Landscape display is a key mode for Android devices, yet many apps lack dedicated landscape UIs. An analysis of 289 popular apps from Google Play shows only 20 provide landscape interfaces; most either stretch the portrait UI or do not support landscape at all. One major reason is the additional development cost, which averages 30%. This motivates the need for automated methods that generate landscape UIs from existing portrait designs, reducing costs and promoting wider adoption. Existing approaches typically rely on fixed adaptation rules to stretch or resize the portrait UI, producing results that differ significantly from developerdesigned landscape layouts and lacking the ability to generate runnable code. To address these limitations, we propose UILAND (UI-LANDscape Adapter for Android Apps), which takes portrait UI code and component coordinates as input and outputs both a design sketch and executable code for the landscape UI. UI-LAND frames landscape adaptation as a multi-objective optimization problem, considering the actual content of the portrait UI rather than relying on rigid rules to determine the optimal layout. We evaluate UI-LAND on 20 real-world apps using metrics including Classification Accuracy (CA), Intra-Class Distance (ICD), and UI Deviation (UD), achieving 92.3%, 37.7%, and 21.2%, respectively, outperforming baseline methods. Furthermore, we submitted 11 pull requests implementing generated landscape UIs on GitHub, with 4 successfully merged, demonstrating practical effectiveness. An ablation study confirms that each module of UI-LAND contributes critically to its performance.