LAUSR

Abstract All over the world, there is increasing demand for wood and other goods from seasonally dry tropical forests; the “Caatinga” forest in northeast Brazil is a case in point. In order to set up sustainable forest management protocols, a comprehensive understanding of the main drivers of forest growth is needed, but few studies have focused on this subject in the Caatinga biome. Traditionally, periodic annual increment (PAI) has been calculated by dividing standing stock by the legally imposed minimum cutting cycle of 15 years, but it is doubtful that this guarantees sustainability. We use data from 20 coupes spread over 10 managed areas and apply both multiple regression and tree regression techniques to correlate PAI with 27 environmental variables including mean annual rainfall and many soil properties. We find that neither the time since harvesting nor the stock before harvest are significantly related to PAI. Instead, using a simple linear regression model, we show that rainfall can explain most (72%) of the variation, while a tree regression model, which captures non-linear relations between rainfall and PAI, explains 96% of the variation. On the other hand, no soil factors contribute significantly to the overall explanation of growth after harvest. We conclude that planning of sustainable management could be greatly improved by use of our regression models and rainfall data which are widely available at local level across the Caatinga. Moreover, this would obviate the need for costly forest inventories.