the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 Jan 2026
Local weather scenarios for soil and crop models: a simple generator based on historic data sampling
Abstract. Weather scenarios are for example required to model future agricultural production and the development of soil properties under climate change. These scenarios should realistically depict regional weather conditions at a daily resolution for the expected climate development. In this technical note, we present the LocalWeatherSampler (LWS) for generating mid-term weather scenarios (20–30 years) for specific regions or locations based on historically recorded weather data. It is demonstrated for an example site in Germany. The core idea is to define wet or dry years and to increase their abundance in future years via a random sampling from history. A temperature trend based on common climate projections can be added afterwards. For the definition of dry/wet years, two different methods are implemented. The historical weather data can be either divided manually into a pool of wet (or dry) years or based on the Standardized Precipitation Index (SPI). By varying the threshold value for wet (dry) years and their probability of appearance within the scenario, the framework allows for the generation of scenarios tailored to specific requirements, such as sequences characterized by extremely dry years or by moderately dry years, as well as extremely wet future sequences. This approach is designed to test or analyze future scenarios of precipitation regimes and temperature trends using models that require realistic daily weather data, such as soil, crop, or hydrological models.
- Preprint
(1226 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2025-6173', Anonymous Referee #1, 24 Feb 2026
-
CC1: 'Reply on RC1', Sara König, 03 Mar 2026
Dear Reviewer, thanks for reviewing our manuscript.
You are right, the presented weather generator is based on simple R methods. It was developed, as mentioned in the manuscript, to generate weather scenarios with daily data which include realistic weather extremes such as droughts fo any specific location (i.e. arable field). We need such data as input for soil-crop modeling, which is our expertise. As we are no experts in climate modelling, we did intense literature research and discussed with various colleagues with expertise in climate modelling (e.g. from DWD and Herion), and had to conclude that there is no tool available serving our purpose. This short technical note is thought to provide a simple solution for this gap that might help others. If you are aware of a similar tool, we are more than happy to include it to our work.
Regarding the statistics: could you please elaborate more on which kind of statistical test you have in mind? Note that we do not aim to predict climate, but to generate site-specific realistic weather dynamics.Citation: https://doi.org/10.5194/egusphere-2025-6173-CC1 -
AC1: 'Reply on RC1', Sara König, 14 Apr 2026
Thanks for reviewing our manuscript. Please see our response to your specific comments below:
“Reviewing of the manuscript titled ‘Local weather scenarios for soil and crop models: a simple generator based on historic data sampling’ submitted to the discussion of Geoscientific Model Development (Manuscript Number: egusphere-2025-6173). The authors develop a simple generator based on the historic precipitation data for the local weather scenarios (e.g., wet or dry). The time period is long and fine (e.g., 1993-2022 or historic weather and 2023-2052 for projections). The manuscript is overly simple and written in a study that is not strict and not including sufficient literature review and discussion. In my opinion, this is a simple formulation using the R statistics for the data analysis and case study, which lacks strong novelty to fill the research gaps from previous studies.”
You are right, the presented weather generator is based on simple R methods. It was developed, as mentioned in the manuscript, to generate weather scenarios with daily data which include realistic weather extremes such as droughts for any specific location (i.e. arable field). We need such data as input for soil-crop modeling, which is our expertise. As we are no experts in climate modelling, we did intense literature research and discussed with various colleagues with expertise in climate modelling (e.g. from DWD and Herion), and had to conclude that there is no tool available serving our purpose. This short technical note is thought to provide a simple solution for this gap that might help others. During our research, we learned from some colleagues that they are facing exactly the same problem. If you are aware of a similar tool, we are more than happy to include it to our work.
“There are also a couple of specific comments. For example, (1) in Figure 1, the reviewer is confused that why in step 4 for Future Scenarios the selected years still 2001-2021?”
Figure 1 serves to illustrate the concept. Here, we used the years 2001 to 2011 for historic weather data and generated future scenarios for the years 2012 to 2021. Shown in step 4 is the whole time series to illustrate how historic data is used for the future scenarios. This is why after 2011 years are written under the x-axis, i.e. for the last year in the time series, the data of the year 2001 were sampled.
“ (2) the paragraph should be line well instead of randomly being located, such as in Section 2.1; (3) Is there any statistics to evaluate the validity and accuracy of the proposed model or method? The authors need substantial works to include more details in each section.”
We are not clear what kind of statistical tests are being referred to here. Note that we do not aim to predict climate, but to generate site-specific realistic weather dynamics.
Citation: https://doi.org/10.5194/egusphere-2025-6173-AC1
-
CC1: 'Reply on RC1', Sara König, 03 Mar 2026
-
RC2: 'Comment on egusphere-2025-6173', Anonymous Referee #2, 13 Mar 2026
This technical note provides a description of a simple and user-friendly applicable code to distribute weather scenarios for a pre-determined duration. At first, pools of drier and wetter years will be set on which randomly precipitation and temperature scenarios will be set thereafter. It is very good that the scripts are published. In general, it is a very required tool, as the authors mention the need for such data sets for soil–crop models. However, the manuscript stresses some climatically terms too much while no definitions are provided.
I recommend avoiding the terms ‘dry’ and ‘wet’ and rather using them in a relative sense, for example as ‘drier’ and “wetter” or as ‘relatively dry’ and ‘comparatively wet’. E.g., an area with very high amounts of precipitation (e.g., >1500 mm) and a threshold of 0.8 wouldn’t make the pool of “drier” years “dry”. The same applies to the word arid. Often, aridity is defined as higher (evapo)transpiration than precipitation. If this is not your case, don’t use the term “arid” throughout the manuscript unless you provide a proper definition of how you define it.
Why is only temperature corrected according to RCP scenarios? If temperature is adjusted to trends, shouldn’t precipitation scenarios also have a trend?
If this technical note creates data sets for precipitation and temperature, why is this not also done for solar radiation when you want to use such scenarios for soil–crop modelling? For most of these models this weather information is totally mandatory. Please explain very briefly how that was achieved. If not, explain briefly why it is not important for this study case. Otherwise, don’t mention the models that require “solar radiation”.
Figure 2 and Table 2: Align your naming of the scenarios.
L73 – Robust results? How is robustness determined?
L113 – Your tool doesn’t control “aridity”!
L125 and L140 – Explain how the manual setting can be done or what the difference is compared to “manually” selecting an RCP scenario. It is always a “manual” selection. What is the actual difference between the selection modes.
Citation: https://doi.org/10.5194/egusphere-2025-6173-RC2 -
AC2: 'Reply on RC2', Sara König, 14 Apr 2026
Thanks for reviewing our manuscript. Please find the responds to your specific comments below:
“This technical note provides a description of a simple and user-friendly applicable code to distribute weather scenarios for a pre-determined duration. At first, pools of drier and wetter years will be set on which randomly precipitation and temperature scenarios will be set thereafter. It is very good that the scripts are published. In general, it is a very required tool, as the authors mention the need for such data sets for soil–crop models. However, the manuscript stresses some climatically terms too much while no definitions are provided.”
Thanks a lot!
“I recommend avoiding the terms ‘dry’ and ‘wet’ and rather using them in a relative sense, for example as ‘drier’ and “wetter” or as ‘relatively dry’ and ‘comparatively wet’. E.g., an area with very high amounts of precipitation (e.g., >1500 mm) and a threshold of 0.8 wouldn’t make the pool of “drier” years “dry”. The same applies to the word arid. Often, aridity is defined as higher (evapo)transpiration than precipitation. If this is not your case, don’t use the term “arid” throughout the manuscript unless you provide a proper definition of how you define it.”
You are right that relative terms for describing the precipitation scenarios are more correct in this regard, we will adapt this throughout the manuscript.
“Why is only temperature corrected according to RCP scenarios? If temperature is adjusted to trends, shouldn’t precipitation scenarios also have a trend?”
Of course, the RCP scenarios also generally have trends for precipitation, i.e. higher amount of heavy rain fall events in the 8.5 scenario. However, this is not reflected in the projections for specific sites, what we also illustrate with Figure 2. More general trends like a total increase in precipitation could be included easily following the same approach as for temperatures. However, in the time frame we are considering for the simulations these changes are comparatively minor. For this reason, we only use temperature trends from RCP scenarios, and manipulate the precipitation with the stochastic sampling approach.
“If this technical note creates data sets for precipitation and temperature, why is this not also done for solar radiation when you want to use such scenarios for soil–crop modelling? For most of these models this weather information is totally mandatory. Please explain very briefly how that was achieved. If not, explain briefly why it is not important for this study case. Otherwise, don’t mention the models that require “solar radiation”.”
Yes, we agree, radiation is relevant for any soil-crop model. The short-wave radiation is relevant for photosynthesis and the radiation balance is relevant for potential evapotranspiration. Both are used in our modeling. However, this data is only available for a couple of stations in Germany. If it is given, we use it. If not, we fall back to sunshine duration and geographical position for the estimation of the radiation components. If sunshine duration is also not given, we use the difference in minimum and maximum temperature together with geographical location for the calculation. This estimation scheme is according to Allen et al. (1998). We have checked these relationships for German weather stations and found a good agreement. We will add a short paragraph to the manuscript to describe this approach.
“Figure 2 and Table 2: Align your naming of the scenarios.
L73 – Robust results? How is robustness determined?
L113 – Your tool doesn’t control “aridity”!”
Thanks for these minor comments, we will clarify this in the revised manuscript.
“L125 and L140 – Explain how the manual setting can be done or what the difference is compared to “manually” selecting an RCP scenario. It is always a “manual” selection. What is the actual difference between the selection modes.”
The difference basically is that you can either just select an RCP scenario, and the tool automatically picks the trend for your coordinates, or you can manually set a trend such as +1 degree. We will clarify this in the revised manuscript.
Citation: https://doi.org/10.5194/egusphere-2025-6173-AC2
-
AC2: 'Reply on RC2', Sara König, 14 Apr 2026
Model code and software
R script: Local weather scenarios for soil and crop models: a simple generator based on historic data sampling Stefan Anton Albert Gasser et al. https://doi.org/10.5281/zenodo.17511186
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 261 | 108 | 29 | 398 | 15 | 40 |
- HTML: 261
- PDF: 108
- XML: 29
- Total: 398
- BibTeX: 15
- EndNote: 40
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Reviewing of the manuscript titled ‘Local weather scenarios for soil and crop models: a simple generator based on historic data sampling’ submitted to the discussion of Geoscientific Model Development (Manuscript Number: egusphere-2025-6173). The authors develop a simple generator based on the historic precipitation data for the local weather scenarios (e.g., wet or dry). The time period is long and fine (e.g., 1993-2022 or historic weather and 2023-2052 for projections). The manuscript is overly simple and written in a study that is not strict and not including sufficient literature review and discussion. In my opinion, this is a simple formulation using the R statistics for the data analysis and case study, which lacks strong novelty to fill the research gaps from previous studies. There are also a couple of specific comments. For example, (1) in Figure 1, the reviewer is confused that why in step 4 for Future Scenarios the selected years still 2001-2021? (2) the paragraph should be line well instead of randomly being located, such as in Section 2.1; (3) Is there any statistics to evaluate the validity and accuracy of the proposed model or method? The authors need substantial works to include more details in each section.