Amazonia is the largest remaining tropical rainforest. As such, it is globally important both in terms of carbon and water cycles as well as biodiversity. As in most areas of the tropics, deforestation is also alarmingly fast in the Amazon. Rainforests are subject to many types of land use pressures, such as food production, mining and timber extraction. Although the rainforest area looks quite uniform in aerial photographs, the soil in the area is remarkably variable. This variation, in turn, is reflected in the variation of species composition. There are also regional differences in the structure of the forest. Mapping the variation in soils and species composition is particularly important in conditions of rapidly advancing deforestation, as soil properties affect productivity - both the productivity of the original forest ecosystem and food production. The properties of the soil also affect the assemblage of species present in the area. Large variability of the soil often also means a great variety of forest types and thus a regionally high number of species. Mapping is therefore also important for conservation purposes. The extent of Amazonia and the inaccessibility of many areas, the large number of species and the lack of knowledge about species ecology have slowed down the mapping. The use of indicator species is an effective way to speed up mapping. In this dissertation, I use two large datasets on Amazonian indicator species collected over the last three decades. I draw from the long tradition of Europe's deciduous and coniferous forest zones, where indicator species are used to indicate soil fertility and forest type. I am looking into whether the same could be applied in Amazonia. I look for potential indicator species by modelling their ecological niches in relation to an important soil variable, the concentration of base cations. I also study general plant ecological questions, such as whether common and abundant species are always ecologically wide-ranging generalists. For these purposes, I use HOF models and weighted averaging. I test how accurately different soil properties can be predicted with indicator species. In prediction, I apply the k-nearest neighbour method (k-NN) and weighted averaging calibration. I also study the influence soil has on the structure of the lowland rainforest understorey. I carry out field measurements of stem density in different size classes of trees, and of canopy openness. In the cloud forests of the Andean slopes, I study the relationship between the forest structure and microclimate, and their effect on the abundance and diversity of epiphytic plants. I carry out forest structure measurements and estimate the abundance of epiphytes in the field and record the microclimate with automatic data loggers. In five research papers, I find out that: 1) Both the Melastomataceae family and the fern genera Adiantum and Lindsaea contain several promising indicator species. In addition, the optima of the species in both groups are spread along the base cation gradient in such a way that several narrowniche indicator species exist in all parts of the gradient, 3) Locally abundant species are not always ecological generalists, but can also be specialised to a certain soil type, 3) Soil base cation concentration can be accurately predicted using indicator, but it is also possible to predict the potassium content quite accurately. Regarding other soil variables, the method is not as good, 4) The differences in forest structure between different soil types prove to be difficult to verify with field measurements, but differences have later been found with remote sensing methods, 5) The conditions in the premontane cloud forest become less favourable for epiphytes with increasing canopy openness. This development is reinforced by both climate change and deforestation.