Hybrid Tree-Ensemble and Metaheuristic Dust Aware Frameworks of Solar PV Power Prediction in Arid Regions: Scoping Review Evidence, Gaps, and Future Directions From the Case of Iraq and the MENA Region (2020–2025)

Abstract

Background: In arid and hyper-arid areas, solar photovoltaic (PV) power generation prediction is hampered by high aerosol optical depth (AOD), frequent dust events, high temperatures, and sparse ground-based. measurement networks. In Iraq, the frequency and intensity of dust storms are especially acute, especially under these conditions. inflict quantifiable effects on the solar irradiance supply and PV systems output. Purpose: This scoping review systematically identifies evidence on hybrid generation prediction frameworks (2020-2025) that are peer-reviewed. hybridizing tree-based ensemble machine learning models with metaheuristic optimization algorithms on solar PV. prediction of power generation with specific emphasis on the applicability of the approach to arid climate and the gaps in Iraq-specific research. Methods: A scoping review in adherence to PRISMA-ScR reporting principles was conducted in accordance across. four databases (Scopus, Web of Science, IEEE Xplore, ScienceDirect) of a scientific character. Studies were screened against five pre-established eligibility criteria and categorized under an eight-level taxonomy (T1-T8). N=25 Core studies. were held in evidence to synthesize evidence, with N=26 Contextual references to give methods, environmental, or geographic supporting evidence. Results: Tree-ensemble models prevail on the corpus: Random Forest appeared in N=7 of 21 empirical Core studies (33.3%) and XGBoost in N=6 of 21 (28.6%). Only N=4 experiments regarding the larger literature affirmed Tier-1 hybrid arrangements the use of metaheuristic tree optimizers. hyperparameter tuning the core arid-climate ensemble of N=1 solar generation prediction in four of the twenty-one empirical Core studies (19.0%) dust and aerosol variables were directly used as inputs to the model, but with 54.5 percent Solar irradiance reduction which can be related broadly down to Iraqi measurement over time and 62.3 percent reduction distantly from American reference point level respectively we bring attention years between 2004-2012. Impact of extreme periods of dust events on PV power loss Tree-Ensemble Forecasting Directly : Iraq Gap: N=1 Core study was the only such study. to an Iraqi dataset, validating a MERRA-2 satellite data channel at Mosul but absent metaheuristic optimization. or dust-aware feature engineering. Research Agenda: There are four suggested priority directions: (i) dust-aware feature. ML-retrieved or MAIAC-corrected AOD engineering; (ii) systematic metaheuristic evaluator of optimizers. models in a data-sparse arid environment; iii) validation against satellite-derived forest-cover estimates specific to Iraq. inputs; (iv) the use of common metric protocols, as well as a dust-event subset evaluation framework dedicated to the standardization of benchmarking procedures for enablement consistent cross-study comparability and reproducibility. This scoping review was not prospectively registered; although the PROSPERO database accepts registrations for scoping reviews, no formal protocol was archived or publicly available before the start of our work. However, we strictly followed the PRISMA-ScR guidelines and also the JBI methodology at all steps of a scoping review to ensure methodological rigor and transparency. Potential limitations of the present study include limited dual-independent literature screening, which is expected to add some selection bias; for instance five references were detected after-search and beyond an a priori defined window period (ie. (2025), October 2025: Taha et al. (2025), 02:00 pm November 2025; Shehab et al. (2026); and Oruç et al. (2026) were intentionally kept solely as additional contextual evidence instead of integrated into the main analysis. Additional limitations include conducting our literature search only in English-language databases, which may mean missing relevant studies published in other languages (see Table 1 for justification of inclusion and exclusion criteria), as well as the lack of a pre-archived, prospectively registered protocol to enhance replicability and transparency of the review process.