Abstract: To address the redundancy of monitoring points and the difficulty in simultaneously reducing point numbers while maintaining representativeness in farmland quality monitoring networks, a multi-objective optimization method for monitoring point layout was developed for typical counties in the black soil region of Northeast China. A total of 406 soil-testing and formulated-fertilization monitoring points in Gongzhuling and Lishu Counties, Jilin Province, were used as the original candidate point set. Soil pH, soil organic matter (SOM), total nitrogen (TN), available phosphorus (AP), and available potassium (AK) were selected as the basic soil attributes. Conditional Latin hypercube sampling (cLHS) was used to construct the initial population, and the non-dominated sorting genetic algorithm II (NSGA-II) was then employed to optimize the monitoring point layout. The number of monitoring points and the comprehensive representativeness error were taken as the optimization objectives. An n-Q relationship curve between monitoring point number and comprehensive representativeness error was constructed, and the recommended monitoring point number and corresponding layout were determined by jointly considering the marginal change in the n-Q curve and the comparison among schemes with different monitoring point numbers. Ordinary kriging was further used to evaluate the preservation of spatial patterns and the predictive performance at unretained samples. The results showed that the comprehensive representativeness error decreased with the increased number of monitoring points, whereas the magnitude of improvement gradually diminished. Based on the marginal change in the n-Q curve and the comparative performance of different monitoring point number schemes, 130 monitoring points were selected as the optimized scheme, representing a reduction of 68% from the original 406 points. The probability density distributions of the integrated fertility index (IFI), pH, SOM, TN, AP, and AK for the optimized monitoring points were generally close to those of the original monitoring points, indicating that the main value ranges and distribution patterns of the major soil attributes were well preserved after optimization. Ordinary kriging models were established using the 130 optimized monitoring points and validated against the unretained samples. The R² values for pH, IFI, TN, and AP were 0.41, 0.38, 0.36, and 0.34, respectively, while the root mean square error (RMSE) and mean absolute error (MAE) for IFI were 0.21 and 0.163, respectively. These results indicated that, the optimized network retained a certain ability to characterize the spatial variation of major soil properties and to predict values at unretained locations, despite the substantial reduction in the number of monitoring points. To further assess the effectiveness of the proposed method, the cLHS+NSGA-II scheme was compared with random sampling, stratified sampling, and spatial k-means under the same monitoring point number of 130. The IFI prediction R² of the proposed scheme was higher than those obtained using the three comparison methods, while its RMSE and MAE were also relatively lower. This comparative advantage indicated that the proposed method could better coordinate point reduction, attribute representativeness, and spatial information preservation than the alternative sampling schemes under the same monitoring scale. Overall, the proposed cLHS+NSGA-II framework can reduce redundant monitoring points while maintaining the major statistical characteristics and spatial consistency of farmland soil attributes. The method provides a practical reference for optimizing farmland quality monitoring point layouts and improving the allocation efficiency of monitoring resources in the black soil region of Northeast China.