Abstract: Desiccation cracking in farmland soils is not only a visible surface phenomenon but also a process that reshapes soil structure, modifies evaporation pathways, and alters the redistribution of near-surface moisture. Despite its hydrological and engineering significance, many existing numerical studies still treat soil-water movement and crack development as weakly coupled or even independent processes, which limits their ability to reproduce crack evolution under field conditions. To address this gap, this study develops a coupled modeling framework for farmland soil moisture evaporation and desiccation cracking, with particular emphasis on the dynamic feedback between moisture loss and crack propagation. The model links a soil-water movement scheme to a three-dimensional cracking model derived from linear elastic fracture mechanics, to simulate moisture variation, volumetric shrinkage, and crack development in an integrated manner rather than as separate stages.The study is based on farmland soil from Yinchuan, Ningxia, where field observations and laboratory response tests were conducted to support model construction and validation. In the field experiment, crack images and crack-depth measurements were collected from replicated test plots during post-saturation drying, allowing the spatial distribution of crack depth to be quantified under natural evaporative conditions. A second experiment was designed to establish the response relationship between soil shrinkage rate and volumetric water content. This relationship serves as the key transfer mechanism between the moisture sub-model and the cracking sub-model, enabling changes in water content to be translated into shrinkage deformation and subsequent fracture development. To reflect the influence of cracks on moisture transport, the coupled framework also incorporates feedback from crack surfaces to the evaporation boundary and seepage behavior.Model parameters were calibrated using a Monte Carlo search procedure based on field observations, and performance was assessed through both geometric and statistical criteria. Crack images from experiments and simulations were compared using Minkowski density descriptors, while crack-depth relative frequency was used to evaluate the vertical distribution of cracking. The coefficient of determination (R²), index of agreement (IA), bias (BIAS), and root mean square error (RMSE) were employed to quantify the agreement between simulated and observed results. The field data showed that crack depth was concentrated mainly within the 3–9 cm interval, indicating that the middle layer of the soil profile was the dominant zone for crack development under the tested drying conditions. The proposed coupled model successfully reproduced this feature and captured the main characteristics of crack-network morphology and vertical stratification.Validation results demonstrate that the model provides strong predictive performance across both planar crack geometry and depth-related statistics, with R² values no less than 0.8759, IA values no less than 0.9507, BIAS values no greater than 0.1039, and RMSE values no greater than 0.1329. Compared with the baseline desiccation-cracking model that relies primarily on surface evaporation as the driving mechanism, the coupled framework performs better in representing crack length, area, network topology, and especially the propagation of deeper cracks. These improvements indicate that explicitly accounting for the interaction between soil-water movement and cracking enhances both the physical realism and the predictive accuracy of field-scale simulations.Overall, the study provides a practical and physically grounded framework for analyzing the co-evolution of moisture evaporation and desiccation cracking in farmland soils. The proposed model can support future research on soil structural degradation under drying, while also offering potential value for irrigation management, soil-water regulation, and the assessment of crack-related risks in agricultural and geo-environmental settings.