Changes in Snow Surface Albedo and Radiative Forcing in the Chilean Central Andes Measured by In-Situ and Remote Sensing Data

1. Introduction

Uncontaminated snow surfaces are among the whitest surfaces on Earth and have one of highest broadband albedos in the solar spectrum range [1]. Surface snow albedo (SSA) is a fundamental factor for the reflection of incident shortwave radiation (ISR) in the global snow-atmosphere budget system [2,3,4,5]. In the visible (VIS) and near-infrared (NIR) spectral ranges (<1µm), SSA is strongly influenced by the presence of light-absorbing particles (LAPs), whereas for wavelengths above 1µm the variability of SSA is mainly controlled by the grain size of snow crystals [6,7]. The deposition of LAPs in snow darkens the surface of the snowpack, decreasing the reflection of ISR and the albedo [8,9,10]. Moreover, grain metamorphism is accelerated in the presence of LAPs, through a process that is typically referred to as the snow-grain-size feedback, [11,12], which results in a higher ISR absorbing capacity [13,14]. Consequently, LAPs increase the radiative forcing (RF) with the potential to accelerate snowmelt and disturb the hydrologic cycle in snow-dominated basins [15]. The deposition of LAPs in snow has also been observed to contribute to regional warming [16].

Several studies have documented the reduction of SSA due to LAPs such as Black Carbon, Brown Carbon and Mineral Dust, and the effects on local hydrological processes [12,13,17,18,19]. Black Carbon (BC), also known as soot [20], is an aerosol generated from the incomplete combustion of fossil fuels, biofuels, and biomass [11,21]. In Asia, BC emissions were found to be a significant precursor of glacier melting in the Himalayas [19], the Tibetan Plateau [22] and regions in Northern China [23]. Brown Carbon (BrC) is an organic carbon (OC) fraction, primarily emitted by the incomplete combustion of biomass or by the oxidation of anthropogenic and biogenic gaseous precursors [24]. The effect of BrC is comparable to the snow darkening effect of black carbon, ranging from around half to the same darkening of snow surfaces as soot [25]. Mineral dust is transported to snow-covered areas from arid regions or from anthropogenic sources like opencast mining and the construction industry [26,27]. In the Upper Colorado River Basin, heavy dust loads deposited on snow due to industrial and urban expansion resulted in a shift of the timing of peak snow-melt driven runoff and an increase in evapotranspiration as vegetation was exposed earlier [28]. Observations also suggest that Saharan dust anticipates the snowmelt which implies a shorter winter season in the European Alps [29] altering its hydrological response.

Research along the Chilean central Andes has primarily focused on the reduction of snow albedo by BC [27,30,31]. Since obtaining field measurements in high mountain environments is a difficult task, studies have mainly included point measurements of SSA that do not account for the spatial variability of the complex climatic and topographic conditions of the Chilean Andes [32,33]. The lack of knowledge of the spatial variability of SSA and the distribution of LAPs over the Andes renders albedo estimations for hydrological modelling to empirical parameters, which increases uncertainties in runoff predictions and does not consider the environmental impacts driven by the deposition of these particles. In this context, remote sensing tools emerge as a valuable tool as they provide data at high spatio-temporal resolution for monitoring snow covered regions at high elevations [34], offering promising opportunities to improve the assessment, observation, and modeling of snow processes [1,35,36].

This study aims to evaluate changes in RF associated with the presence of BC, which is emitted by anthropogenic activities such vehicular traffic and the mining industry that occur in the Andes [37]. We separate the solar spectrum into a broadband (Bbr) and enclose (Ecr) ranges to isolate the impact of BC [38]. The broadband range considers wavelengths of 300-2500 nm, while the enclose range considers a narrower band with wavelengths between 841-876 nm. We use observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) and modeled ISR from the Chilean Solar Explorer (SE) to evaluate temporal and spatial changes in RF, which are validated against in-situ SSA and ISR measurements collected in the Upper Aconcagua River Basin (UARB) in the Chilean central Andes. The MODIS Dust Radiative Forcing in Snow (MODDRFS) [39,40] product was used to estimate the continuous and integrated daily RF for each spectral range, which are analyzed for temporal trends and variability throughout the 2004-2016 period. The paper is organized as follows: Section 2 describes material and methods, and Section 3 presents the results and discussion. Finally, conclusions are presented in Section 4.

2. Materials and Methods

2.1. Study Area

Central Chile is characterized by a Mediterranean climate, with mild-wet winters and dry summers [41]. Precipitation occurs during the austral winter due to sporadic cold fronts, and orographic effects may produce very localized rainfall patterns [42]. Since 2010 the central Chile region has experienced a so-call mega-drought, characterized by a series of dry years that typically show a 25% to 45% deficit in annual precipitation [43]. Several studies have reported significant decreases in snow cover duration and extent [34,44] and significant losses in glacier mass [45] over the last decade.

The study area corresponds to the Upper Aconcagua River Basin (UARB) in the Chilean Central Andes (Figure 1). The watershed is located in the Valparaíso district approximately 80 km northeast of Santiago, Chile’s largest city and capital. The UARB extends between 32°15’–33°15’S and 70°0’–70°45’W, with an area of approximately 3,660 km² and elevations ranging from 800 to 6,000 meters above sea level (m.a.s.l.). Mean watershed altitude is approximately 3,000 m.a.s.l. The UARB includes four meteorological stations, which are managed by the Dirección General de Aguas (DGA), and provide daily precipitation and temperature data (Resguardo Los Patos, Riecillos, and Vilcuya) and snow depth (Portillo, see Figure 1). A laboratory-refuge, managed by the Centre for Environmental Technologies (CETAM) of Universidad Técnica Federico Santa María (USM), named NUNATAK-1 (NNTK-1) (32°50’43’’S, 70°07’47’’W and 2,813 m.a.s.l.), measures atmospheric variables such as incident and reflected radiation (shortwave and longwave), air quality (aerosols and gases) and meteorology, including snow height since 2016. Monthly mean temperature in the UARB typically oscillates between 9.7°C to 10.9°C and 19.6°C to 21.1°C during the austral winter and summer, respectively, with an annual mean of 15.2°C. Mean annual precipitation is 360 mm, with approximately 280 mm occurring in fall-winter seasons (May to August). Maximum snow depth occurs in July or August, with values typically exceeding 1,700 mm (snow water equivalent of 813 mm) and a historical maximum of 2,590 mm in August 1997 (snow water equivalent of 1,201 mm).

2.2. Database

2.2.1. NUNATAK-1 laboratory-refuge

NUNATAK-1 (NNTK-1) is a laboratory-refuge located at an altitude of 3,000 m.a.s.l. (Figure 1) located within the winter quarters of the Mountain School of the Chilean Army in Portillo, border crossing between Chile and Argentina. Its main purpose is to study environmental pollution and atmosphere-cryosphere interactions, specifically how air pollution affects the melting of glaciers and climate change [32,46]. NNTK-1 measures variables such as concentration and size distribution of aerosol, atmospheric concentrations of BC and BrC, and net solar irradiance. A portable Kipp & Zonen 4-Component Net Radiometer (CNR4) is used to measure incoming and reflected solar irradiances and estimate surface snow albedo. The radiometer includes a pyranometer, with a sensitivity of 10-20 µVW^-1m^-2 and maximum uncertainty of 5% (95% confidence level), which covers the solar spectrum in the 300 to 2,800 nm range. The instrument registers the incident and reflected radiation at 1-minute frequency. For this study, we only used information from the pyranometer for the 2016-2019 period, which was available at the time of data processing. The in-situ surface snow albedo (SSA_{in situ}) was calculated for the 300-2,800 nm range, as follows:

$$SS{A}_{in situ}=\frac{R_r}{R_i}$$

(1)

where R_r and R_i correspond to the reflected and incident radiation, respectively. The ISR and SSA data collected at a frequency of 1 min were aggregated at a daily timescale (only considering daylight hours) with the purpose of comparing them to those obtained from the SE and MODIS. The in-situ SSA data was filtered to: (i) eliminate data recorded before sunrise and after sunset, (ii) discard spurious measurements caused by sensor obstruction and (iii) discard data from days with incomplete records.

2.2.2. MODIS data processing

The Moderate Resolution Imaging Spectroradiometer (MODIS) is one of the main instruments of the NASA Earth Observing System (EOS) and operates onboard the Terra and Aqua satellites, with records starting in 2000 and extending until the present [47]. The MODIS-Aqua (MYD10A1) and MODIS-Terra (MOD10A1) provide daily and complementary observations of the Earth, passing over our study area approximately at 10 AM and 2 PM (local time, GMT-3), respectively.

We used the reflectance data from band 2 (841-876 nm) from the Surface Reflectance Daily L2G Global 500m SIN Grid, V6 product (M*D09GA) to assess the level of absorption of incident shortwave radiation (ISR) in the range where the effects of BC are maximized (~880 nm) [38,48]. The SSA in a broadband spectral range for snow (300 to 2,500 nm) was obtained from the Snow Cover Daily L3 Global 500m SIN Grid, V6 product (M*D10A1) [47]. Each data pixel has an indicator for the quality of the measurement (QA), which is used to filter the data. Data is used only if the QA indicates the best quality (QA=0) or good quality (QA=1), as a criterion to increase the accuracy of radiative forcing estimations from satellite imagery and reduce the uncertainties in the process [39,40]. For validation purposes, data is extracted from the pixel containing NNTK-1 and compared to the in-situ measurements (see Section 2.2.1).

2.2.3. Solar Explorer

The Solar Explorer (SE) (http://solar.minenergia.cl/exploracion) is a tool developed by Chilean Energy Ministry to calculate the photovoltaic potential in the Chilean territory [49]. The model calculates the shortwave radiation at the top of the atmosphere (RTOA) and at surface level including ISR and longwave radiation for the 2004 to 2016 period. For these calculations the model uses the parametrization of solar radiation developed by the Goddard Climate and Radiation Branch (CLIRAD-SW) [50] in combination with information of atmosphere radiation from the TIM/SORCE satellite (Total Irradiance Monitor/Solar Radiation and Climate Experiment) [51], topographic data from the Shuttle Radar Topography Missions (SRTM) site [52,53], and the Geostationary Operational Environmental Satellite (GOES EAST) [54] for cloud detection. The ISR data have been validated against in-situ radiation data from 82 weather stations distributed throughout Chile and provides hourly RTOA and ISR with a spatial resolution of 1 km per pixel (Molina [49]). We used ISR data in our RF calculations after validation against field data from NNTK-1 (see Section 2.2.1).

2.2.4. Validation of SSA and ISR

Snow surface albedo from M*D10A1(SSA_Bb) and modeled ISR from the Solar Explorer were validated against in-situ SSA and ISR measurements from the NNTK-1 station prior to their use in estimating the radiative forcing. Since data were available for different periods at each platform, we could only validate MODIS SSA and ISR data for years 2016-2019 and 2016, respectively. SSA_Bb from the MODIS-Terra and MODIS-Aqua satellites were compared separately against SSA_in-situ, and a correlation analysis was performed on a monthly basis for the entire period.

2.3. Radiative Forcing

2.3.1. MODIS Dust Radiative Forcing in Snow (MODDRFS)

We used the MODDRFS originally developed by Painter et al. [39] and improved by Miller et al. [40] to estimate the radiative forcing (RF) over snow. This algorithm was developed to study the RF on snowpack resulting from the deposition of mineral dust in the 350-850 nm wavelength range. The instantaneous RF is obtained as follows (Miller et al. [40]):

$$RF=\underset{{\lambda }_o}{\overset{\lambda }{{\displaystyle \int }}}{F}_o\left(\lambda ,t\right)D\left(\lambda \right)d\lambda dt$$

(2)

$$F_o\left(\lambda ,t\right)=F_{dir,local}\left(\lambda ,t\right)+F_{dif}\left(\lambda ,t\right)$$

(3)

$$F_{dir,local}=F_{dir}\frac{sin{\theta }_s *\sin s * cos\left({\varphi }_s-\alpha \right)+cos{\theta }_s coss}{cos{\theta }_s}$$

(4)

$$D\left(\lambda \right)=SS{A}_{fresh}-SS{A}_{effective}$$

(5)

where $F_o\left(\lambda ,t\right)$ represents the total ISR at a wavelength  and at time t. The total ISR $F_o\left(\lambda ,t\right)$ has a local and diffuse direct component (Equation 3), where $F_{dir,local}\left(\lambda ,t\right)$ is given by direct radiation $F_{dir}\left(\lambda ,t\right)$ and diffuse radiation $F_{dif}\left(\lambda ,t\right)$. Direct radiation is dependent on the conditions at a particular location such as the angle of solar zenith _s, the angle of solar azimuth _s, slope s and orientation  of the site (Equation 4). $D\left(\lambda \right)$ corresponds to the difference between the surface snow albedo in ideal conditions, e.g., fresh or pure snow (SSA_fresh), and the effective surface snow albedo (SSA_effective) at a wavelength  (Equation 5). In our case, SSA_effective is provided by the in-situ and satellite measurements, and we estimate SSA_fresh using the radiative transfer model SNICAR [55]. The integration limits were defined based on the wavelength range defined for Bbr and Ecr; 300-2,500 nm for the M*D10A1 product (Bbr) and 841-876 nm for M*D09GA (Ecr). Finally, the daily integrated RF (DIRF) from sunrise to sunset is given by:

$$DIRF={{\displaystyle \int }}_{sunrise}^{sunset}{{\displaystyle \int }}_{{\lambda }_o}^{\lambda }{F}_0\left(\lambda ,t\right)D\left(\lambda \right)d\lambda dt$$

(6)

2.3.2. Santa Barbara DISORT Atmospheric Radiative Transfer (SBDART)

We use the Santa Barbara DISORT (Discrete Ordinate Radiative Transfer) Atmospheric Radiative Transfer (SBDART) [56,57,58] model to obtain indirect estimates of the fraction of radiation extracted from the SE that corresponds to the spectral bands provided by the MODIS reflectance products (M*D10A1 and M*D09GA band 2). SBDART models the radiative transfer processes in the atmosphere and Earth’s surface (ISR), and requires input such as time, geographic location, atmospheric profile (tropical, mid-latitude or sub-artic) in terms of water vapor density and pressure, solar zenith (sza) and azimuth (saza), type of surface albedo, and spectral boundaries. After estimating the total radiation in the 280-4000 nm range using SBDART, we can obtain the radiation $F_o\left(\lambda ,t\right)$ that corresponds to the spectral ranges of MODIS products. We find that the Ecr and Bbr ranges correspond to approximately 3% and 99%vof the total $F_o\left(\lambda ,t\right)$ extracted from the SE, respectively (see Figure A1). These percentages show small variability throughout the day, which results primarily from changes in solar zenith and azimuth (±0.3% y ±0.1%, respectively). We also determined that values obtained at 10 AM (MODIS Terra) are close to the daily mean for both spectral ranges, with differences below 1%.

2.3.3. Snow, Ice and Aerosol Radiative Model (SNICAR)

SNICAR [55] is a radiative transfer model developed to simulate spectral snow albedo as a function of snow microstructure, stratigraphy, the content of LAPs and the surface irradiance flux [59]. For this study, SNICAR was used to estimate the fresh and clean snow albedo (SSA_fresh), which is required for the calculation of the radiative forcing (Equations 2 and 5). For the estimation of SSA_fresh, we assumed clear-sky conditions, a snow grain size of 100 µm, which corresponds to a null concentration of pollutants [1,60], and a snow density of 400 kgm^-3, which is representative of seasonal snow [61]. Optical ice properties were estimated based on the ice refraction index proposed by Picard et al. [62]. An optically semi-infinite and uniform snowpack (0.3 m depth) was considered for the study area [60]. SNICAR albedo simulations showed that SSA_fresh is mainly determined by solar position (zenith and azimuth); albedo values at 10 AM throughout the UARB vary only by 1%. Similar variations (1%) were found when computing albedo throughout the day at NNTK-1’s location. These results, i.e., the small spatio-temporal variability of SSA_fresh throughout the study area, allow us to assume that SSA_fresh estimated at NNTK-1’s location can be extrapolated to the entire UARB to investigate the spatial variability of the radiative forcing (see Section 3.2).

2.3.4. Estimation of SSA, D, and RF at NNTK-1

As described on Equation 6, D(λ) is required to estimate the daily integrated radiative forcing (Equation 6). In turn, the estimation of D(λ) requires the SSA and ISR associated to both spectral ranges (Bbr and Ecr). The period of study is defined between 2004-2016, which corresponds to the period where the remote sensing data overlap (MODIS and SE) and the Daily Integrated Radiative Forcing (DIRF) can be estimated. We estimated SSA, D, and DIRF for all days with clear-sky conditions using the remote sensing data extracted at the location of NNTK-1. Additionally, we estimated D for every pixel situated within the UARB for days following a precipitation event. For this, we only considered events with daily images containing more than 1000 pixels with clear-sky condition, which is equivalent to a 10% of the total watershed area. Estimates of D are obtained using Equation 5, where SSA_fresh is taken as a constant throughout the UARB and equal to the value computed at the location of NNTK-1.

3. Results and discussion

3.1. Data validation

MODIS products reveal excessive cloudiness in 2016, such that a comparison against in-situ SSA (NNTK-1) measurements was not possible for this year. For the 2017-2019 period, we observe that the percentage of days with clear-sky condition (valid data) in June-July did not exceed 25% of the whole MODIS data set, while for August this percentage increased to 61%. Comparing the two satellites between June and August, MOD10A1 (Terra) presented a larger number of days with clear-sky conditions, as well as a higher monthly correlation value (between 0.51-0.91), except for July 2017 and 2018 where the correlation values were lower than 0.13. During June-August, MYD10A1 (Aqua) showed correlation values ranging between 0.27-0.81, however, lower values were observed for the months of June, July, and October in year 2017 (0.02, 0.20, and 0.21, respectively). The reason for the low correlation values in these months for both satellites may be attributed to two main factors: (i) few days with useful remote sensing data due to excessive cloudiness, and (ii) days where the elevation of the snow line (zero isotherm) is relatively close to the elevation of NNTK-1, with the implication of low or partial snow coverage at NNTK-1’s pixel (<70%), which underestimates SSA by more than 10%, with respect to the field measurements (see Figure A2). The Solar Explorer data had been previously validated in the study area [49]. Despite this, we computed the correlation between ISR from SE and from NNTK-1, confirming a very high correlation (>0.9) during year 2016.

3.2. Temporal analysis of SSA, D(λ) and RF during 2004-2016 period at NNTK-1

Figure 2 shows SSA values for both spectral ranges (Bbr and Ecr) for the pixel corresponding to NNTK-1. After filtering, 312 days (20% of the study period) with snow and clear-sky conditions were found. SSA exhibits a similar behavior for Bbr (Figure 2a) and the Ecr (Figure 2b), with minimum values during May that increase as the austral winter progresses. The distribution of values until the third quartile is similar for both spectral ranges, ranging between 0.12-0.35 in May, 0.25-0.67 in June, 0.50-0.70 in July, and 0.55-0.75 in August. Monthly mean values (shown as white boxes in Figure 2) are similar, with a maximum variation of 0.02 for the month of May.

To quantify the influence of LAPs on the radiative forcing, we need to compute D (Equation 5) for both spectral ranges. D values are obtained using SSA_fresh simulated by SNICAR during the austral winter between 2004 and 2016 considering clear-sky and the other conditions outlined in Section 2.3. We obtained daily mean values of 0.863±0.007 for the Bbr range and 0.933±0.007 for the Ecr (see Figure 3). . Figure 3 shows estimates of D for the entire study period. For each month, all DEcr values (Figure 3b) are higher than DBbr (Figure 3a), including those below quartile 3 and the monthly mean values. D_Ecr values within quartile 1 and 3 (above q1 and below q3) are approximately 0.07 units higher for the month of July, whereas for the other months (May, June, and August) values in the Ecr range are 0.04 units higher than Bbr. The monthly mean values of D (white boxes) in the Ecr are larger by 0.05 units for May and August (during 2004 to 2016 period) compared to those for Bbr. The monthly mean D_Ecr values were also higher than D_Bbr for June and July, with differences of 0.07 and 0.08 units respectively. We note that the differences between D_Ecr and D_Bbr increase if the analysis is made considering the daily mean values. The values of D for the Ecr were generally higher (as much as 0.25 units) than in the Bbr, with the exception of only 60 days within the 2004-2016 period (Figure 4).

Figure 5 shows results of DIRF for the Bbr and Ecr, which are consistent with our previous results of SSA and D. DIRF reaches the higher monthly mean values (white boxes) in May, with 8.02 and 0.27 MJ m^-2 d^-1 for the Bbr and Ecr, respectively, and decreasing to 2.82 (Bbr) and 0.12 MJ m^-2 d^-1 (Ecr) in July. This scenario follows the trend observed for D between May and July (Figure 3), however, monthly mean values of DIRF in August increased to 3.58 (Bbr) and 0.13 (Ecr) MJ m^-2 d^-1, whereas D continued decreasing. This difference is explained by the fact that ISR values reach minimum values during June and July, and then starts to increase (Figure A4).

In Figure 5 DIRF values reaching 13.21 MJ m-2 d-1 in the Bbr and 0.42 MJ m-2 d-1 in the Ecr. DIRF_Ecr values are within the range reported by Rowe et al. 2019 at the same location. We note that although these authors had used a different spectral range to analyze BC effects in snow (650-700nm), the DIRF values are of the same order of magnitude (0.1-2.3 MJ m^-2 d^-1). We found a maximum Instantaneous RF in the Bbr of 571.54 W m^-2 (observed on August 29th, 2004) which is attributed to the melting of most of the snow at the corresponding pixel (snow cover value of 0.23). The instantaneous radiative forcing found in other regions around the world such as the Himalaya or across the United States, using the 350-850 nm wavelength range, are comparable with the values obtained here for the Bbr (e.g., [28]).

To compare the DIRF values between the Ecr and Bbr spectral ranges, which have different magnitudes of received radiation (3% and 99%, as shown in Section 2.3.2), we normalize the values with respect to Bbr. This normalization allows for the determination of the percentage impact on Ecr in terms of the DIRF (DIRF_Ecr/Bcr), which can then be compared to the normalized radiation received in the Ecr range (ISR_Ecr/Bbr). From the monthly mean values in Figure 4, the normalized DIRF_Ecr/Bcr values are 3.42%, 3.74%, 4.22%, and 3.74% for the months of May to August, respectively. Similarly, the normalized ISR_Ecr/Bbr values are 3.17%, 3.22%, 3.23% and 3.17% for the same month, respectively. As observed, there is a difference between the ISR_Ecr/Bb and DIRF_Ecr/Bcr ratios for each of the months evaluated, with higher ratios observed for DIRF_Ecr/Bcr. This suggests that the variations in SSA in the Ecr are higher, and that this increase may be attributed to the presence of LAPs as these particles have a greater impact on the albedo at wavelengths between 841-876 nm (Ecr) than in the range of 300-2500 nm (Bbr), regardless of the total amount of received radiation [15,63,64].

With the purpose of following daily changes in SSA, we selected consecutive days without liquid precipitations and with equal or less snow cover after the initial day to make sure that local snowfall at Portillo did not occur (see Table 1). Figure 6 shows the variations in spectral SSA (for MODIS bands 1-4) for the periods that satisfy the former conditions throughout our study period. As expected, we observe that SSA decreases with time (from one day to the next) for most cases, which is consistent with the aging of snow and the size increase of snow grains that occur as time progresses after the storm [9,65]. For some of the selected periods, such as June 28-29th 2004 (Figure 6a), August 15-16th 2007 (Figure 6b), July 2-th 2016 (Figure 6g), and June 6-7th 2016 (Figure 6m), the spectral SSA values showed an evident decrease in the Ecr (841-876 nm) and the closest MODIS band (620-670 nm), larger than in the 459-479 nm and 545-565 nm bands. However, for the majority of days shown in Figure 6. the decrease in spectral SSA did not show much variation between bands (e.g., Figure 6c–e,h,l,m). Although there was one period where a slight increase in SSA was observed (Figure 6d), these data show a systematic decrease in spectral SSA for the visible wavelengths (MODIS bands 1-4) and particularly for the Ecr, which is usually associated to the presence of BC [19,66]. We note that the other bands in the visible wavelengths may be associated to the presence of mineral dust [39].

Table 1 shows daily SSA variations for MODIS bands 1,3 and 4, and for the Bbr and Ecr spectral ranges, as well as daily D and DIRF values for the Bbr and Ecr ranges. In general, we observe that differences between SSA for bands 1,3,4 and D (Ecr and Bbr) is approximately -0.3, with a larger difference observed for second day (see Figure 6). The maximum daily rates of albedo decrease for the Bbr and Ecr were observed in August 2013 and August 2007, respectively. These reductions in albedo seem to be much higher than values reported by previous studies in the central Andes area [32,33,34]. Cereceda-Balic et al. 2018 reported a reduction of 0.08 units per day in the broadband albedo (SSA_Bbr) resulting from the deposition of BC from vehicle emissions in the Portillo location. We attribute these differences to: i) differences in contamination sources associated to the periods and location considered in the studies, and ii) the fact that the remote sensing measurements consist of only one measurement, either at 10am or 2pm, whereas the in-situ study used continuous data from a radiometer [32].

The results presented in Table 1 also show the differences in radiative forcing (DIRF) for the two spectral ranges of interest (Ecr and Bcr) for two consecutive days. On the second day, DIRF_Bbr shows a 60% increase compared with the average radiative forcing of the evaluated days, whereas DIRF_Ecr exhibited an increase of approximately 50%. When examining DIRF_Ecr/Bbr, the radiative forcing in the Ecr band represents on average a 4.46% of the radiative forcing in the Bbr band, with values reaching a 12.21% (August 18^th, 2013). On the other hand, the incident solar radiation ratio ISR_Ecr/Bbr is on average 3.18% for the days considered in Table 1, with values that reach 3.21% for June 6-7^th, 2016, and 3.14% for August 18^th, 2013. The above confirms previous results indicating that radiative forcing has a greater percentage impact on the narrow Ecr band than on the much broader Bcr band.

4. Conclusions

In this work we used high-quality in-situ measurements of snow surface albedo (SSA) coupled with remote sensing data to make the first estimates of the radiative forcing (RF) over snow in the Chilean Central Andes. We have extended the spectral range of MODDRFS model to wavelengths between 300-2500 nm (Bbr) and obtained a snow reflecting capacity for the NNTK-1 laboratory refuge area, one of the few stations located in the Chilean Andes with capabilities to monitor SSA and related environmental parameters. We focused on isolating the effects of the deposition of BC in snow, which is known to be a key factor influencing SSA variability in the study, by considering the 841-876 nm wavelength range (Ecr). The higher sensitivity of SSA in the Ecr (than in the Bcr) suggests that BC particles (and probably other LAPs) are present in the Andean snowpack impacting snowpack radiation reflection. In the southern area of the UARB, several emission sources of BC can be identified as it is close to Santiago (Chile’s capital and largest city), the international Argentina-Chile route (Paso Los Libertadores) and mining activities (33°08’55’’S, 70°16’53’’W). Overall, we find that reductions in albedo are generally higher than those reported by previous studies in the area using in-situ characterizations, which we attribute to differences in atmospheric contamination that may exist in the different time periods used in the studies and the temporal resolution of the remote sensing data. Nonetheless, the remote sensing MODIS data proved to be a valuable tool for estimating the radiative forcing in regions that are typically inaccessible and improve our understanding of the variability in SSA, which is essential to accurately predict snowmelt and glacier retreat, which in turn control river discharge and freshwater availability in mountain regions.