Assimilation of satellite data and insitu data for the improvement of global radiation maps in the Netherlands
(2014) In Student thesis series INES NGEM01 20141Dept of Physical Geography and Ecosystem Science
- Abstract
- For this research, two satellite products were used to see if it was possible to improve the resolution and quality of the global radiation interpolation in the Netherlands. The first data source was from the Climate Monitoring Satellite Application Facility (CM-SAF). The second data source was the Surface Insolation under Clear and Cloudy Skies (SICCS) from the KNMI. Both products were available for the period of January 2006 to December 2011 and came in the form of images with monthly and daily averages. To combine the satellite images with the in- put provided by the KNMI’s 32 measurement stations, these interpolation/merging methods were used: Thin Plate Splines (TPS), Mean Bias interpolation (MB), Interpolated Bias interpolation (IB),... (More)
- For this research, two satellite products were used to see if it was possible to improve the resolution and quality of the global radiation interpolation in the Netherlands. The first data source was from the Climate Monitoring Satellite Application Facility (CM-SAF). The second data source was the Surface Insolation under Clear and Cloudy Skies (SICCS) from the KNMI. Both products were available for the period of January 2006 to December 2011 and came in the form of images with monthly and daily averages. To combine the satellite images with the in- put provided by the KNMI’s 32 measurement stations, these interpolation/merging methods were used: Thin Plate Splines (TPS), Mean Bias interpolation (MB), Interpolated Bias interpolation (IB), Kriging with External Drift; Exponential model (KED-EXP) & Kriging with External Drift; Spherical model (KED-SPH) All these methods made use of the in-situ measurements as main input for the interpolation and all methods except TPS used the satellite products as auxiliary data.
Interpolations were made for the average of the six year period and on monthly measurements, for each month, in each year. Daily interpolations were made for April 2010 until July 2010.
Different validation methods were used to analyze the output. The results showed that; for the six year average both products and all interpolation methods did a good job on predicting global radiation. The R2 was lowest for the IB on the CM-SAF product with a value of 0.19. However, the MAPE (mean absolute percentage error) did not exceed 1.39% on the CM-SAF product and 1.42% on the SICCS product. These values corresponded with an absolute bias of 1.77 W/m2 and 1.8 W/m2. The monthly results showed similar results. The R2 values tended to differ more, especially in the IB and MB interpolation. In most cases this could be explained by the quality of the
Satellite images. The MAPE was low in all cases. A maximum MAPE of 8.38% was found (when using proper satellite images), in November, which corresponded with an absolute bias of ± 4 W/m2. Data splitting returned similar results. MAPE’s did increase up to 9.27% when leaving out 1/4th of the measurement stations but this value corresponds with an absolute bias of 2.71W/m2. These low absolute errors showed that all interpolation methods return an accurate interpolation. However, because the interpolation methods rely on the quality of the satellite images, the SICCS product would be a better product. These images were complete in all months while the CM-SAF product lacked data in December.
Since it turned out that all interpolations performed well, daily data was analyzed for the period of April until July 2010.
It turned out that for the daily data KED and the IB interpolations performed significantly better than the TPS or MB interpolation. The biggest average MAPE was found for the TPS method (10.7% in May). The smallest average errorof0%wasfoundfortheIBmethod. How- ever this method was paired with very low R2 values which made the model unpredictable. The average KED R2 and MAPE ranged from 0.57 to 0.75 and from 0.08% to 0.95 %. This made the method a stable and accurate interpolation method. The satellite images on their own would not be good enough to use directly as a global radiation map, for this time interval. The over- and underestimated bias of the satellite images ranged from -89.63 to 64.49 W/m2. This showed that, a combination of station data and satellite data would improve the quality and resolution of daily global radiation maps. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/4611204
- author
- Van Tiggelen, Jurgen LU
- supervisor
- organization
- alternative title
- Can MSG help us to determine where to place solar cells in the Netherlands?
- course
- NGEM01 20141
- year
- 2014
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- geography, geomatics, physical geography, global radiation, data-assimilation, satellite, The Netherlands
- publication/series
- Student thesis series INES
- report number
- 320
- language
- English
- additional info
- Supervisor Raymond Sluiter, the Royal Netherlands Meteorological Institute (KNMI), Bilt, the Netherlands.
- id
- 4611204
- date added to LUP
- 2014-08-27 09:15:24
- date last changed
- 2014-08-27 09:15:24
@misc{4611204,
abstract = {{For this research, two satellite products were used to see if it was possible to improve the resolution and quality of the global radiation interpolation in the Netherlands. The first data source was from the Climate Monitoring Satellite Application Facility (CM-SAF). The second data source was the Surface Insolation under Clear and Cloudy Skies (SICCS) from the KNMI. Both products were available for the period of January 2006 to December 2011 and came in the form of images with monthly and daily averages. To combine the satellite images with the in- put provided by the KNMI’s 32 measurement stations, these interpolation/merging methods were used: Thin Plate Splines (TPS), Mean Bias interpolation (MB), Interpolated Bias interpolation (IB), Kriging with External Drift; Exponential model (KED-EXP) & Kriging with External Drift; Spherical model (KED-SPH) All these methods made use of the in-situ measurements as main input for the interpolation and all methods except TPS used the satellite products as auxiliary data.
Interpolations were made for the average of the six year period and on monthly measurements, for each month, in each year. Daily interpolations were made for April 2010 until July 2010.
Different validation methods were used to analyze the output. The results showed that; for the six year average both products and all interpolation methods did a good job on predicting global radiation. The R2 was lowest for the IB on the CM-SAF product with a value of 0.19. However, the MAPE (mean absolute percentage error) did not exceed 1.39% on the CM-SAF product and 1.42% on the SICCS product. These values corresponded with an absolute bias of 1.77 W/m2 and 1.8 W/m2. The monthly results showed similar results. The R2 values tended to differ more, especially in the IB and MB interpolation. In most cases this could be explained by the quality of the
Satellite images. The MAPE was low in all cases. A maximum MAPE of 8.38% was found (when using proper satellite images), in November, which corresponded with an absolute bias of ± 4 W/m2. Data splitting returned similar results. MAPE’s did increase up to 9.27% when leaving out 1/4th of the measurement stations but this value corresponds with an absolute bias of 2.71W/m2. These low absolute errors showed that all interpolation methods return an accurate interpolation. However, because the interpolation methods rely on the quality of the satellite images, the SICCS product would be a better product. These images were complete in all months while the CM-SAF product lacked data in December.
Since it turned out that all interpolations performed well, daily data was analyzed for the period of April until July 2010.
It turned out that for the daily data KED and the IB interpolations performed significantly better than the TPS or MB interpolation. The biggest average MAPE was found for the TPS method (10.7% in May). The smallest average errorof0%wasfoundfortheIBmethod. How- ever this method was paired with very low R2 values which made the model unpredictable. The average KED R2 and MAPE ranged from 0.57 to 0.75 and from 0.08% to 0.95 %. This made the method a stable and accurate interpolation method. The satellite images on their own would not be good enough to use directly as a global radiation map, for this time interval. The over- and underestimated bias of the satellite images ranged from -89.63 to 64.49 W/m2. This showed that, a combination of station data and satellite data would improve the quality and resolution of daily global radiation maps.}},
author = {{Van Tiggelen, Jurgen}},
language = {{eng}},
note = {{Student Paper}},
series = {{Student thesis series INES}},
title = {{Assimilation of satellite data and insitu data for the improvement of global radiation maps in the Netherlands}},
year = {{2014}},
}