|
|
|
| A new non-linear approach to approximate the function linking climate and physiognomy |
 |
Text Size: A A A |
|
There is a relationship linking climate and physiology independent of taxonomic composition; however, we have little idea of the form of the function, how complex it is , and its parameters. Non-linear relationships should be sought to improve the precision of paleoclimatic reconstruction from leaf physiognomy.
Prof. ZHOU Zhekun and his team of Xishuangbanna Tropical Botanical Garden (XTBG) conducted a study ti explore a different way of revealing the information content of physiognomic space. They explored a new non-linear approach to approximate the function linking climate and physiognomy. The general regression neural network (GRNN) is a type of artificial neural network (ANN) that can approximate to both linear and nonlinear regressions As such the GRNN is a useful technique to investigate the climate and physiognomy relationship. They tested GRNN on two different physiognomy data sets and compared their results with those obtained from other computational methods. They also tested the GRNN using different physiognomic characters and fossil sites from North America.
The study found that the new algorithm (CLANN) revealed a high-resolution climatic signal in leaf form. Tests showed that the predictions were repeatable, and robust to information loss and applicable to fossil leaf data. The CLANN neural network algorithm was applicable to fossil leaf data and could form a new climate proxy.
The study entitled “Artificial neural networks reveal a high-resolution climatic signal in leaf physiognomy” has been published online in Palaeogeography, Palaeoclimatology, Palaeoecology.
|
|
