CC BY
0
DOI 10.18551/rjoas.2020-08.16
VALIDATION AND UTILIZATION OF ECMWF RAINFALL DATA MODEL TO EVALUATE CHANGES OF THE OLDEMAN CLIMATE TYPE IN LOMBOK
Baihaqi Anas*, Posgraduate Student Dryland Resources Management Program, University of Mataram, West Nusa Tenggara,
Indonesia
Kusnarta I Gusti Made, Yasin Ismail
Department of Soil Science, Faculty of Agriculture, University of Mataram, West Nusa Tenggara, Indonesia
*E-mail: baihaqianas89@gmail.com
ABSTRACT
This study aims for validating and correcting the ERA-Interim ECMWF (European Centre for Medium-Range Weather Forecasts) model data to change observational data that has a short range in the analysis of the changes of the Oldeman climate type in Lombok in the last 40 years. The validation method that has been used in this research is Double-Mass Curves (DMC) by correcting the slope that occurs on the ECMWF rainfall accumulation curve with its observations. The comparing of RMSE values and correlations that occur both before and after the validation process, are used to test the reliability level of the validation method. The results show that there are positive and strong correlations between ECMWF rainfall data and observational data with values ranged from 0.764 to 0.904. RMSE values decreased significantly from 110.1 to 85.4, In general, Lombok Island can be divided into E, D, C and B according to Oldeman climate type. Changes mostly occur in the eastern coastal area which turned wetter (from climate types E to D), the west coast area which became drier (from climate type D to E), and the highlands in the north of the island where the climate type became wetter (from climate type C to B). From those results, we can conclude that the DMC method is highly recommended to be used to validate and correct the ECMWF model data. This also shows that the ECMWF rainfall data is suitable to replace the observational rainfall data for the purpose of climate change analysis after going through a validation and correction process.
KEY WORDS
Validation, utilization, ECMWF, Oldeman, double-mass curves.
Oldeman climate type is a method to determine cropping patterns following the average of monthly rainfall and water requirements of rice and secondary crops in the tropics areas as a reference (Fadholi & Supriatin, 2012; Sasminto & Sutanhaji, 2014). Cropping patterns and its changes according to the trend of climate change (Nugroho & Nuraini, 2016) are strongly important things that have to be considered in dryland farming, especially in Lombok.
In evaluating and modeling climate change requires the availability of long weather data range (Bracho-Mujica et al., 2019). But not every single location in Lombok is represented by rainfall observation data. Even though some locations have records of rainfall data, only a few of them have long range data to be analyzed. Therefore the usage of non-observation data can be considered (ex. satellite or model).
TRMM satellite images have a long rainfall data series. Nevertheless, it has a low grid resolution (0.25o x 0.25o) which requires a downscaling method before the data can be processed (Zhang et al., 2018). On the other hand, the ERA-Interim dataset of ECMWF has a high-resolution model (0.125o x 0.125o) and has a long-term reanalysis rainfall data that available from 1979 to the present. Therefore these data are easy to download, widely used,
and recommended for monitoring climate change in a place that has a short range of observational data (Cerlini et al., 2017).
However, a data validation method is still needed to determine the reliability of the ECMWF model data before being used for analysis purposes (Eker et al., 2019). This study aims for validating and correcting rainfall data model compared to observational rainfall data. Correction factors obtained from the validation process are expected to produce reanalysis rainfall data values that are close to the value of observational rainfall data.
By using the validated ECMWF rainfall data model which has values those are close to the observation data, we are no longer limited by the short-ranged observational rainfall data in the analysis of climate change. So that in the future, even using model data, changes in cropping patterns based on Oldeman climate type in Lombok can still be analyzed comprehensively close to the actual conditions.
METHODS OF RESEARCH
The data that used in this study are observational rainfall data and model rainfall data, including: 1) Monthly rainfall data from 8 samples of rain observation stations on Lombok Island with variable length data periods and elevation (Table 1); 2) Rainfall reanalysis model data from ECMWF downloaded in the form of *.nc file format from the ECMWF web https://apps.ecmwf.int/datasets/data/interim-full-mnth/levtype=sfc/ with 'Select time' 00:00:00 and 12:00:00, 'Select step' 12, and 'Total Precipitation' for the 'Select parameter' option. The *nc formatted data was exported to *.xls by OpenGrADS to simplify the analysis process. For validation, the amount of exported data is 8 points following the same coordinates as the rain observation station.
Double-Mass Curves (DMC) analysis is used to justify and bring the value of ECMWF rainfall closer to rainfall observations. Before applying the DMC analysis, the length of the ECMWF series data was adjusted to the length of each observed rainfall data. The analysis is done by equating the slope of the ECMWF curve with the slope of the observation curve using a correction value (k). The k value that has been generated will be used to correct ECMWF data from 1979 to 2017. DMC is analyzed by the Excel program.
The reliability level of ECMWF data both before and after corrected by DMC is examined by calculating RMSE (Root Mean Square Error) and Spearman correlation (r) by using the Excel program. Climate type classifying and spatial zoning base on the Oldeman climate type is calculated and mapped by using Excel program and QGIS 2.8 in 4 periods: 1979 - 1988, 1989 - 1998, 1999 - 2008 and 2009 - 2017 after the validation process was carried out. The changes in the Oldeman climate type are reviewed spatially from these four periods.
RESULTS AND DISCUSSION
Data Series Validation and Correction of ECMWF Rainfall against its Observation. The DMC between the accumulation of pre-adjusted ECMWF rainfall data with observations shows that the error between the two data is highly significant, especially for Batukliang Utara, Pringgabaya, Jerowaru and Sembalun station. It can be seen from the difference in slope and width of the angle formed between the two curves (Figure 1).
On the other hand, figure 1 also shows that several stations such as Ampenan, Pringgabaya, and Jerowaru have a higher (over-estimated) ECMWF curve slope compared to the observation curve. Base on the elevation of these stations, it is concluded that generally, it occurs at stations that located in lowland areas (with elevations below 100 m)
In the adjustment process, each pre-adjusted ECMWF rainfall data in each station is multiplied by its correction factor (k) that ranged from 0.384037 to 1.54371 (Table 2), so that the accumulation value is close to or equal to the accumulated value of the observed rainfall data. Changes that have been applied to the pre-adjusted ECMWF rainfall data make the curve coincide (have the same slope) with the accumulated rainfall observation curve (Figure 2). This altered ECMWF rainfall value refers to as the Adjusted ECMWF.
Table 1 - Rain Observation Stations on Lombok Island
Nu. Station Data Period Elevation (M)
1 Ampenan 1995 - 2017 1S
2 Sekotong 2009 - 2017 14
3 Praya 1995 - 2017 124
4 Batukliang Utara 2009 - 2017 346
б Sikur 1995 - 2017 325
6 Pringgabaya 2009 - 2017 20
7 Jerowaru 2009 - 2017 15
S Sembalun 2009 - 2017 1151
Table 2 - RMSE and Correlation value (r) monthly data series of Pre-Adjusted ECMWF and Adjusted
ECMWF vs Observation rainfall data
Station k RMSE (vs Obs) Correlation (vs Obs)
Pre-Adj ECMWF Adj ECMWF Status Pre-Adj ECMWF Adj ECMWF Status
Ampenan 0.857532 88.4 79.6 better 0.764 0.764 fair
Praya 1.321103 111.7 10S.6 better 0.846 0.S47 better
Sekotong 1.146823 67.5 66.3 better 0.904 0.904 fair
Batukliang Utara 1.54371 140.7 124.2 better 0.778 0.77S fair
Sikur 1.093732 90.7 90.2 better 0.866 0.S71 better
Pringgabaya 0.384037 136.2 47.7 better 0.775 0.775 fair
Jerowaru 0.544715 113.4 52.2 better 0.818 0.S39 better
Sembalun 1.272636 132.2 114.6 better 0.853 0.S53 fair
Average 110.1 S5.4 0.826 0.S29
Table 3 - Changes in the oldeman climate type at 8 stations based on the 10-year period
_ . Oldeman Climate Type_
S n_1979 - 1988_1989 - 1998_1999 - 2008_2009 - 2017
Ampenan D3 D3 E E
Praya C3 D3 C3 C3
Sekotong C3 D3 D3 D3
Batukliang Utara C2 C3 C3 B2
Sikur C3 D3 C3 D3
Pringgabaya E D3 D3 D3
Jerowaru E E E E
Sembalun C3 C3 C3 D3
-Accum_Obs -Accum_ECMWF
Figure 1 - DMC of Pre-Adjusted ECMWF and Observation rainfall data
The average correlation values (r) between the ECMWF rainfall data series against its observations are 0.826 (pre-adjusted) and 0.829 (adjusted) with correlation values for each station varying between 0.764 to 0.904. The analysis result indicates that there is a strong and positive correlation between ECMWF rainfall data and its observation both before and after adjusting. The small difference in the correlation values also confirms that there are no significant changes from the ECMWF rainfall patterns in the adjustment process.
However, the decrease of the RMSE value that occurs after the adjustment process shows that the method has a robust impact to bring the ECMWF value closer to its observations. Before the adjustments of ECMWF rainfall data, the RMSE value ranges from 67.5 to 140.7 mm/month with 110.1 mm/month (3.7 mm/day) as its average value. After adjustments, the range of RMSE value decreases between 47.7 to 124.2 mm/month with 85.4 mm/month (2.8 mm/day) as its average value.
Figure 2 - DMC of Adjusted ECMWF and Observation rainfall data
Figure 3 - Maps of changes in Oldeman climate type in Lombok (a) 1980, (b) 2005 based on the ECMWF correction data
Figure 4 - Maps of changes in Oldeman climate type in Lombok (a) 1980, (b) 2005 based on the latest research (Asy-Syakur et al., 2011)
Figure 5 - Maps of changes in Oldeman climate type in Lombok (a) 1979 - 1988, (b) 1989 - 1998, (c) 1999 - 2008, and (d) 2009 - 2017
The overall results show that the validation and adjustment process of ECMWF rainfall data from 8 stations were able to produce better RMSE values. It explains that this method is very well applied to correct the eCmWF model data series related to the usage of this model data for further climate analysis purposes.
Spatial Analysis Validation on its Observation. Spatial analysis as a result of the validation and correction process approaches the results of previous studies (Figures 4) conducted by Asy-Syakur et al., (2011) with some differences. In general, the spatial distribution of Oldeman climate types in 1980 obtained from the correction of ECMWF data (Figure 3a) has a lot of agreement with the climate type patterns in Figure 4a from the latest research. It can be seen from both maps that climate type E is spread on the east coast, climate type C is located in the highlands in the middle of the island, and type D covers most areas on the north coast and lowlands. The discrepancy can only be seen in the difference in the area of climate type C which is located in the northern plateau of the island. This is due to the rain observation stations operating in 1980 (which is used by Asy Syakur et al., 2011) were not as many as those operating in 2005. So the distribution of Oldeman climate types described cannot represent the real climate condition.
More similarity was found in the comparison of the spatial distribution of Oldeman climate types in 2005 obtained from the correction of ECMWF data (Figure 3b) with the climate type patterns in Figure 4b which is based on observational rain data. The distribution of climate types E appears to dominate on the east and south coasts of both maps. The climate type D, especially D4, is located in the lowlands that are in the southern part of the island. While the D3 climate type is spread in the middle plains which are located around the highlands.
The difference is only seen in the distribution of climate types found in the highlands. Variations in climate types ranging from C2, B2 to B1 are shown in Figure 4b. Whereas what appears in Figure 4a is only the uniformity of the C3 climate type. This difference is based on the subjective determination of climate types C2, B2 to B1 at that location by the researcher. Basically, these locations are not represented by rainfall observation stations.
Based on the satisfactory results of the validation that has been produced, the correction value (k) is used confidently for adjusting the ECMWF rainfall data at each station from 1979 to 2017 and considered as representative data representing the ground station
data. Adjustments to ECMWF data were followed by dividing the time series into 10-year periods (1979 - 1988, 1989 - 1998, 1999 - 2008, and 2009 - 2017) to be mapped and analyzed spatially.
Spatial Analysis of Changes in the Oldeman Climate Type. Table 3 shows the Oldeman climate type in Lombok ranged between C, D, and E in each decade. In more detail, a number of changes have taken place in several regions represented by 8 stations. 3 of them have changed to become drier, 2 of them became wetter, 1 location has changed fluctuatively, and there are 2 locations that have not changed until the fourth decade.
Ampenan, Sekotong, and Sembalun are areas whose climate types became drier. The regions are changed from climate types D3 to E, C3 to D3, and C3 to D3, respectively. On the other hand, North Batukliang and Pringgabaya are areas whose climate type became wetter. The two regions are changed from climate types C3 to B2 and E to D3, respectively. Sikur is a region that experienced fluctuating climate type changes from C3 to D3, C3 and back to D3. While Jerowaru and Praya are 2 regions whose climate types did not change in the last 40 years with climate types C3 and E, respectively.
Spatial analysis results (Figure 5) show that in general, climate type C which includes C2 and C3, dominates in the northern part of Lombok Island where the area is a plateau region as a result of orographic effects (Jiang, 2003). Areas with climate type D, which include D3 and D4, are generally scattered in the lowlands on the southern part of Lombok Island and some areas on the north coast. Whereas climate type E generally dominates the southern and eastern coastal regions.
Changes in climate types in each period are seen in the northern part of Central Lombok Regency which changes from climate type C3 to B2 (became wetter), east coast changing from climate types C3 and D3 to E (became drier), and west coast changing from climate types E to D4 and D3. Minor changes have occurred in the central part of the Central Lombok Regency which has experienced a change between C3 and D3 climate types. Nevertheless, the northern and southern coastal areas tend to be stagnant and do not experience significant climate type changes.
CONCLUSION
From those results, we can conclude that: 1) There are positive and strong correlations between ECMWF rainfall data and observational data with values ranging from 0.764 to
0.904. RMSE values decreased significantly from 110.1 to 85.4; 2) In general, Lombok Island can be divided into several climate zones according to the Oldeman climate type, namely E, D, C and B. Changes in Oldeman climate type mostly occur in the eastern coastal area which turns wetter (from climate types E to D), the west coast area which has become drier (from climate type D to E), and the highlands in the north of the island where the climate type has become wetter (from climate type C to B).
They indicated that the DMC method is highly recommended to be used to validate and correct the ECMWF model data. This also shows that the ECMWF rainfall data is suitable to replace the observational rainfall data for the purpose of climate change analysis after going through a validation and correction process.
REFERENCES
1. Asy-Syakur, A. R., Nuarsa, I. W., & Sunarta, I. N. (2011). Pemutakhiran peta agroklimat klasifikasi oldeman di pulau lombok dengan aplikasi sistem informasi geografi. https://www.researchgate.net/publication/303457934.
2. Bracho-Mujica, G., Hayman, P. T., & Ostendorf, B. (2019). Modelling long-term risk profiles of wheat grain yield with limited climate data. Agricultural Systems, 173, 393402. doi:10.1016/j.agsy.2019.03.010.
3. Cerlini, P. B., Meniconi, S., & Brunone, B. (2017). Groundwater supply and climate change management by means of global atmospheric datasets. Preliminary results. Procedia Engineering, 186, 420-427. doi:10.1016/j.proeng.2017. 03.245.
4. Eker, S., Rovenskaya, E., Langan, S., & Obersteiner, M. (2019). Model validation: A bibliometric analysis of the literature. Environmental Modelling & Software, 117, 43-54. doi:10.1016/j.envsoft.2019.03.009.
5. Fadholi, A., & Supriatin, D. (2012). Sistem pola tanam di wilayah priangan berdasakan klasifikasi iklim oldeman. Jurnal Geografi Gea, 12(2).
6. Jiang, Q. (2003). Moist dynamics and orographic precipitation. Tellus A: Dynamic Meteorology and Oceanography, 55(4), 301-316.
7. Nugroho, B. D. A., & Nuraini, L. (2016). Cropping pattern scenario based on global climate indices and rainfall in banyumas district, central java, indonesia. Agriculture and Agricultural Science Procedia, 9, 54-63. doi:10.1016/j.aaspro.2016.02.124.
8. Sasminto, R. A., & Sutanhaji, A. T. (2014). Analisis spasial penentuan iklim menurut klasifikasi schmidt-ferguson dan oldeman di kabupaten ponorogo. Jurnal Sumber Daya Alam dan Lingkungan, 1(1), 51-56.
9. Zhang, T., Li, B., Yuan, Y., Gao, X., Sun, Q., Xu, L., & Jiang, Y. (2018). Spatial downscaling of TRMM precipitation data considering the impacts of macro-geographical factors and local elevation in the Three-River Headwaters Region. Remote Sensing of Environment, 215, 109-127. doi:10.1016/j.rse.2018.06.004.
