Accueil » CG landscapes will get even more realistic, thanks to ecoclimates #SIGGRAPH2022

CG landscapes will get even more realistic, thanks to ecoclimates #SIGGRAPH2022

This article is available in: French

Every year at SIGGRAPH, researchers showcase their latest computer graphics and interactive techniques projects. This is the perfect place to get a glimpse of things to come in the CG industry, and to have a better idea of what tools artists could use in the future.

Ecoclimates: Climate-Response Modeling of Vegetation is a good example of this. This paper by Wojtek Pałubicki, Miłosz Makowski, Weronika Gajda, Torsten Hädrich, Dominik L. Michels, and Sören Pirk focuses (as the title implies) on vegetation simulation, a field that allows us to create realistic landscapes by simulating how plants grow and propagate. Here are a few details about the paper, as well as some examples created by the team, their video presentation and a few ideas for future research.

One of the simulations included in the paper

Ecoclimates: the key to realistic modeling of vegetation?

The key idea behind the paper is to focus on ecoclimates, in other words the link between plant ecosystems and weather. For example, trees in a forest sweat and create humidity, they will receive more rainfall but less light if they are located at the bottom of a slope and not on the top of a hill. Such factors will, of course, shape how the vegetation evolves and grows day after day, year after year. Some plant species may be disadvantaged in a specific area while others whill thrive.

Without getting too technical (the full paper will give you more details if needed), here is an overview of the process used by the team:

  • A user provides input: ” set of plant species, a digital elevation model and data describing macroclimatic variation over time.”
  • the ecoclimate is then used, with three sub-models: vegetation, soil, weather. There is of course some feedback between those three models. It should also be noted that two timescales are used: one for vegetation and water (timescale of one month) and one for weather (10 to 60s). This makes sense since weather can change quickly, while the size of a tree won’t really change over a few seconds or even days.
  • The output is therefore the simulated landscape, with an evolving vegetation and weather.

It should be highlighted that since the evolutions of the climate over time is user-defined, this model can be used to evaluate vegetation response to climate change in a specific area, for example using different climate change scenarios for a forest located in the South of France.

Results: what can be achieved with this technique?

Here are a few results shared by the researchers.

  • A virtual deforestation experiment. As the paper explains: “Our method models the feedback between vegetation, soil, and weather. To illustrate this, we conduct a deforestation experiment while keeping the macroclimate in our weather model constant. In (a) we show a tropical rainforest with cumulus clouds. In (b) we remove a large portion of the rainforest thereby modifying the vapor emission from the terrain. Consequently, fewer cumulus clouds form, especially over the deforested area. After continuing ecosystem growth cloud formation increases slightly (c). Only after significant portions of the rainforest have regrown, cumulus cloud formation is restored (d).”
  • This comparison between real-life locations (arid ecosystem in Niger, peatlands in Siberia) and landscape simulated using the proposed method shows that realistic yet complex vegetation patterns can be achieved.
  • A tropical forest (with a large annual rainfall) reacting to continued decrease precipitation over the years.
  • Here is the video presentation of the project, with additional examples and simulations:

Conclusion & future research

The research teams shows that you can simulate ecoclimates using models for vegetation, soil, and weather, as well as the feedback between these three key elements. Focusing on microclimates (local climatic variation such as the fact that there is more light, less humidity at the edge of a forest than inside it) allows the researchers to take into account small scale effects that can have a huge impact on the overall appearance of a landscape. This method will be very useful to get a better understanding on how climate change can affect vegetation over time in a specific location. Another strength of this method is the fact that it doesn’t need that much computing power: to create the results shown in this article, and using an Intel Core i5, 4 x 2.5GHz, 6Go RAM, NVIDIA Geforce GTX 1050, the computation time of one simulation year is only 0.6 to 8 seconds. In other words, this approach is fast enough to be used in vegetation simulation tools for VFX, or even in videogames depending on the use case. It should also be noted that in such use cases, since the simulation also provides weather data, you could use this data to display clouds, or tweak it for aesthetic purposes.

This approach does have limitations, however. For example, the diurnal cycle is not taken into account (temperatures don’t drop and rise each day), neither is the fact that there are variations of directional lighting between summer and winter. These are some of the ideas listed by the team to improve this ecoclimate model. The paper also suggest to “extend the method for testing ecoclimate hypotheses and validating these findings with other analytical and empirical observations” or to combine the method with models of the pedosphere (the outermost layer of the Earth) and cryosphere (areas where solid water can be found, such as glaciers, permafrost, frozen lakes).

For more information, you can check out the full paper. Ecoclimates: Climate-Response Modeling of Vegetation is a reseach project by Wojtek Pałubicki, Miłosz Makowski, Weronika Gajda, Torsten Hädrich, Dominik L. Michels and Sören Pirk from Adam Mickiewicz University (Poland), KAUST (Saudi Arabia) and Adobe Research.

Laissez un commentaire

A Lire également