In order to develop ecology into a predictive and manipulative science, we need to have a method to reconstruct the nonlinear dynamics from a very limited dataset (data size-problem) and a method of controlling chaos (manipulation-problem).The discovery of nonlinear dynamics in natural ecological systems has attracted much attention after May$B!G(Bs historical work. Nonlinear time series analysis (NTSA) is used to investigate deterministic chaos. However, most ecological time series are too short to perform NTSA, which requires a time series whose size is in the thousands. We developed "ensemble reconstruction" of the dynamics from a very short ecological time series whose size is smaller than ten. Alternate bearing is a common phenomenon in tree crops such as citrus and nuts. Most tree crops such as citrus, nuts and acorn, show large fluctuations in annual yield, so-called alternate bearing and/or mast fruiting. In this presentation, I show a method of reconstructing the global and local dynamics from a very short ecological time series. For the global dynamics, we tested the feasibility of applying response surface methodology and for the local dynamics, I demonstrate the developed "ensemble reconstruction". We used an ensemble dataset consisting of the yields of 48 individual trees over 7 years to test our proposed method and successfully validated this method by one-year forward prediction three times in 2002, 2003 and 2004. Reconstructed models were also validated by one-year forecasting fitness and by controllability in numerical experiments based on the OGY method developed for controlling chaos.