Surface Geological Evolution in the Chang'e 5 Landing Area (Em4 Unit) Revealed by a New Age‐Retrieving Method From Regression Learning

The chronology function and production function have been widely used to derive the model ages of lunar mare regions from crater size‐frequency distributions. Challenges remain in homogenous counting area selection, crater saturation and crater rim identification. Geological unit‐based ages are also difficult to study the continuous surface evolution among adjacent areas. Using regression‐learning models, we have tried a new method on the Em4 unit of the Chang'e 5 landing area to explore a quantitative relationship between ages and surface morphometric expressions using texture features. Four features (Contrast, Energy, Entropy and Homogeneity), together with a stepwise linear model (SL) and a linear support‐vector‐machine model (LS), are well selected to produce a pixel‐level continuous age map of the Em4 unit. Mean age values of 1.75 ± 0.26 Ga and 1.69 ± 0.22 Ga obtained respectively from the two models are consistent with the ages of Chang'e 5 samples returned from this area. Both texture features and age maps are separated along the NW‐SE sinuous rilles (Rima Sharp and Rima Mairan). Comprehensively considering the geology, geomorphology, and newly retrieved ages of the study region, we have proposed a three‐stage evolution process for the Em4 unit. Our new age‐retrieving method is useful for obtaining a pixel‐level high‐resolution age map in the study region and has the potential to be widely used in other lunar mare areas. Plain Language Summary The geological evolution of the lunar surface can provide inspiration for comparative studies on Earth. Lunar surfaces with different ages serve as indicators of the lunar geological history. Traditionally, crater counting has been the primary method for determining lunar surface ages. However, the method faces challenges such as selecting homogenous counting areas and identifying crater rims accurately. In this study, we have developed two regression learning models using texture features of the young geologic unit (Em4) from which the Chang'e 5 mission returned samples. The models can generate pixel‐level age maps of the Em4 unit, revealing continuous age distribution of the area. Both texture features and age maps exhibit asymmetrical patterns along the NW‐SE sinuous rilles (Rima Sharp and Rima Mairan). Together with the geology and geomorphology of the study region, the newly retrieved maps of ages indicate a three‐stage evolution process for the Em4 unit. Key Points Texture features (Contrast, Correlation