2

ABSTRACT: Ecological niche model (ENM) generally takes mechanistic (or physiological) form and estimates the species distribution based on causal factors (e.g. rainfall or sunlight). Data for such causal factors are difficult or expensive to collect and are frequently in small quantity. Terrain-related variables are used as proxy of causal factors, since topographic data can be easily acquired by remote sensing. Taiwan red cypress (TRC) often form pure forests and grow in fog-forest belt (FFB) with rich moisture carried by northeastern season wind (NESW). TRCs grow in the north of study area, but almost do not grow in the Huisun Experimental Forest Station (HEFS) in the south of study area. The study attempted to determine if the NESW blocked by topographic obstacles is a key to this distribution by reversely thinking the positive effect of topographic shelters associated with Taiwan fir in Hohuan Mountains. Topographic sheltering index (TSI) was developed and ENMs were built by maximum entropy (MAXENT) with combinations of elevation, TSI, slope, and vegetation index. The absence of TRC forests in most FFB areas of the HEFS is shown to be attributable to the adverse effects of mountain ranges in the northeast of the study area. The model with elevation plus TSI is the best among ENMs. This indicates that predictors currently used in models, except elevation and TSI, are useless for explaining abovementioned distribution, and TSI can help to explain it. Moreover, the greater the TSI, the fewer TRC trees grow, especially in HEFS. TSI can improve model performance; however, the improvement is relatively limited. Hence, further refined TSI (progressive lessening of the NESW as the latitude decreases fro