A Multi-Dimensional Framework for Diagnosing Inconsistencies in Remote Sensing-Based Ecosystem Assessment

Oluwafemi David Bejide 

Remote sensing-based ecosystem assessment increasingly relies on the integration of spectral, structural, and model-derived datasets; however, inconsistencies among these data sources can produce divergent representations of vegetation dynamics and carbon processes. This study proposes a multi-dimensional framework to diagnose such inconsistencies by integrating scale, dimensional, and variable components within a unified analytical structure. Using Ekiti State, Nigeria as a heterogeneous tropical test landscape, scale inconsistency was assessed through comparison of CASA-derived and MODIS Net Primary Productivity (NPP), dimensional inconsistency through threshold-based validation of land use/land cover (LULC) against GEDI-derived canopy height (≥5 m and ≥10 m), and variable inconsistency through non-proportional changes between aboveground biomass (AGB) and canopy height model (CHM).
Results indicate persistent scale-related discrepancies, with CASA systematically overestimating NPP relative to MODIS and exhibiting weak spatial correspondence (R² ≈ 0.03 in 2020 and 0.04 in 2025). Dimensional inconsistency was substantial (mean ≈ 0.41), reflecting systematic mismatch between spectral and structural representations of forest extent. In contrast, variable inconsistency was relatively low (mean VI = 0.18), indicating broadly proportional structural–carbon responses across most of the landscape, with localized divergence in disturbed and non-vegetated areas. Integration of these components yielded a Multi-Dimensional Inconsistency Index (MDII) of 0.26, indicating moderate overall inconsistency dominated by dimensional mismatch.
These findings demonstrate that inconsistencies in remote sensing-based ecosystem assessment are systematic and multi-faceted, arising from differences in spatial scale, data representation, and model structure, and highlight the importance of integrated frameworks for improving the reliability of ecosystem monitoring in complex landscapes.