Dataset Specifications

Name :CH-POA300-v1
Release date:February 2025
Author:Y. Frischholz, CRYOS, EPFL
Type:Grids
Reference system:WGS84 (epsg:4326)
Spatial resolution:0.0028° (approx. 300m)
Spatial extent ([lat] x [lon]):[45.75,47.87] x [5.75,10.75]
Timezone:UTC
Time resolution:1 hour
Time extent:2009-2019 (current version)
Format:NetCDF
Approx. size:70GB/year

CH-POA300-v1 is a set of maps of hourly solar irradiance (W/ m 2 ) projected onto south oriented tilted surfaces, similar to monofacial solar panels, at a spatial resolution of 300m over Switzerland. Such irradiance projection is referred to as POA, short for plane-of-array irradiance. POA for tilts of 0° (horizontal), 30°, 70° and 90° (vertical) are provided. The current temporal extent is: 2009-2019. Pre-computed seasonal and annual statistics of these maps are also provided in the dataset.

1 Base data

1.1 Surface solar radiation

The base surface solar radiation (SSR) data is taken from the HelioMont [Castelli et al., 2014] products provided by MeteoSwiss. This satellite-based retrieval method for SSR is specialized for alpine regions and performs better in such regions than more general methods such as the SARAH methods [Carpentieri et al., 2022]. Heliomont uses the high resolution channel of the Meteosat Second Generation imager HRSEVIRI, conferring a spatial resolution of approx. 1.7km to the SSR maps. The used variables are the horizon-free direct horizontal irradiance ( BHInh ), diffuse horizontal irradiance ( DHInh ) and the ground albedo (ALB). The "nh" subscript stands for "no horizon", meaning the considered variable is theoretically free of any terrain effect (shadowing, reflection).

1.2 Digital elevation model

The base digital elevation model (DEM) is the NASADEM at 1 arcsecond (30m) of spatial resolution. To meet the desired resolution of 300m, it is down-sampled by averaging.

2 Procedure of SSR projection onto tilted surfaces

The procedure used to derive up-sampled POA maps from the base SSR products ( BHInh and DHInh ) is almost identical to that of Ratnaweera et al. [2023], originally described in Müller and Scherer [2005]. It only differs in the computation of the reflected component of POA, for which we add more realistic details. It consists in imposing horizon effects (self- and projected-shadowing) to the original horizon-free SSR estimates at a relatively high spatial resolution. Plane-of-array irradiance can be decomposed into three components:

POA W m [ / 2 ] = POADirect + POADif f use + POAReflected (2.1)

The direct component POAdirect is computed as the projection of the direct normal irradiance (DNI) onto the considered surface.

POADirect = S x ( , y , ( θ sun , φ sun ) t ) ∗ ( n sun · n sur f ) ∗ DNI x y ( , , ( θ sun , φ sun ) t ) (2.2) DNI = 1 cos ( θ sun ) ∗ BHInh x y t ( , , ) (2.3)

where, x and y are the considered pixel position on the map and ( θ sun , φ sun ) t ) is the zenith and azimuth angles couple defining the sun position at time t. The surface orientation and sun position are respectively described by the normal vectors n sur f and n sun defined by their respective tilt/zenith ( θ ) and azimuth ( φ ) angles: n = ( sin ( θ ) cos ( φ ), sin ( θ ) sin ( φ ), cos ( θ )). The tilt angle of the surface is defined similar to the solar zenith angle, from the vertical position. The scalar product of these normal vectors is a factor describing the relative position of the sun to the surface orientation. Multiplying this factor to the DNI is an efficient way of projecting DNI onto the tilted surface. The binary shadow mask S x ( , y , θ sun , φ sun ) is derived from the high resolution DEM (Section 2.1) and is used as activating function.

The diffuse component POAdif f use is derived from the sky-view factor (SVF) (Section 2.1), assuming an isotropic distribution of the diffuse down-welling radiation across the visible sky. In complex terrain, with a high horizon, this is a conservative approach.

POADif f use = SV F ( x y , , ( θ sur f , φ sur f )) ∗ DHInh x y t ( , , ) (2.4)

The reflected component POAref lected is derived using the terrain albedo and the terrain view factor defined as 1-SVF , assuming an isotropic reflectance distribution of the ground surface and no multi-reflection. This is also a conservative approach. Contrasting with Ratnaweera et al. [2023] the incident radiance is further decomposed into direct and diffuse components for the local terrain slope and aspect.

POAReflected = (1 -SV F ( x y , , ( θ sur f , φ sur f ))) ∗ ALB x y t ( , , ) ∗ ( Bterrain + Dterrain ) (2.5)

where Bterrain and Dterrain are the terrain direct and diffuse local incident radiation respectively:

Bterrain = ( n sun · n terrain ) ∗ DNI x y ( , , ( θ sun , φ sun ) t ) (2.6)
Dterrain = SV F ( x y , , ( θ terrain , φ terrain )) ∗ DHInh x y t ( , , ) (2.7)

2.1 Sky-view-factor and shadow maps

The DEM-derived products are pre-computed using the HORAYZON set of algorithms [Steger et al., 2022]. The sky-view factor (SVF) is an agnostic measurement and is computed once at high resolution using an adapted version of the example available on the linked repository. The binary shadow mask is a function of solar angles, thus of time. It is computed once for a single year, also using adapted version of an available example. Neglecting small inter-annual variations, the hour-of-year is used to join any timestamp to the corresponding shadow mask.

3 Similar works

Ratnaweera et al. [2023] published POA seasonal and annual statistics for Switzerland. Tilts of 20°, 30°, 40°, 50°, 60° and 70° are available. The base radiation data is also the HelioMont SSR dataset. The horizon effects were applied at a resolution of 25m of resolution. From the pre-cited publication, it is not clear what part of the HelioMont data was used to produce these statistics.

Meteotest, on behalf of the Swiss Federal Office of Energy (SFOE) published POA monthly, seasonally and yearly statistics. Tilts of 0°, 30°, 75° and 90° are available. The base radiation data comes from the proprietary software Meteonorm [Remund et al., 2020]. A 10m DEM was used for applying the horizon effects, but the available datasets have a spatial resolution of 50m. Twenty years (2000-2019) of data were used to derive these statistics.

Table 3.1: Summary of the time resolutions and the available variables of each dataset. 1:CH-POA300, 2:Ratnaweera et al., 3:SFOE
HourlyDailyMonthlySeasonallyAnnually
POA 0°111, 31, 31, 3
POA 30°111, 31, 2, 31, 2, 3
POA 70°1111, 21, 2
POA 90°111, 31, 31, 3

4 Validation

To validate the computed POA 0° maps, five years (2015-2019) of measurements of global horizontal irradiance (GHI) at 118 ground stations of the SwissMetNet (SMN) network are used.

Ground stations equipped with tilted radiometers being less common, the validation of POA on tilted surfaces (30°, 70°, 90°) is conducted on a single Alpine-PV test site located at the Totalp (lat=46.8377°, lon=9.8134°, WGS84), near Davos (GR) [Anderegg et al., 2023].

The high time resolutions of both the SMN stations (10 minutes) and the test site (10 seconds) measurements allow for a validation of the hourly modeled POA values at multiple resolutions: hourly, daily, monthly, seasonally, and annually. The comparisons at the seasonal and annual resolutions are further completed by joining the available results of other similar works (see Section 3). At higher time resolution, there is no other dataset available for comparison.

4.1 POA0°

Figure 4.1 shows the spatial distribution of the selected SMN stations. Only stations with less than 1% of missing data over the chosen time period are kept.

A first comparison between the SMN stations and the CH-POA300 dataset is provided in Figure 4.2 for the hourly, daily and monthly time resolutions, split seasonally. Metrics are provided for 5 altitude bands: 0-1000m (lowlands stations), 1000-2000m (pre-alpine and alpine stations) and 2000-3000m (high alpine stations), 1000m-4000m (all alpine stations) and 0-4000m (all stations).

Figure 4.1: Spatial distribution of the subset of the SwissMetNet stations used for validation. The measurements of 10 minutes GHI from 118 stations are used. All selected stations have less than 1% of missing data points.

Modelled GHI WIm?

Figure 4.2: Five years (2015-2019) comparison between the CH-POA300 dataset and the measurements at the hourly, daily and monthly time resolutions. Metrics are provided for different altitude bands and aggregated seasonally and annually. In parentheses, next to each metric, the difference to the same metric computed for the base data (without horizon effects) is provided. The vertical lines appearing at the hourly resolution annually and in the summer probably stem from the inaccurate shading at the resolution of 300m.

Figures 4.3 and 4.4 report the comparison at the season time resolution (i.e. seasonal sums).

Figure 4.3: Relative bias of CH-POA300 and the SFOE maps to the SMN stations measurements in seasonal and annual sum averages. Stations are sorted by elevation. The metrics are provided for 5 different elevation bands.

1

1

1

8

1

1

1

H

1

1

%]

summer

5

winter

1
Figure 4.4: Maps of the relative bias between the measurements (2015-2019) of seasonal and annual GHI average sum at the SMN stations (points) and the corresponding CH-POA300 (2015-2019) and SFOE (2000-2019) statistics.

4.2 POA30°, 70° and 90°

Figure 4.5 presents the comparison at the hourly, daily and monthly time resolutions. Figure 4.6 shows the same results in a time-series style. Table 4.1 presents the measurements and relative biases at the seasonal and annual time resolutions.

1

Modelled POA [WIm?]

Figure 4.5: Two years (2018-2019) comparison of hourly, daily and monthly POA between the CH-POA300 data and the measurements at the location of the Totalp test site.
Figure 4.6: Two years (2018-2019) of hourly, daily and monthly POA time series of the CH-POA300 data and the measurements at the location of the Totalp test site.

Table 4.1: Comparison of average seasonal and annual POA sums, between the CH-POA300 maps (2015-2019), the maps of Ratnaweera et al. [2023] (NA), the maps of the SFOE (2000-2019) and the actual measurement at the Totalp test-site (2018-2022). The corresponding relative bias is given in percentages in the parentheses.

POA [kWh/m²]Measurements (Totalp) (2018-2022)CH-POA300 (2015-2019)Ratnaweera et al. NAOFEN (2001-2020)
Tilt=30°Winter697.5617.1 (-11.5%)640.4 (-8.2%)697.0 (-0.07%)
Summer959.4987.2 (+2.9%)1005.7 (+4.8%)1024.0 (+6.73%)
Annual1656.91604 (-3.18%)1646.1 (-0.7%)1721.0 (+3.9%)
Tilt=70°Winter817.4706.8 (-13.5%)728.8 (-10.8%)NA
Summer855.2778.2 (-9.0%)807.3 (-5.6%)NA
Annual1672.6.31485.0 (-11.2%)1536.0 (-8.2%)NA
Tilt=90°Winter856.0664.7 (-22.4%)NA763.0 (-10.9%)
Summer707.1592.5 (-16.2%)NA619.0 (-12.5%)
Annual1563.31257.3 (-19.6%)NA1382.0 (-11.6%)

References

  • D. Anderegg, S. Strebel, and J. Rohrer. Alpine Photovoltaik Versuchsanlage Davos Totalp : Erkenntnisse aus 5 Jahren Betrieb. Sept. 2023. doi: 10.21256/zhaw-2524. URL https://digitalcollection.zhaw.ch/handle/11475/28797. Accepted: 2023-09-29T09:13:16Z Publisher: ZHAW Zürcher Hochschule für Angewandte Wissenschaften.
  • A. Carpentieri, D. Folini, M. Wild, L. Vuilleumier, and A. Meyer. Satellite-derived solar radiation for intra-hour and intraday applications: Biases and uncertainties by season and altitude, Oct. 2022. URL http://arxiv.org/abs/2212.11745. arXiv:2212.11745 [physics].
  • M.Castelli, R. Stöckli, D. Zardi, A. Tetzlaff, J. Wagner, G. Belluardo, M. Zebisch, and M. Petitta. The HelioMont method for assessing solar irradiance over complex terrain: Validation and improvements. Remote Sensing of Environment , 152: 603-613, Sept. 2014. ISSN 00344257. doi: 10.1016/j.rse.2014.07.018. URL https://linkinghub.elsevier.com/retrieve/ pii/S0034425714002673.
  • M.D. Müller and D. Scherer. A Grid- and Subgrid-Scale Radiation Parameterization of Topographic Effects for Mesoscale Weather Forecast Models. Monthly Weather Review , 133(6):1431-1442, June 2005. ISSN 1520-0493, 0027-0644. doi: 10.1175/MWR2927.1. URL http://journals.ametsoc.org/doi/10.1175/MWR2927.1.
  • N. Ratnaweera, A. Kahl, and V. Sharma. Geospatial segmentation of high-resolution photovoltaic production maps for Switzerland. Frontiers in Energy Research , 2023.
  • J. Remund, M. Schmutz, and Graf. Meteonorm Version 8, 2020. URL https://meteonorm.com/assets/publications/5BV. 3.8_pvsec_2020_mn8.pdf.
  • C. R. Steger, B. Steger, and C. Schär. HORAYZON v1.2: an efficient and flexible ray-tracing algorithm to compute horizon and sky view factor. Geoscientific Model Development , 15(17):6817-6840, Sept. 2022. ISSN 1991-9603. doi: 10.5194/gmd-15-6817-2022. URL https://gmd.copernicus.org/articles/15/6817/2022/.