Site adaptation of albedo data

Overview

Satellite-derived albedo data provides global coverage with a long acquisition history, making it the only practical source for long-term albedo characterization at project sites. However, when compared to ground measurements, a systematic deviation (bias) is often observed. Site adaptation corrects this bias by locally adjusting the satellite-derived albedo time series to better represent on-site observations.

Accurate long-term albedo data is particularly important for bifacial PV systems, where albedo directly influences reflected solar radiation reaching the rear side of the modules. A statistically robust assessment requires at minimum 10 years of historical data.

Ground-based albedometer measurements are generally more representative of local surface conditions than satellite estimates. However, they are typically available only for limited periods, insufficient to characterize long-term albedo variability at seasonal and interannual scales. The satellite-based model fills this gap, provided its systematic deviations are corrected through site adaptation.

Sources of systematic deviation

The bias between satellite-derived and ground-measured albedo is driven by several factors:

• Field of view (FOV) representativeness. The area seen by a ground-mounted albedometer (installed at standard height, typically less than 2 m) may not represent the full extent of a utility-scale PV plant. Satellite pixels (approximately 0.5 km spatial resolution) may also miss fine-scale spatial variability, particularly over fragmented land cover. For homogeneous surfaces, satellite estimates are generally close to ground observations.

• 16-day weighted averaging. Solargis albedo time series uses MODIS satellite data, where daily values are calculated as a weighted average over a 16-day window. This averaging approach reduces data gaps caused by cloud cover but limits the reproduction of daily variability. Long-term trends, including monthly, seasonal, and interannual variability, are captured reasonably well.

• Ephemeral snow cover. Snow events are typically accompanied by persistent cloudiness, which prevents the satellite from observing the surface. Additional complications arise from possible misdetection of clouds versus snow, and from rapid snow degradation through melting or soiling. For this reason, snow-affected periods are normally treated separately from snow-free periods in the site adaptation process.

Note: For sites without snow events, snow-specific corrections are not required.

Requirements for site adaptation

Site adaptation of albedo requires that ground measurements meet three prerequisites before the process can be applied:

1. Instrumentation and data quality. Measurements must be obtained with a properly installed and maintained albedometer and must pass quality assessment. Measurements that do not meet quality requirements cannot be used.

2. Length of the measurement period. A minimum of one year of measurements is required to capture the multiple components of albedo variability, including seasonal cycles. Measurements over shorter periods generally do not adequately describe long-term variability and introduce errors when used to correct the satellite-based time series. These errors propagate into PV energy calculations.

3. Surface representativeness. The ground measurements and the satellite data must cover an equivalent surface. Significant differences in the area observed (e.g., due to very heterogeneous land cover within the satellite pixel) may prevent reliable site adaptation.

Important: Using short-period ground measurements for site adaptation introduces systematic errors into the long-term albedo estimate and should be avoided.

Site adaptation method

The site adaptation process consists of four sequential steps.

Step 1: Analysis of ground measurements

Ground albedo measurements, after quality assessment, are inspected to verify that all three prerequisites described above are met. If any prerequisite is not satisfied, site adaptation cannot be performed.

Step 2: Temporal homogenization and data pre-processing

Ground measurements and satellite-derived data are homogenized to a common temporal resolution to make values directly comparable. Daily mean albedo is typically used as the common resolution, since it is the lowest resolution available from the satellite data. Daily mean values from sub-daily ground measurements are calculated as a weighted average of intraday data, accounting for data gaps due to missing records or quality-flagged values.

Once homogenized, both time series are compared to identify and exclude periods with extreme or non-systematic situations. Excluding these periods prevents anomalous events from propagating into the correction coefficients.

Step 3: Calculation and application of correction coefficients

The homogenized data is analyzed for systematic deviations. The analysis considers whether the deviation has a long-term component, including monthly or seasonal structure. Based on this analysis, the appropriate correction method and time scale are selected.

Correction factors are calculated over the overlapping period of ground-measured and satellite-based data. The following considerations apply to the selection of the correction time scale:

• Daily, monthly, and seasonal patterns should each be analyzed to determine the optimal time scale for generating correction factors.

• Monthly and seasonal corrections require careful attention to possible inconsistencies or abrupt shifts in the resulting time series at seasonal transitions.

• Incomplete measurement periods (less than one full year) require an assessment of whether the available data adequately represents albedo variability at the site before monthly or seasonal corrections are applied.

The calculated correction coefficients are then applied to the full historical satellite-based albedo time series, producing a site-adapted dataset that better represents local surface conditions.

Figure 1: Example of site-adapted albedo time series. Ground observations (green), original satellite-derived data (red), and site-adapted satellite data (blue) are shown for an illustrative period.

Step 4: Verification of results

The site-adapted time series is evaluated against the original satellite data and the ground measurements. Statistical metrics are calculated for the overlapping period, including:

• Mean Bias Deviation (MBD)

• Mean Absolute Deviation (MAD)

• Root Mean Square Deviation (RMSD)

• Standard Deviation (STD)

• Correlation Coefficient (CC)

An improvement relative to the uncorrected satellite data is expected. In particular, the MBD should be close to zero after site adaptation. The corrected time series is also inspected with respect to the treatment of extreme values and, where applicable, snow-affected periods.

Figure 2: Scatter plots of satellite-derived albedo versus ground observations before (left, red) and after (right, blue) site adaptation. The dashed line represents the 1:1 reference.

Further reading

• "Surface albedo and reflectance: Review of definitions, angular and spectral effects, and intercomparison of major data sources in support of advanced solar irradiance modeling over the Americas": C. Gueymard, V. Lara-Fanego, M. Sengupta, Y. Xie

• "How to use albedo for more accurate bifacial PV estimates"

• "Satellite-based albedo data"

• "Solar irradiance ground measurements"

• "Solar irradiance site adaptation"