Although the Water Erosion Prediction Project (WEPP), the Water and Tillage Erosion Model-Sediment Delivery Model (WaTEM-SEDEM), and Landlab are well-respected, process-based approaches to hillslope erosion and deposition, they have never been applied to the same landscapes and directly compared. WEPP is a field-scale hillslope erosion model developed for agricultural watersheds and driven by hydrologic inputs; WaTEM-SEDEM extends USLE principles into a spatially distributed framework for routing soil loss and deposition; and Landlab provides a numerical landscape-evolution environment that simulates diffusion and overland flow.WEPP is a field‐scale hillslope erosion model developed for agricultural watersheds and driven by hydrologic inputs. WaTEM‐SEDEM extends USLE principles into a spatially distributed framework for routing soil loss and deposition. Landlab provides a process‐based numerical landscape‐evolution environment that simulates diffusion and overland flow. This study asks: how accurately can WEPP, WaTEM‐SEDEM, and Landlab predict decadal‐scale erosion and deposition patterns in low‐relief Iowa watersheds (2009–2021) when validated against co‐registered, differenced LiDAR‐derived digital elevation models?

HUC-12 watersheds were delineated across four physiographic regions using ten-meter DEMs in QGIS and ArcGIS Pro. Cover‐factor rasters were generated from NDVI derived from Landsat 8 and Sentinel-2 imagery processed in Python and QGIS. Daily precipitation and temperature data were obtained from Daymet via the pydaymet API and aggregated into constant and annual erosivity scenarios. Models were parameterized with literature-based soil erodibility and land‐cover inputs and executed in batch on a high-performance computing cluster. Observed elevation change was calculated by differencing co-registered 2009 and 2021 LiDAR DEMs, then propagated through benchmark sampling to establish cell‐wise confidence intervals. Model performance was assessed using RMSE, bias, and correlation metrics, alongside sensitivity tests for DEM resolution and slope‐analysis window size.

Preliminary results indicate that all three models capture broad spatial patterns of erosion and deposition but differ in magnitude. WaTEM-SEDEM yields the lowest median absolute error, while WEPP tends to overpredict peak erosion by 10–15% and Landlab underpredicts deposition by 5–10%. Annual erosivity forcing reduces bias by approximately 10% and increases model‐observation correlation (r ~0.65), and larger slope‐analysis windows improve Landlab’s accuracy. Explicit DEM error modeling lowers RMSE by ~0.05 m for all models.

These findings underline the critical roles of temporal climate variability, analysis‐scale selection, and rigorous uncertainty propagation in enhancing predictive reliability of process‐based landscape evolution models.