EGU22-2212, updated on 27 Mar 2022
https://doi.org/10.5194/egusphere-egu22-2212
EGU General Assembly 2022
© Author(s) 2022. This work is distributed under
the Creative Commons Attribution 4.0 License.

Applying artificial intelligence in modeling the relationship of tree ring growth index with different climate variables

Nasrin Salehnia1,2 and Jinho Ahn1,2
Nasrin Salehnia and Jinho Ahn
  • 1School of Earth and Environmental Science, Seoul National University, Seoul, Korea
  • 2Center for Cryospheric Sciences, Seoul National University, Siheung, South Korea (salehnia61@gmail.com; jinhoahn@gmail.com)

Trees are one of the best sources of high-resolution proxy data to understand the past climate. By analyzing Tree Ring Width (TRW) data with climate variables and indicators, we can find main clues, which can help us to encode paleoclimate signals. Nowadays, scientists try to model TRW data with various climate data versus reconstructing the mentioned data for an extended period through different statistical methods. One of the newest methods is AI (Artificial Intelligence). This study aimed to model TRW data with the most effective climate variables by comparing the statistical methods vs. the AI method. First, seven climate variables were gathered from the nearest synoptic station (Sokcho) to the TRW site (Whachae Peak-Sorak), in northeast South Korea, during 1901–1998. The climate variables include maximum temperature (Tmax), minimum temperature (Tmin), mean temperature (Tm), diurnal temperature range (DTR = Tmax – Tmin) (°C), precipitation (Pr) (mm), and vapor pressure (VP) (hPa). The in-situ data were applied to correct the Climate Reach Unit (CRU, Version 4.03). Moreover, we have checked two meteorological drought indices, namely the Palmer drought index (PDSI) and standardized precipitation index (SPI). We applied two regression methods (namely multiple linear regression (MLR) and stepwise regression (SR)) and one AI (Nonlinear autoregressive with exogenous input (NARX)) method. In the first step of analyzing data, we did not see any specific significant results for the relationship between drought effects and TRW data in the case study. Then in the second step, modeling continued with the climate variables. Finally, the results demonstrated that among the three used methods, the NARX method achieved the best outcomes, as MLR with r = 0.44 (p < 0.003); SR with r = 0.27 in p < 0.001; and the NARX model was the best outcomes with r = 0.78. This study revealed that regression methods were not strong enough to reconstruct TRW data. Whereas, by noticeable results, the AI method has obtained the best performance.    

How to cite: Salehnia, N. and Ahn, J.: Applying artificial intelligence in modeling the relationship of tree ring growth index with different climate variables, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-2212, https://doi.org/10.5194/egusphere-egu22-2212, 2022.

Displays

Display file