The application of Machine Learning in the field of sustainability, more specifically in the development of predictions in ecology, is experiencing its moment of splendor, and the Estación BiolorDo Park MagicñAna is very closely linked to it.
Thanks to a group of Spanish researchers and the development of a model based on this technology, it is already possible to predict how they will evolve herb populationstoceas of a certain piece of land, using climatological data and soil properties, with the aim of estimating the direct effects of climate changetotico in the different greenhouse gas emission reduction scenarios.
“Digitization is directly helping the scientific progress and the conservation of the environment. Doandeach time all temperature or precipitation records are madeiThey are at hand in a few locations today, and thanks to technologies such as Machine Learning, we can continuously monitor these magnitudes at millions of points. The higher the quality of the data, the better it workston the models that we can build with them”, explains Javier Garcito Algarra, academic directorandmonkey of toEngineering Areaiay Sciences at the U-tad University Center.
Machine Learning in Doñana Park
herb plantstoceas are essential to maintain the biodiversity of an environment as varied as the Doñana National Park. They provide sustenance and shelter for a multitude of animals that spend part of the year in the marshes and serve to fix the soil in an area subject to periodic flooding.you say
The different herb speciestoceas compete with each otheri for the dissolved nutrients in a very saline environment, and the abundance of one of them can be detrimental to others, although sometimes it is alsoandn can be beneficial. The disappearance of a species as a result of climate changetodude, coulditrigger a phenomenon known as ‘extinctioncascade‘of AHi that it is very important to predict if the alteration of the climatic conditionstoticas can significantly vary the species present (fauna and flora) in the marsh.
Machine learning allows the monitoring of more than twenty species of plants in a hypersaline environment in the Northeast of Doñana Park
This is where the machine learning, how well Javier Garc explainsito Algarra: “The essence of machine learning is to build a functionn mathethics based on data available to date and to be able to make predictions with it. The study of the consequences of climate changetotico is a natural application of these tandtechniques and can be a very helpful tool”.
In the case of the Biological Station, this is achieved by monitoring more than twenty species of plants from the hypersaline environment of the Los Caracoles marsh, in the northeast of Doñana Park.
From the U-tad University Center explain that they have used a two-step model to predict abundance when change the temperature and dissolved minerals in the soil. In this way, it is possible to predict, even with an accuracy level of 85% cutoMany specimens of each species coexistton under certain conditions.
Said model has uncovered the great impediments that emerge with this type of study, since obtaining abundant data, in this case, on herbaceous species in the field, lasted a total of five seasons, one per year. To get an idea, the academic director has revealed that the CSIC team, “He had to carry out individual counts in a marsh area where these species reproduce. The construction and verification of the model ledor oneyear of additional work”
One of the drawbacks that researchers have encountered throughout the process is the quality and representativeness of the data. “Although it is a job carried out by a very professional team, until the task is completed you cannot know if the models are going to be better than the previous ones available.ito. It is a bit like cooking, you have good ingredients and experience inventing recipes, but until you have finished the work you cannot be sure if you have reached the point oroptimal“, Explain.
So, with all these problems, how has the quality of predictions been improved? García comments that the problem lies in the fact that the prediction is very complicated, “because in order to knowtoMany plants of species A will haveto next year, I need to know those of species B, C, etc. To know those of species B, alsoandn needinamely those of A, is a circular condition. What we did was make a first prediction using very few variables, which breaks that condition. That first prediction is very dandbil, but combinedtoBy comparing it with the original data, a greater precision is achieved than with the models that we used.ito date and that was the goal”.
The research, carried out over two years, has been carried out by a multidisciplinary team of professionals formed by Javier Garcito Algarra, research director and academic directorandmonkey of toEngineering Areaiay Sciences at the University Center U-tad, Javier Galeano and Iciar Civantos-Gomez, hundrediphics of the Complex Systems Group of the political universityandnica of Madrid, together with the ecologo ORScar Godoy and David Garcia-Callejas of the University Institute of Marine Research (INMAR) at the University of Ctodiz and Ignasi Bartomeus at the Estación Biolormagic of DoñAna (EBD-CSIC).
Other fields of activity
These types of advances are an example of how digitization, big data, analiData ethics and interactive visualization tools contribute to scientific advancementifico and have direct application in the development and conservationn of the environment.
In addition, the study can focus on other fields of activity, such as commercial agriculture. “The same thing that we have done with wild species can be applied to crops such as cereals, forage plants or legumes. These types of tools are already used to select varieties that can better withstand changes in temperature or are resistant to fungi or partosites that proliferate under the new conditions”, concludes the academic directorandmonkey of toEngineering Areaiay Sciences at the U-tad University Center.
George is Digismak’s reported cum editor with 13 years of experience in Journalism