Trace elements such as zinc (Zn), copper (Cu) and boron (B) are important micronutrients for crop production. Their bioavailability is essential to crops yield quantity and quality in many tropical soils. Nutrient bioavailability depends partly on the soil nutrient status, and in particular on the reactive and soluble fractions. Adsorption/desorption and precipitation/dissolution processes control the partitioning of the reactive pool over the solid and solution phase. However, so far the solid-solution partitioning of trace elements has mostly been studied in temperate and often contaminated soils. We studied, therefore, the solid-solution partitioning of Zn, Cu and B for 172 soils from Burundi, Rwanda and Kenya, using extensive soil characterization in combination with multi-surface modelling and two types of empirical Freundlich type partition relations. Our aim was to enhance the understanding of the soil chemical processes that control the solid-solution partitioning of the three micronutrients in these soils from the tropics with a multi-surface model, and to use this knowledge as benchmark to develop partition relations that require less input data and are more convenient tools for predicting the concentration in solution based on existing soil data. We show that the generic multi-surface model applied to these tropical soils performs similarly for Zn and Cu as in previous studies on temperate and contaminated soils. The Zn and Cu speciation was dominated by adsorption to soil organic matter, with an increased importance of metal (hydr)oxides with increasing pH. Given its generally low concentrations in these soils, dissolved organic matter was found to be important only for the solution speciation of Cu. The adsorption of B was mainly to metal (hydr)oxides at low pH, and with increasing pH soil organic matter became more important. The multi-surface model overestimated the dissolved B concentration for most soil samples, which we attributed to an inaccurate estimation of reactive B. Interestingly, the variation in observed and modeled solid-solution partitioning expressed as Kd of Cu and B among the soils was relatively small (∼1 log L kg−1), and the concentration in solution was consistently mainly controlled by the reactive concentration. Generally, the optimized partition relations resulted in a smaller prediction error compared to the multi-surface models. The partition relations in which the concentration in solution was optimized, resulted generally in an overestimation for the lowest observed concentrations, and an underestimation for highest concentrations of all three elements. Partition relations with optimized Freundlich parameters Kf and n resulted in more robust predictions since the prediction error was not related to the actual measured concentration. The partition relations from this study are easy-to-use tools for predicting the dissolved concentrations of Zn, Cu and B in soils from the tropics with low contents of these micronutrients and can therefore enhance the use of current existing soil information data for Sub-Saharan Africa.