Guizhou Province is a composite of ecologically fragile areas of karst landscape and contiguous destitute areas, and examining the spatial and temporal dynamics of poverty and the factors that influence it has important theoretical and practical significance for targeted poverty alleviation in Guizhou Province. Taking the incidence of poverty as a research index, we analyzed the spatial-temporal dynamic evolution of poverty in Guizhou Province from 2003 to 2015 using the ESTDA framework, and analyzed the factors influencing the spatial pattern using the Geo-detector model. The main conclusions drawn from our analyses are as follows:(1) The poverty among counties shows significant positive spatial autocorrelation and we detected a spatial agglomeration effect of the neighboring distribution of similar counties in relation to poverty. At the local level, a two-level differentiation trend in county poverty is obvious, and the spatial structure of poverty shows a typical "core edge" model. (2) The local spatial structure and spatial dependence direction of poverty have strong stability. There are 53 counties showing collaborative growth, indicating that there is a strong spatial integration of poverty in Guizhou Province. (3) The spatial association model and spatial transfer inertia are characterized by a certain degree of path-dependence or spatial lock-in characteristics. (4) There are certain differences in the influence and significance of various factors relating to poverty. The disposable income per rural capita is the main factor affecting poverty, and the influences of altitude, slope, and vegetation coverage are relatively small. The influence of any two factors on poverty after interaction is stronger than that of any single factor, and the explanatory force shows two types of enhancement:nonlinear or bilinear.