Pedoclimatic context often defines a soil’s basic functions, but human management can have superimposing impacts on soil health and carbon (C) stocks. Therefore, it can be challenging to interpret soil health measurements, define benchmarks, and predict the effects of management on soil organic carbon (SOC) stocks in the context of a region’s climates, soils, and cropping systems. In chapter 2, I examined the effects of soil texture, a dominant inherent soil property, and cropping system on soil health indicators in New York State (NYS), USA, soils. Available water capacity measured on disturbed samples was mostly affected by texture, while soil respiration, protein, and wet aggregate stability were mostly impacted by cropping system. Pasture and Mixed Vegetable systems tended to have the highest biological and physical soil health and Annual Grain and Processing Vegetable cropping systems had the lowest. In chapter 3, I developed production environment soil health (PESH) benchmarks for eight physical and biological indicators in the context of region, soil texture, and cropping system. Long Island had lower PESH benchmarks for soil organic matter (SOM) than the rest of NYS, implying that regional PESH benchmarks within a state or region are warranted if the pedoclimatic context varies greatly.In chapter 4, I assessed the effects of tillage system on SOC stocks in long-term continuous corn silage and corn grain experiments. No-till did not lead to consistent benefits in SOC stocks relative to plow-till, across the experiments. The differences in SOC stocks between tillage treatments can be explained by pedoclimatic variables and the amount of C inputs. Prediction of soil health indicators that are expensive to measure can improve the cost-effectiveness of comprehensive assessment of soil health. In chapter 5, I developed pedotransfer functions for available water capacity, field capacity (FC), and permanent wilting point (PWP) using random forest (RF) and traditional multiple linear regression modeling. In chapter 6, pedotransfer functions for soil protein were developed using RF and traditional multiple linear regression modeling. Soil protein was sensitive to management at 36 of 57 long-term experiments and the full RF model was able to predict 92% of those significant effects.