May and June. The lower value for ET by end June, since ETrF did not change, is due to a low ETr observed by that date, which agrees with decreased values of ET for all vegetation types by June 30th (Table 4). Lower ETrF and ET values were observed for lameiros by September 2nd, i.e. a few weeks after hay cutting, reflecting less frequent irrigation by late summer, which induced a slow development of lameiros vegetation due to water stress, as indicated by a lower NDVI observed at that date (Table 4). In fact, by late summer, farmers often use the available water to irrigate other crops (e.g. vegetable

I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172


Table 4 Descriptive statistics of METRIC intermediate and final outputs averaged for each satellite date and land cover type. METRIC


Land cover type (LCT) IM Mean













29-Mai 22-Jun 30-Jun 02-Set Average

0.80 0.81 0.80 0.56 0.74

0.03 0.04 0.03 0.06 0.04

0.72 0.70 0.72 0.46 0.65

0.07 0.07 0.01 0.08 0.06

0.58 0.69 0.73 0.59 0.64

0.07 0.02 0.01 0.03 0.03

0.66 0.68 0.70 0.45 0.62

0.06 0.03 0.02 0.08 0.05

0.71 0.81 0.82 0.71 0.76

0.03 0.01 0.01 0.02 0.02

0.66 0.66 0.72 0.55 0.65

0.03 0.04 0.05 0.11 0.06

ET (mm/d)

29-Mai 22-Jun 30-Jun 02-Set Average

3.78 4.48 3.05 2.07 3.34

0.39 0.37 0.19 0.37 0.33

3.64 3.89 2.63 1.87 3.01

0.09 0.27 0.00 0.74 0.27

2.89 3.53 2.39 2.22 2.76

0.42 0.35 0.10 0.39 0.32

3.37 3.60 2.57 1.66 2.80

0.32 0.36 0.08 0.55 0.33

1.79 3.35 1.39 2.91 2.36

0.40 0.41 0.52 0.34 0.42

2.00 3.13 1.00 2.32 2.11

0.51 0.38 0.36 0.68 0.48

ETrF []

29-Mai 22-Jun 30-Jun 02-Set Average

0.74 0.73 0.73 0.55 0.69

0.10 0.09 0.07 0.10 0.09

0.71 0.63 0.61 0.48 0.61

0.02 0.05 0.03 0.20 0.10

0.53 0.53 0.54 0.54 0.54

0.09 0.06 0.10 0.10 0.00

0.65 0.57 0.60 0.42 0.56

0.07 0.07 0.10 0.16 0.10

0.34 0.50 0.33 0.65 0.45

0.09 0.09 0.14 0.04 0.15

0.36 0.47 0.20 0.57 0.40

0.11 0.07 0.07 0.18 0.16

Rn (W/m2 )

29-Mai 22-Jun 30-Jun 02-Set Average

716.9 695.4 747.0 595.3 688.66

39.97 38.52 36.57 55.32 42.59

723.7 702.3 710.2 605.2 685.34

17.12 22.49 1.87 26.73 17.05

690.3 646.7 706.0 559.2 650.54

43.75 39.00 33.92 57.38 43.51

701.8 669.8 710.7 586.4 667.18

27.56 32.40 16.25 43.26 29.87

687.0 641.1 715.6 544.6 647.07

59.54 46.64 82.34 46.11 58.66

714.7 678.0 697.2 560.8 662.68

44.20 39.44 31.81 67.70 45.79

TsDEM (Kelvin)

29-Mai 22-Jun 30-Jun 02-Set Average

295.4 302.0 293.8 302.4 298.38

0.29 0.75 0.89 2.09 1.01

297.3 305.2 295.9 303.8 300.55

0.86 1.09 0.42 2.27 1.16

297.9 303.5 294.1 300.6 299.03

2.46 1.11 0.62 1.21 1.35

297.1 304.7 296.4 304.0 300.57

1.66 1.78 0.58 1.02 1.26

293.7 299.6 290.2 295.7 294.79

1.07 1.23 0.51 0.90 0.93

292.9 300.5 291.4 297.3 295.49

1.23 1.57 1.93 0.75 1.37

G (W/m2 )

29-Mai 22-Jun 30-Jun 02-Set Average

67.14 62.60 68.46 97.68 73.97

4.88 8.25 10.58 12.58 9.07

83.70 86.75 79.19 105.49 88.78

14.43 15.24 0.45 6.48 9.15

112.01 85.11 84.36 93.69 93.79

6.59 8.36 6.69 9.88 7.88

91.93 84.67 82.11 105.33 91.01

12.04 5.58 5.62 3.51 6.69

88.83 64.41 67.03 78.58 74.72

18.94 7.62 11.51 7.10 11.29

120.03 114.64 105.96 94.00 108.66

11.42 10.20 12.76 7.26 10.41

H (W/m2 )

29-Mai 22-Jun 30-Jun 02-Set Average

397.4 311.9 422.7 296.4 357.08

4.41 6.65 3.37 40.59 13.76

404.3 342.9 416.5 323.4 371.79

10.42 9.70 2.70 49.78 18.15

406.8 326.2 433.8 266.5 358.35

13.02 15.52 13.84 27.57 17.49

395.5 333.5 418.8 326.4 368.55

13.27 24.85 5.32 22.37 16.45

480.5 357.3 521.6 225.9 396.31

18.76 21.27 12.21 19.91 18.04

462.5 365.8 521.1 257.4 401.69

10.95 21.40 19.76 21.04 18.29

For the descriptive statistics (mean and standard deviation SD) the average data from PRR and SLT test sites are considered. However data for Salto of the Jun 30th image were not considered because of clouds present in that area in the Landsat image. NDVI – normalized difference vegetation index; ET – evapotranspiration (daily); ETrF – fraction of reference evapotranspiration; Rn – net radiation; TsDEM – surface temperature according to the digital elevation model; G – soil heat slux; H – sensible heat flux; IM – irrigated meadows; NIM – non-irrigated meadows; CP – common pastures; RC – rainfed crops; DF – deciduous forests; EF – evergreen forests.

crops) rather than meadows. This may be confirmed for the study area of Paredes do Rio (Fig. 4c, PRR), where the highest ETrF values observed by September 2nd for the IM class corresponded to areas located within valleys, where soil moisture is commonly higher. Irrigated and non-irrigated lameiros have shown different behaviours relative to ET, with NIM having ETrF decreasing from May to end June, likely due to a decrease in soil water availability, while IM had a constant ETrF. Smaller ETrF values at all image dates, as well as smaller NDVI indicate that NIM were much more stressed than IM vegetation. The ETrF results were converted through Eq. (5) into grass reference Kc act . For irrigated lameiros and the images of May and June, the Kc act values (0.88–0.89) are in general agreement with various sources of Kc data for irrigated pastures, often managed for hay. Cancela et al. (2006), for irrigated pastures in lowlands of Galicia, north of the study area, found mid season Kc (Kc mid ) of 1.05 for pristine conditions and a smaller value of 0.8 later in season. Greenwood et al. (2009) reported for various irrigated perennial forages cultivated under pristine conditions a Kc mid near 1.05 for northern Victoria, Australia, Allen et al. (1998) refer Kc from 0.85 to 1.05 for rotating grazing. Jia et al. (2009) report for irrigated bahiagrass in Central Florida a smaller Kc mid = 0.9, while Sumner and

Jacobs (2005) report Kc mid = 0.85 for the same grass in Southwest Florida. For Northeast China, Zheng et al. (2012) found a similar value for irrigated grass, Kc mid = 0.82. Considering that values obtained in this study for the irrigated meadows (0.88–0.89) correspond to the average of a large number of pixels and likely refer to the midseason of the grass crop, results provided in the literature confirm those obtained in this study. The ET and ETrF values for non-irrigated lameiros (NIM) in the June images were smaller than for IM (Table 4). The average values for Kc (0.85–0.73) were decreasing from end May to end June due to lack of precipitation since June 11th. These values compare well with the literature. Kc mid for various non irrigated grass mixtures in northeast Germany were in the range 0.65–0.86, with rye grass varying 0.61–0.89 (Mueller et al., 2005); higher values were observed when a water table was high enough to supply the crop. Paz et al. (1996) found Kc mid = 0.80 for non irrigated pastures in Galicia, in an area at north of the study region. Mata-González et al. (2005) refer Kc act ranging 0.79–0.85 for grasslands in Montana. Allen et al. (1998) proposed Kc mid = 0.75 for extensive grazing pastures. Therefore, the estimated average Kc act varying from 0.85 by end May to 0.73 by end June are in agreement with values reported in the literature.


I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172

The maximum ET values for common pastures (baldios) were observed for the image of June 22nd (Table 4). ET had behaviour similar to that of IM and NIM but having lower ET rates, with Kc act varying little, from 0.64 to 0.65. These values are smaller than those for IM and NIM because these pastures are not irrigated contrarily to IM and to NIM, the latter being irrigated until early spring. Relatively low Kc act reflect actual vegetation characteristics, a mix of perennial herbaceous and shrubs under continuous grazing throughout the summer. ETrF (and Kc act ) values observed for CP at Salto by September 2nd were smaller than those for Paredes do Rio (Fig. 4c) due to more intensive livestock grazing and larger livestock concentration in Salto. Melesse et al. (2008) observed similar differences when comparing pastures with moderate vs heavy grazing intensity in North Dakota, USA. The average Kc act obtained for CP compare well with Kc of other ecosystems such as the arid ones of central Argentina, where a Kc mid = 0.65 is reported by Contreras et al. (2011). Similar Kc act of 0.5–0.85 for open vegetation and shrubs in Ethiopian highlands were reported by Descheemaeker et al. (2011). Wight and Hanson (1990) found rangeland Kc act values from 0.75 to 0.90 over most of the growing season, with transpiration coefficients within the 0.40–0.60 range. In a former study in Montana, seasonal Kc act for native vegetation, that did not completely cover the ground, were below 0.55 (Aase et al., 1973). Lower values are referred for Mongolian desert steppe grasslands, with a Kc act of 0.45 when less arid conditions were verified (Yang and Zhou, 2011). The information available allows therefore to say that the average Kc act values found, 0.64–0.65, are within those reported for other parts of world. The rainfed crops (RC), mainly rye and potatoes, have shown an ET pattern across the four images similar to that observed for the non-irrigated meadows (Table 4), with a peak for the image of June 22nd. The ETrF values for RC are also similar to those of NIM. The NDVI values for RC and NIM also have a similar behaviour. The highest ETrF value was observed by May 29th, i.e. during crop mid season, and later decreased as observed in the image of September 2nd. The presence of various crops makes it difficult to assess the ETrF or Kc values obtained. Bodner et al. (2007) report for nonpristine and non irrigated rye Kc mid values of 0.60–0.68, and for a pristine rye crop a Kc mid = 0.90. Relative to the potato crop, data on Kc for potato cultivated in low elevation lands south of the area under study showed Kc mid = 1.15 when pristine conditions apply (Sousa and Pereira, 1999). Differently, Allen et al. (2005a) applied METRIC to vast area of Idaho and found Kc act generally varying 0.85–1.0. Jayanthi et al. (2007) developed a canopy reflectancebased crop coefficient for potato applied at farm level in Idaho, and found Kc act = 0.85; however these values decreased for poor soils to Kc act = 0.77. Hane and Pumphrey (1984) found Kc mid for potatoes around 0.8 during maximum leaf area. Considering results in the literature and that non pristine conditions apply to rainfed crops in the area, it is likely that average values obtained for RC (0.78–0.72) are adequate. Forested areas had lower ET and higher G and H values than agricultural LCTs in May and June. In contrast, in September, both EF and DF had higher ET and lower G and H values (Table 4) than agricultural LCTs. This is likely due to stomatal control of forest trees (Jarvis and McNaughton, 1986; Pataki and Oren, 2003). The period of observations (May to September) corresponds to a period with high NDVI for DF and EF. A clear decrease in ET was observed for the June 30th image, when a peak in sensible heat flux was observed (Table 4), which relates to a decrease in transpiration caused by a decrease in surface conductance. According to Wilson and Baldocchi (2000), the partitioning of energy fluxes into sensible and latent heat fluxes is related to surface conductance, which decreases for some forest species after the leaf expansion period if low soil moisture occurs. The absence of precipitation after June

11th suggests low soil water content, which would favour stomatal control, thus low ET. Evergreen forests have ET depending upon soil water availability, atmospheric demand, characteristics of species, and age and density of stands. ET values for evergreen forests were within the range of values observed by Poyatos et al. (2007) for pine stands in Mediterranean mountains. Attarod et al. (2011), for a Japanese red pine forest, referred a peak Kc of 0.93 or 0.81 when rainfall was insufficient; after the peak, in July, Kc decreased to 0.7–0.8. Zheng et al. (2012) found for Pinus sylvestris a Kc mid = 0.62. Considering these data, it can be assumed that values computed for EF are likely to be acceptable, i.e. Kc = 0.6 by end June, when atmospheric demand is highest and stomatal control was likely higher, and Kc = 0.78 by early September, when stomatal control is lesser. ET of deciduous forests also depends upon soil water availability, atmospheric demand, characteristics of species in terms of stomatal control, and age and density of stands, which also influence Kc . While Zheng et al. (2012) reported for poplar species Kc mid = 0.87, other authors report lower values. Gazal et al. (2006) reported that stands near an intermittent stream had Kc act = 0.43 ± 0.14 and that near a perennial stream Kc act increases to 0.74 ± 0.24. Hou et al. (2010) report Kc act = 0.62 at midseason for Populus euphratica in an arid region of northwest China, and Nagler et al. (2005) referred Kc act from 0.63 to 0.78 for deciduous trees in Rio Grande, New Mexico. It can be concluded that a Kc act = 0.56 by end June and Kc act = 0.68 by early September are acceptable values.

3.2. Land cover type characterization The results analysed above, for the plots selected (Table 1), are consistent and compared well with results from a range of other studies, thus providing confidence in using ET and energy balance results for LCT characterization. Therefore, the combination of NDVI–ET–TsDEM–Rn–G–H values from the image dates of May 29th (29M), June 22nd (22J) and June 30th (30J) was considered for the LCTs characterization. PCA was used with this objective. Based on the analysis of the scree plot of the eigenvalues (not shown), the first three principal components were retained and subsequently varimax rotated to obtain more spatially localized and uncorrelated PCs. The rotated PC loadings (R1, R2, and R3), the eigenvalues, and the percentages of explained variances and cumulative variances of the retained PCs are presented in Table 5, while the cloud distribution of variables between the three PCs, based on the loading values, is illustrated in Fig. 5. The three leading PCs explain 84.8% of the total variance of the whole dataset. The high loadings (identified in bold in Table 5) indicate good correlations between the variables and the PCs. The first PC, R1, explains 37.29% of the total variance (Table 5). It has high positive loadings on TsDEM variables for the three dates, and high negative loadings on NDVI variables (Table 5 and Fig. 5a). As illustrated in Fig. 5a and b, the variables with positive loadings on R1 are mainly related to common pastures (CP) and rainfed crops (RC) while those with negative loadings are mainly associated with irrigated meadows (IM) and deciduous forests (DF). When comparing the IM and RC, higher TsDEM for RC reflects water stress of rainfed crops due to lack of irrigation. Contrarily, the lower surface temperature for IM indicates the cooling effect of evapotranspiration and low stress of irrigated vegetation. This is in agreement with higher NDVI values of IM (Table 4) and its association with ET as shown relative to the second PC loading, R2, in Fig. 5a and b. An inverse trend between NDVI and land surface temperature was also observed by other authors (Gao et al., 2011; Melesse et al., 2008). This inverse trend NDVI vs TsDEM is reflected when comparing deciduous forests with common pastures (Fig. 5b), the latter having lower NDVI.

I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172


Fig. 5. Varimax rotated principal components loading plot, with the three rotated PC loading identified as R1, R2 and R3: (a) distribution of variables based on loading variables; (b) distribution of land cover type sites: abbreviations EF, DF, CP, RC, and IM refer, respectively, to evergreen forest, deciduous forest, common pastures, rainfed crops, and irrigated meadows (lameiros).

The second PC, R2, explains 27.94% of the total variance (Table 5). It has high positive loadings on ET variables for the three dates, and high negative loadings on H variables (Table 5 and Fig. 5a), thus evidencing the contrast between latent and sensible heat fluxes. It results a positive association of R2 to IM and, to a less extent to

RC (where non-irrigated meadows are included), and a negative association with forests, both EF and DF. This reflects the stomatal control of forest trees as discussed above, resulting in low ET and high H in contrast with meadows (Fig. 5b). Baldocchi et al. (2004) also reported higher H values for oak woodlands (DF in this study)


I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172

Table 5 Varimax rotated PC loadings, eigenvalues, explained variances and cumulative variances. The first column includes the variables used for the PCA. R1



Eigen value Percentage of explained variance Cumulative percentage of explained variance

7.0728 37.29

5.0287 27.94

3.1641 17.58




Loading 1

Loading 2

Loading 3

NDVI29M NDVI22J NDVI30J ET29M ET22J ET30J Rn29M Rn22J Rn30J TsDEM29M TsDEM22J TsDEM30J G29M G22J G30J H29M H22J H30J

–0.7384 –0.8393 –0.7863 0.1154 –0.4157 0.0817 –0.0144 –0.0196 –0.0401 0.7042 0.8138 0.7715 0.4258 0.4231 0.4401 –0.2798 0.1602 –0.2613

0.4262 0.2774 0.0459 0.9563 0.8541 0.9375 0.0581 0.0981 0.0461 0.4243 0.3875 0.3611 –0.5420 –0.4378 –0.3510 –0.8503 –0.7228 –0.8686

–0.2115 –0.2858 –0.0983 0.1969 0.1130 0.1120 0.9690 0.9733 0.9692 –0.1287 0.1386 –0.0676 0.6212 0.7157 0.6899 0.2626 0.4936 0.3204

with aerodynamically rough features than for grasslands having an aerodynamically smother canopy. The third PC, R3, explains 17.58% of the total variance (Table 5). It has high positive loadings on Rn and G variables for the three dates and negative loadings on NDVI (Table 5 and Fig. 5a). These low negative loadings are associated with IM, the vegetation group having higher NDVI during June (Fig. 5b). Contrasting, IM show to have the lowest G values, since in lameiros the ground is well covered by vegetation, and EF have shown to have the highest G (Table 4), probably because there is few vegetation under the evergreen trees. The fraction of vegetation cover and type of vegetation affects soil heat flux due to radiation extinction by the canopy; a smaller G is associated with denser vegetation as reported by Verhoef et al. (2012) among others. Differently, the high positive loadings relative to Rn are associated to high Rn values of EF (Table 4), which may refer to albedo. The use of referred METRIC products for the characterization of landscape classes, where lameiros and common pastures are key elements, provided for the integration of phenological properties, management practices such as irrigation, vegetation density, and surface energy balance during the spring-summer period. 4. Conclusions The use of several METRIC outputs, such as NDVI, ET, Ts, Rn, H and G, for different dates supported the characterization of land cover types in test sites of the montane landscape of Northern Portugal. This biophysically based characterization provided for better understanding of main LCTs and to assess and compare surface energy balance components in relation to LCTs. Results allowed appropriate comparison of crop coefficients characterizing each type of vegetation with reference values in the literature, hence providing for a positive test of METRIC application in estimating ET in addition to former model applications to other regions, environments and vegetation. The use of several image dates from May and June for the LCTs characterization provided appropriate information to differentiate among LCTs and a more integrated response of the energy balance parameters’ behaviour. The use of PCA and adopting varimax

rotated loadings supported appropriate differentiation among variables and LCTs, as well as the explanation of differences among LCTs. The results obtained may be of high value and importance for future monitoring of montane landscape health, with primary focus on lameiros sustainability, which may be impacted in future due to demographic degradation, changes in water policies and in montane economy, as well as climate change. The results obtained may contribute to support integrated land and water management of these landscapes, particularly for the preservation of the landscapes focusing on ancestral irrigated lameiros. Nevertheless, further developments are required relative to create scenarios for montane landscape dynamics as affected by demography, policies and climate change, as well as relative to the use of METRIC in mountain landscapes that could lead to more accurate information on water use. Further studies and mapping at a wider scale are also required in the perspective of improved monitoring of montane landscapes. Acknowledgements This study was funded by the project PTDC/AGRAAM/67182/2006, “Fundac¸ão para a Ciência e a Tecnologia”, Portugal. The first author also acknowledges the same institution for the PhD grant (SFRH/BD/24373/2005) and the Kimberly Research Centre – University of Idaho for supporting her training. Thanks are due to Dr. Jeppe Kjaersgaard and Dr. Magali Garcia of the University of Idaho for providing METRICtm training and for review by the University of Idaho that was partially supported by the National Science Foundation – Idaho EPSCoR Program under NSF award number EPS-0814387, to Mr. A. Moura for helping with field work, and to Dr. Tayeb Raziei for advising in PCA. References Aase, J.K., Wight, J.R., Siddoway, F.H., 1973. Estimating soil water content on native rangeland. Agricultural Meteorology 12, 185–191. Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration – Guidelines for computing crop water requirements. FAO Irrigation and drainage paper 56, FAO – Food and Agriculture Organization of the United Nations, Rome, Italy. Allen, R.G., Tasumi, M., Morse, A., Trezza, R., 2005a. A Landsat-based energy balance and evapotranspiration model in western US water rights regulation and planning. Irrigation and Drainage Systems 19, 251–268. Allen, R.G., Walter, I.A., Elliott, R.L., Howell, T.A., Itenfisu, D., Jensen, M.E., Snyder, R.L., 2005b. The ASCE Standardized Reference Evapotranspiration Equation. ASCE – American Society of Civil Engineers, New York, 204pp. Allen, R.G., Trezza, R., Tasumi, M., 2006. Analytical integrated functions for daily solar radiation on slopes. Agricultural and Forest Meteorology 139, 55–73. Allen, R.G., Tasumi, M., Morse, A., Trezza, R., Wright, J.L., Bastiaanssen, W., Kramber, W., Lorite, I., Robison, C.W., 2007a. Satellite-based energy balance for mapping evapotranspiration with internalized calibration (METRIC) – Applications. Journal of Irrigation and Drainage Engineering 133, 395–406. Allen, R.G., Tasumi, M., Trezza, R., 2007b. Satellite-based energy balance for mapping evapotranspiration with internalized calibration (METRIC) – Model. Journal of Irrigation and Drainage Engineering 133, 380–394. Allen, R.G., Wright, J.L., Pruitt, W.O., Pereira, L.S., Jensen, M.E., 2007c. Water requirements. In: Hoffman, G.J., Evans, R.G., Jensen, M.E., Martin, D.L., Elliot, R.L. (Eds.), Design and Operation of Farm Irrigation Systems. , 2nd ed. ASABE, St. Joseph, MI, pp. 208–288. Allen, R.G., Tasumi, M., Trezza, R., 2010. METRICtm . Mapping Evapotranspiration at High Resolution. Applications Manual for Landsat Satellite Imagery. University of Idaho, Kimberly, ID. Allen, R., Irmak, A., Trezza, R., Hendrickx, J.M.H., Bastiaanssen, W., Kjaersgaard, J., 2011a. Satellite-based ET estimation in agriculture using SEBAL and METRIC. Hydrological Processes 25, 4011–4027. Allen, R.G., Pereira, L.S., Howell, T.A., Jensen, M.E., 2011b. Evapotranspiration information reporting. I. Factors governing measurement accuracy. Agricultural Water Management 98, 899–920. Anderson, M.C., Allen, R.G., Morse, A., Kustas, W.P., 2012. Use of Landsat thermal imagery in monitoring evapotranspiration and managing water resources. Remote Sensing of Environment 122, 50–65. Attarod, P., Aoki, M., Bayramzadeh, V., Ahmadi, M.T., 2011. Micrometeorological observations above a Japanese red pine forest within the growing season. Turkish Journal of Agriculture and Forests 35, 597–609.

I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172 Bagayoko, F., Yonkeu, S., Elbers, J., van de Giesen, N., 2007. Energy partitioning over the West African savanna: multi-year evaporation and surface conductance measurements in Eastern Burkina Faso. Journal of Hydrology 334, 545–559. Baldocchi, D.D., Xu, L., Kiang, N., 2004. How plant functional-type, weather, seasonal drought, and soil physical properties alter water and energy fluxes of an oak–grass savanna and an annual grassland. Agricultural and Forest Meteorology 123, 13–39. Bastiaanssen, W.G.M., Menenti, M., Feddes, R.A., Holtslag, A.A.M., 1998. A remote sensing surface energy balance algorithm for land (SEBAL). 1. Formulation. Journal of Hydrology 212–213, 198–212. Bastianoni, S., Pulselli, F.M., Focardi, S., Tiezzi, E.B.P., Gramatica, P., 2008. Correlations and complementarities in data and methods through Principal Components Analysis (PCA) applied to the results of the SPIn-Eco Project. Journal of Environment Management 86, 419–426. Bodner, G., Loiskandl, W., Kaul, H.-P., 2007. Cover crop evapotranspiration under semi-arid conditions using FAO dual crop coefficient method with water stress compensation. Agricultural Water Management 93, 85–98. Cancela, J.J., Cuesta, T.S., Neira, X.X., Pereira, L.S., 2006. Modelling for improved irrigation water management in a temperate region of Northern Spain. Biosystem Engineering 94, 151–163. Chávez, J.L., Gowda, P.H., Howell, T.A., Copeland, K.S., 2009. Radiometric surface temperature calibration effects on satellite based evapotranspiration estimation. International Journal of Remote Sensing 30, 2337–2354. Contreras, S., Jobbágy, E.G., Villagra, P.E., Nosetto, M.D., Puigdefábregas, J., 2011. Remote sensing estimates of supplementary water consumption by arid ecosystems of central Argentina. Journal of Hydrology 397, 10–22. Cunha, M., Marc¸al, A.R.S., Silva, L., 2010. Very early prediction of wine yield based on satellite data from VEGETATION. International Journal of Remote Sensing 31, 3125–3142. Descheemaeker, K., Raes, D., Allen, R., Nyssen, J., Poesen, J., Muys, B., Haile, M., Deckers, J., 2011. Two rapid appraisals of FAO-56 crop coefficients for semiarid natural vegetation of the northern Ethiopian highlands. Journal of Arid Environment 75, 353–359. Duchemin, B., Hadria, R., Erraki, S., Boulet, G., Maisongrande, P., Chehbouni, A., Escadafal, R., Ezzahar, J., Hoedjes, J.C.B., Kharrou, M.H., Khabba, S., Mougenot, B., Olioso, A., Rodriguez, J.C., Simonneaux, V., 2006. Monitoring wheat phenology and irrigation in Central Morocco: on the use of relationships between evapotranspiration, crops coefficients, leaf area index and remotely-sensed vegetation indices. Agricultural Water Management 79, 1–27. Gao, Z., Gao, W., Chang, N.-B., 2011. Integrating temperature vegetation dryness index (TVDI) and regional water stress index (RWSI) for drought assessment with the aid of LANDSAT TM/ETM+ images. International Journal of Applied Earth Observation and Geoinformatics 13, 495–503. Gazal, R.M., Scott, R.L., Goodrich, D.C., Williams, D.G., 2006. Controls on transpiration in a semiarid riparian cottonwood forest. Agricultural and Forest Meteorology 137, 56–67. Glen, E.P., Huete, A.R., Nagler, P.L., Nelson, S.G., 2008. Relationship between remotely-sensed vegetation indices, canopy attributes and plant physiological processes: what vegetation indices can and cannot tell us about the landscape. Sensors 8, 2136–2160. Gonzalez-Dugo, M.P., Neale, C.M.U., Mateos, L., Kustas, W.P., Prueger, J.H., Anderson, M.C., Li, F., 2009. A comparison of operational remote sensing-based models for estimating crop evapotranspiration. Agricultural and Forest Meteorology 149, 1843–1853. Gowda, P., Chávez, J., Howell, T., Marek, T., New, L., 2008. Surface energy balance based evapotranspiration mapping in the Texas High Plains. Sensors 8, 5186–5201. Greenwood, K.L., Lawson, A.R., Kelly, K.B., 2009. The water balance of irrigated forages in northern Victoria, Australia. Agricultural Water Management 96, 847–858. Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1998. Multivariate Data Analysis, 5th ed. Prentice-Hall International Inc. Hane, D.C., Pumphrey, F.V., 1984. Yield-evapotranspiration relationships and seasonal crop coefficients for frequently irrigated potatoes. American Potato Journal 61, 661–668. Hou, L.G., Xiao, H.L., Si, J.H., Xiao, S.C., Zhou, M.X., Yang, Y.G., 2010. Evapotranspiration and crop coefficient of Populus euphratica Oliv forest during the growing season in the extreme arid region northwest China. Agricultural Water Management 97, 351–356. Huete, A., 1998. A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment 25, 295–309. Humphries, C.J., Press, J.R., Sutton, D.A., 1996. Árvores de Portugal e Europa. FAPAS, Fundo para a Protecc¸ão dos Animais Selvagens e Câmara Municipal do Porto. INE, 2001. Recenseamentos Gerais da Agricultura. Dados Comparativos 1989–1999. Instituto Nacional de Estatística, Lisboa. INMG, 1991. O Clima de Portugal. Fascículo XLIX. Normais climatológicas da região de Trás-os-Montes e Alto Douro e Beira Interior correspondentes a 1951–1980. INMG, Lisboa. Jarvis, P.G., McNaughton, K.G., 1986. Stomatal control of transpiration: scaling up from leaf to region. Advanced Ecological Research 15, 1–49. Jayanthi, H., Neale, C.M.U., Wright, J.L., 2007. Development and validation of canopy reflectance-based crop coefficient for potato. Agricultural Water Management 88, 235–246. Jia, X., Dukes, M.D., Jacobs, J.M., 2009. Bahiagrass crop coefficients from eddy correlation measurements in central Florida. Irrigation Science 28, 5–15.


Julien, Y., Sobrino, J.A., Verhoef, W., 2006. Changes in land surface temperatures and NDVI values over Europe between 1982 and 1999. Remote Sensing of Environment 103, 43–55. Julien, Y., Sobrino, J.A., 2009. The Yearly Land Cover Dynamics (YLCD) method: an analysis of global vegetation from NDVI and LST parameters. Remote Sensing of Environment 113, 329–334. Kalma, J., McVicar, T., McCabe, M., 2008. Estimating land surface evaporation: a review of methods using remotely sensed surface temperature data. Surveys in Geophysics 29, 421–469. Lolis, C.J., Bartzokas, A., Metaxas, D.A., 1999. Spatial covariability of the climatic parameters in the Greek area. International Journal of Climatology 19, 185–196. Maselli, F., Gilabert, M.A., Conese, C., 1998. Integration of high and low resolution NDVI data for monitoring vegetation in Mediterranean environments. Remote Sensing of Environment 63, 208–218. Mata-González, R., McLendon, T., Martin, D.W., 2005. The inappropriate use of crop transpiration coefficients (Kc) to estimate evapotranspiration in arid ecosystems. A review. Arid Land Research and Management 19, 285–295. Melesse, A.M., 2004. Spatiotemporal dynamics of land surface parameters in the Red River of the North Basin. Physics and Chemistry of Earth, Parts A/B/C 29, 795–810. Melesse, A.M., Frank, A., Nangia, V., Hanson, J., 2008. Analysis of energy fluxes and land surface parameters in a grassland ecosystem: a remote sensing perspective. International Journal of Remote Sensing 29, 3325–3341. Mueller, L., Behrendt, A., Schalitz, G., Schindler, U., 2005. Above ground biomass and water use efficiency of crops at shallow water tables in a temperate climate. Agricultural Water Management 75, 117–136. Mutiga, J.K., Su, Z., Woldai, T., 2010. Using satellite remote sensing to assess evapotranspiration: case study of the upper Ewaso Ng’iro North Basin, Kenya. International Journal of Applied Earth Observation and Geoinformatics 12 (Suppl. 1), S100–S108. Nagler, P.L., Cleverly, J., Glenn, E., Lampkin, D., Huete, A., Wan, Z., 2005. Predicting riparian evapotranspiration from MODIS vegetation indices and meteorological data. Remote Sensing of Environment 94, 17–30. Nemani, R.R., Running, S.W., 1997. Land cover characterization using multitemporal red, near-IR and thermal-IR data from NOAA-AVHRR. Ecological Applications 7, 79–90. Painho, M., Caetano, M., 2006. Cartografia de ocupac¸ão do solo: Portugal Continental, 1985–2000, Corine Land Cover 2000. Instituto do Ambiente, Amadora. Pataki, D.E., Oren, R., 2003. Species differences in stomatal control of water loss at the canopy scale in a mature bottomland deciduous forest. Advanced Water Resources 26, 1267–1278. Paz, A., Neira, A., Castelao, A., 1996. Soil water regime under pasture in the humid zone of spain: validation of an empirical model and prediction of irrigation requirements. Agricultural Water Management 29, 147–161. Pereira, L.S., Perrier, A., Allen, R.G., Alves, I., 1999. Evapotranspiration: review of concepts and future trends. Journal of Irrigation and Drainage Engineering 12, 45–51. Pettorelli, N., Vik, J.O., Mysterud, A., Gaillard, J.-M., Tucker, C.J., Stenseth, N.C., 2005. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology and Evolution 20, 503–510. Pôc¸as, I., Cunha, M., Marc¸al, A.R.S., Pereira, L.S., 2011a. An evaluation of changes in a mountain rural landscape of Northeast Portugal using remotely sensed data. Landscape and Urban Planning 101, 253–261. Pôc¸as, I., Cunha, M., Pereira, L.S., 2011b. Remote sensing based indicators of changes in a mountain rural landscape of Northeast Portugal. Applied Geography 31, 871–880. Pôc¸as, I., Cunha, M., Pereira, L.S., 2012. Dynamics of mountain semi-natural grassland meadows inferred from SPOT-VEGETATION and field spectroradiometer data. International Journal of Remote Sensing 33, 4334–4355. ˇ J., Llorens, P., 2007. Modelling evapoPoyatos, R., Villagarcía, L., Domingo, F., Pinol, transpiration in a Scots pine stand under Mediterranean mountain climate using the GLUE methodology. Agricultural and Forest Meteorology 146, 13–28. Rana, G., Ferrara, R.M., Martinelli, N., Personnic, P., Cellier, P., 2007. Estimating energy fluxes from sloping crops using standard agrometeorological measurements and topography. Agricultural and Forest Meteorology 146, 116–133. Raziei, T., Bordi, I., Pereira, L.S., 2008. A precipitation-based regionalization for Western Iran and regional drought variability. Hydrological Earth System Science 5, 1309–1321. Sandholt, I., Rasmussen, K., Andersen, J., 2002. A simple interpretation of the surface temperature/vegetation index space for assessment of surface moisture status. Remote Sensing of Environment 79, 213–224. Santos, C., Lorite, I.J., Tasumi, M., Allen, R.G., Fereres, E., 2008. Integrating satellitebased evapotranspiration with simulation models for irrigation management at the scheme level. Irrigation Science 26, 277–288. Sousa, V., Pereira, L.S., 1999. Regional analysis of irrigation water requirements using kriging: application to potato crop (Solanum tuberosum L.) at Trás-os-Montes. Agricultural Water Management 40, 221–233. Sumner, D.M., Jacobs, J.M., 2005. Utility of Penman-Monteith, Priestley-Taylor, reference evapotranspiration, and pan evaporation methods to estimate pasture evapotranspiration. Journal of Hydrology 308, 81–104. Tasumi, M., Trezza, R., Allen, R.G., Wright, J., 2005. Operational aspects of satellitebased energy balance models for irrigated crops in the semi-arid U.S. Irrigation and Drainage Systems 19, 355–376.


I. Pôc¸as et al. / International Journal of Applied Earth Observation and Geoinformation 21 (2013) 159–172

Tasumi, M., Allen, R.G., 2007. Satellite-based ET mapping to assess variation in ET with timing of crop development. Agricultural Water Management 88, 54–62. Teles, A.N., 1970. Os lameiros de montanha do norte de Portugal. Subsídios para a sua caracterizac¸ão fitossociológica e química. Agronomia Lusitana XXXI. Verhoef, A., Ottlé, C., Cappelaere, B., Murray, T., Saux-Picart, S., Zribi, M., Maignan, F., Boulain, N., Demarty, J., Ramier, D., 2012. Spatio-temporal surface soil heat flux estimates from satellite data; results for the AMMA experiment at the Fakara (Niger) supersite. Agricultural and Forest Meteorology 154–155, 55–66. Von Storch, H., Zwiers, F.W., 1999. Statistical Analysis in Climate Research. Cambridge University Press, Cambridge, 494pp. Wight, J.R., Hanson, C.L., 1990. Crop coefficients for rangeland. Journal of Range Management 43, 482–485.

Wilson, K.B., Baldocchi, D.D., 2000. Seasonal and interannual variability of energy fluxes over a broadleaved temperate deciduous forest in North America. Agricultural and Forest Meteorology 100, 1–18. Xiao, X., Zhang, Q., Braswell, B., Urbanski, S., Boles, S., Wofsy, S., Moore III, B., Ojima, D., 2004. Modelling gross primary production of temperate deciduous broadleaf forest using satellite images and climate data. Remote Sensing of Environment 91, 256–270. Yang, F., Zhou, G., 2011. Characteristics and modeling of evapotranspiration over a temperate desert steppe in Inner Mongolia, China. Journal of Hydrology 396, 139–147. Zheng, X., Zhu, J.J., Yan, Q.L., Song, L.N., 2012. Effects of land use changes on the groundwater table and the decline of Pinus sylvestris var. mongolica plantations in southern Horqin Sandy Land, Northeast China. Agricultural Water Management 109, 94–106.