In this study, the impacts of climate related variables such as temperature, maximum temperature, precipitation, solar radiation on wheat yield are estimated by using a regional panel data (NUTS2) for Turkey. The estimated coefficients are used to simulate the impacts of climate change on wheat yield in Turkey until 2100. The humidty coefficient (nks) of Aydeniz climate classification, July Maximum Temperature and August Maximum Temperature are also used in the estimations. The regional effects (for seven regions of Turkey) are estimated separately for each decade. Apart from point estimation, the upper and lower bounds of the estimations are also provided. The "average scenario" points out a decline of about 8% in the wheat yields. In the "worst scenario" the decline enlarges to about 23%. However, the "best scenario" based on the statistically significant lower bound estimation values shows only a 1% of decline in the wheat yields of Turkey. This large interval can be explained by the high varance of the estimations and long forecasting period until 2100. In the study, to take into account the non-stationarity of variables and cross-sectional dependence of panel data, the heterogenus panel data estimation method of AMG (Augmented Mean Group Estimator) is preferred. The AMG estimator of Eberhardt is robust to the cross-sectional dependence of panel series. We argue that this feature makes AMG a good estimator for climate change estimations based on panel data. The residuals of Fixed Effect, Random Effect, FGLS and PCSE estimators are found to be non-stationary pointing unhealthy estimations results.