Understanding the spatial patterns and the driving factors of plant leaf and soil stoichiometry are critical for improving the parameterization of future ecological models and to predict the responses of ecosystems to environmental changes. This study aimed to determine how the latitudinal patterns of leaf and soil C:N:P stoichiometry are affected by climate and vegetation types in the dryland ecosystems. The concentrations of leaf C, N, and P in herb community as well as soil nutrient concentrations along a 500–km–long latitudinal gradient in Northern Shaanxi of the Loess Plateau, were measured. The results showed that the soil C, N, P and C:N:P ratios at all three depths (0–10, 10–20, and 20–40 cm) showed significant latitudinal trends (except for soil C:N ratios) (P < 0.01). In general, the soil C, N and C:N:P ratios decreased exponentially while soil P increased first and then decreased with the latitude. The soil C, N, C:P, and N:P ratios at all three depths (0–10, 10–20 and 20–40 cm) were positively correlated with MAT and MAP (P < 0.05), while soil P and C:N ratios at all three depths were weakly correlated with MAT and MAP (P > 0.05). In addition, leaf C:N:P stoichiometry was significantly correlated with the latitude, MAT, and MAP (except for N:P ratios) (P < 0.01), such that, leaf C, C:N, and C:P ratios decreased as the latitude increased and MAT and MAP decreased, and leaf N, P concentrations increased as the latitude increased and MAT and MAP decreased, while leaf N:P ratios were weakly correlated with the latitude, MAT, and MAP (P > 0.05). Furthermore, the leaf C:N:P stoichiometry of herbaceous communities was related to the soil properties (except for soil P), and we found that the C:P ratios between the soil and leaves were strongly correlated. Compared with the global scale, the relatively high N:P ratios indicated that the vegetation growth of the herb community in the dryland of the Loess Plateau was more susceptible to P limitation.