IMAGE framework/Uncertainties: Difference between revisions

From IMAGE
Jump to navigation Jump to search
No edit summary
(Changed input based on Detlef's comments)
 
(7 intermediate revisions by 3 users not shown)
Line 1: Line 1:
{{FrameworkIntroductionPartTemplate
{{FrameworkIntroductionPartTemplate
|Status=On hold
|Application=Applications;
|Reference=Van Vuuren, 2007; Brinkman et al., 2005;
|PageLabel=Uncertainties
|PageLabel=Uncertainties
|Sequence=7
|Sequence=7
|Reference=Van Vuuren, 2007;
}}
|Description=Obviously, the scientific quality as described above, and treatment of uncertainty are a constant and important concern. While we describe uncertainties and limitations of all IMAGE modules in [[chapters 3-8]], and for the entire model in the [[IMAGE framework summary]], we reflect here on the various aspects of uncertainty, and how they are treated in IMAGE:
<div class="page_standard"><h2>Uncertainties</h2>
* Structural (key) data uncertainty is due to incomplete knowledge of historical time series of model data, for example on energy demand and supply, emissions, and on land use and land-use change. Also other key input data like soil maps, temperature and precipitation maps are uncertain, although continuously improved data sets are prepared. Mostly, this uncertainty is not addressed explicitly, but the best available data are used and eventually harmonized with other modelling teams and partners. Structural, methodological uncertainty (incomplete knowledge of relationships) is still large for many parts of the IMAGE model, e.g. for the impact of Climate change on crop yields, or for local climate change. This uncertainty can be addressed to a certain extent by the possibility to accommodate alternative model formulations, e.g. for crop growth/natural vegetation, carbon cycle, land-use allocation, climate change (via climate sensitivity and temperature/precipitation patterns). While IMAGE strives to include state-of-the art models for all parts of the framework, the size of the uncertainties might rather call for reduced-form models being able to capture the uncertainty range (see further below).
Limitations of IMAGE and critical uncertainties that influence results are described in the [[IMAGE framework summary]] and for each IMAGE component separately in the model component pages (via [[Framework overview]]). Generic aspects of uncertainty in IMAGE are outlined below.
* Another way to address structural uncertainty is by participating in model inter-comparison studies and other multi-model studies, to gain insights on how IMAGE results compare to those from other models. Whether they are well inside the range of outcomes, or at the high or low end, provides information on the functioning of the model and how representative the results are of ranges found in literature.
 
* The overall model uncertainty arising from uncertain processes and data can be assessed by performing more or less systematic sensitivity analyses. This was done among others for the CO2-fertilization factor in crop- and natural vegetation growth and for many parameters of the energy model [[TIMER model|TIMER]] ([[Van Vuuren, 2007]]).
===Structural key data uncertainty===
* Uncertainty in future scenario drivers like population, economic growth, and technology is most commonly addressed by exploring variants around assumed reference pathways,  e.g. high/low variants of population projections, or by assuming contrasting future scenarios (without a specific reference). Likewise,  uncertainty in policy targets and societal trends are addressed by exploring alternative scenarios, varying one or more key input parameters such as learning-by-doing parameters composition of human diets,  or other lifestyle choices.
Structural fundamental data uncertainty is due to incomplete knowledge of historical time series of model data, such as energy demand and supply, emissions, land use, and land-use change. Other critical input data, such as soil maps and temperature and precipitation maps, are uncertain, but data sets are continuously improved. This uncertainty is not addressed explicitly, but the best data available are used and harmonized with other modelling teams and partners.  
* A specific aspect of uncertainty is presented by the level of aggregation, with socio-economic processes being represented by 26 regions, and terrestrial biosphere being modelled at 5 or 30 minute grid cells. At both levels, the region and the grid cell, all behaviour is average behaviour, ignoring all heterogeneity within a region (on income distribution, economy, farming systems, etc), or at a grid cell (on climate, soils, landscape composition). Big differences between countries within a specific world region are masked and all future trends apply to the average, although countries may develop on different pathways. Therefore, interpretation of land use on a country or sub-country level, although possible from a 5 min map must be done with due care.
 
===Structural and methodological uncertainty===
There is structural and methodological uncertainty (incomplete knowledge of relationships) in many parts of the IMAGE framework, for instance, the impact of climate change on crop yields and local climate change. This uncertainty can be addressed to some extent by alternative model formulations, such as for crop growth/natural vegetation, carbon cycle, land-use allocation, climate change (via climate sensitivity and temperature/precipitation patterns). Structural uncertainty can also be addressed in model inter-comparison studies and other multi-model studies to compare IMAGE results with the range of outcomes from other models and results for ranges found in literature and provide information on model functioning (see [[Applications]]. The overall model uncertainty arising from uncertain processes and data can be assessed in systematic sensitivity analyses. This has been done, for example, on the CO<sub>2</sub> fertilisation factor in crop and natural vegetation growth ([[Brinkman et al., 2005]]) and for many parameters of the energy model TIMER ([[Van Vuuren, 2007]]).


===Uncertainty in future scenario drivers===
Uncertainty in future scenario drivers, such as population, economic growth, and technology, is mainly addressed by exploring variants in assumed reference pathways, such as high/low variants of population projections or by assuming contrasting future scenarios. Similarly, uncertainty in policy targets and societal trends is addressed by exploring alternative scenarios, varying one or more key input parameters, such as learning-by-doing parameters, the composition of human diets, and other lifestyle choices.


}}
===Level of aggregation===
A distinct source of uncertainty arises from the level of aggregation, with socio-economic processes represented by 26 regions and the terrestrial biosphere modelled at 5 or 30-minute grid cells. Behaviour modelled at the regional level is mainly described for averages, such as energy supply or food consumption. This does not consider heterogeneity within a region (e.g., in income distribution, economy, farming systems or different countries). For some parameters, socio-economic heterogeneity is considered – different income groups and rural/urban population. Significant differences between countries in a world region are masked, and all future trends apply to the average, although countries may develop along different pathways. Thus, land use on a country or sub-country level is possible on a 5-minute map but must be interpreted cautiously.
</div>

Latest revision as of 17:05, 7 October 2021

Parts of IMAGE framework

  1. Setting the stage
  2. IMAGE 3.0 in a nutshell
  3. Recent model developments
  4. A brief history of IMAGE
  5. Organisational set-up and scientific quality
  6. Uncertainties
Projects/Applications
Models/Databases
Relevant overviews
References


Uncertainties

Limitations of IMAGE and critical uncertainties that influence results are described in the IMAGE framework summary and for each IMAGE component separately in the model component pages (via Framework overview). Generic aspects of uncertainty in IMAGE are outlined below.

Structural key data uncertainty

Structural fundamental data uncertainty is due to incomplete knowledge of historical time series of model data, such as energy demand and supply, emissions, land use, and land-use change. Other critical input data, such as soil maps and temperature and precipitation maps, are uncertain, but data sets are continuously improved. This uncertainty is not addressed explicitly, but the best data available are used and harmonized with other modelling teams and partners.

Structural and methodological uncertainty

There is structural and methodological uncertainty (incomplete knowledge of relationships) in many parts of the IMAGE framework, for instance, the impact of climate change on crop yields and local climate change. This uncertainty can be addressed to some extent by alternative model formulations, such as for crop growth/natural vegetation, carbon cycle, land-use allocation, climate change (via climate sensitivity and temperature/precipitation patterns). Structural uncertainty can also be addressed in model inter-comparison studies and other multi-model studies to compare IMAGE results with the range of outcomes from other models and results for ranges found in literature and provide information on model functioning (see Applications. The overall model uncertainty arising from uncertain processes and data can be assessed in systematic sensitivity analyses. This has been done, for example, on the CO2 fertilisation factor in crop and natural vegetation growth (Brinkman et al., 2005) and for many parameters of the energy model TIMER (Van Vuuren, 2007).

Uncertainty in future scenario drivers

Uncertainty in future scenario drivers, such as population, economic growth, and technology, is mainly addressed by exploring variants in assumed reference pathways, such as high/low variants of population projections or by assuming contrasting future scenarios. Similarly, uncertainty in policy targets and societal trends is addressed by exploring alternative scenarios, varying one or more key input parameters, such as learning-by-doing parameters, the composition of human diets, and other lifestyle choices.

Level of aggregation

A distinct source of uncertainty arises from the level of aggregation, with socio-economic processes represented by 26 regions and the terrestrial biosphere modelled at 5 or 30-minute grid cells. Behaviour modelled at the regional level is mainly described for averages, such as energy supply or food consumption. This does not consider heterogeneity within a region (e.g., in income distribution, economy, farming systems or different countries). For some parameters, socio-economic heterogeneity is considered – different income groups and rural/urban population. Significant differences between countries in a world region are masked, and all future trends apply to the average, although countries may develop along different pathways. Thus, land use on a country or sub-country level is possible on a 5-minute map but must be interpreted cautiously.

Retrieved from "https://models.pbl.nl/index.php?title=IMAGE_framework/Uncertainties&oldid=36606"