<![CDATA[Rainfall prediction is critical for enabling precision irrigation, particularly in tropical agricultural regions vulnerable to climate variability. This review systematically examines 15 peer-reviewed articles published between 2019 and 2024, using the PRISMA framework to evaluate the performance and applicability of rainfall prediction models for precision agriculture. The models are categorized into statistical (e.g., ARIMA), artificial intelligence (e.g., ANN, LSTM, ELM), and hybrid approaches (e.g., Neural Prophet–LSTM, ANFIS). Quantitative synthesis based on RMSE, MAE, MAPE, and R² reveals that hybrid models generally yield the highest predictive accuracy (e.g., RMSE = 0.0633; R² = 0.98), while AI models perform well on daily, nonlinear datasets but require extensive computational resources and expertise. In contrast, ARIMA remains the most practical and reliable option for monthly forecasting in data-scarce environments, offering a balance between accuracy and operational feasibility (e.g., RMSE = 69.506; MAPE = 31.41%). Contextual factors such as data availability, digital infrastructure, and user capacity significantly influence model suitability. The review also highlights real-world implementations and practical challenges—such as sensor limitations and technical skill gaps—associated with deploying advanced models. Ultimately, this review provides a comparative perspective to guide model selection based on statistical performance and implementation readiness. It further supports national food security goals by aligning predictive modeling with the operational needs of climate-resilient agriculture in supporting climate-resilient agriculture in tropical regions. Abstrak Prediksi curah hujan merupakan komponen penting dalam mendukung irigasi presisi, terutama di wilayah pertanian tropis yang rentan terhadap variabilitas iklim. Kajian ini secara sistematis menelaah 15 artikel ilmiah terbitan tahun 2019 hingga 2024 dengan menggunakan kerangka PRISMA, untuk mengevaluasi kinerja dan relevansi model prediksi curah hujan dalam konteks pertanian presisi. Model yang dianalisis mencakup pendekatan statistik (misalnya ARIMA), kecerdasan buatan (seperti ANN, LSTM, ELM), serta model hibrida (seperti Neural Prophet–LSTM dan ANFIS). Sintesis kuantitatif berdasarkan indikator RMSE, MAE, MAPE, dan R² menunjukkan bahwa model hibrida umumnya memberikan akurasi prediksi tertinggi (misalnya RMSE = 0,0633; R² = 0,98), sementara model AI efektif untuk data harian yang kompleks namun membutuhkan sumber daya komputasi dan keahlian teknis yang tinggi. Di sisi lain, ARIMA tetap menjadi pilihan paling praktis untuk peramalan bulanan di wilayah dengan keterbatasan data dan infrastruktur, karena mampu menyeimbangkan akurasi dan kemudahan operasional (misalnya RMSE = 69,506; MAPE = 31,41%). Faktor kontekstual seperti ketersediaan data, kesiapan infrastruktur digital, dan kapasitas pengguna sangat memengaruhi kesesuaian model. Kajian ini juga mengidentifikasi tantangan implementasi nyata, termasuk keterbatasan sensor dan rendahnya literasi teknologi. Secara keseluruhan, ulasan ini memberikan panduan komparatif dalam memilih model berdasarkan performa statistik dan kesiapan penerapan, serta mendukung upaya ketahanan pangan nasional melalui pemodelan prediksi yang kontekstual dan adaptif terhadap iklim.]]>
