The Representativeness of the Existing Network of Ecological and Climatic Stations Eddy Covariance for the Territory of Russia was analyzed

Within the framework of the May series of scientific seminars of the laboratory, the report of researchers Anna Derkacheva and Robert Sandlerskiy "Results of the analysis of the representativeness of the existing network of ecological and climatic stations (EX) for the territory of Russia" was presented. The primary assessment of geographical representativeness of 29 ecological and climatic stations for climate, relief, soils and vegetation is given. Based on the Random forest method, both for all stations and for each one separately, territories with similar environmental parameters and territories whose physical and geographical conditions are not covered by observations in any way were identified.

Climate changes stimulate the development of a network of monitoring stations in the Russian Federation. The first step is to quantify how well existing stations describe the diversity of the environment. The report presents the primary results of the assessment of the geographical representativeness of 29 ecological and climatic stations for climate, relief, soils and vegetation. Based on the Random forest method, territories with similar environmental parameters were identified, as well as territories that are not covered by observations.

The degree of representativeness was estimated because of the prediction of attributing a 1x1 km pixel to the type "presence of ecological and climatic stations in a pixel similar in environmental variables". To generalize the environment variables, the principal component method was used: 10 factors were identified for all variables, describing 95% of their variation. Relative to the latter, the representativeness was 53%. The analysis of the obtained images shows the insufficiency of 29 points for the correct display of geographical space. On the other hand, the very location of existing ecological and climatic stations is initially determined by historical rather than geographical reasons. Thus, the existing number of ecological and climatic stations is not enough to carry out representativeness assessments. In addition, the proposed algorithm based on Random forest is not quite adequate to the task. Nevertheless, the estimates obtained can be the basis for the overall planning of the development of the monitoring network.