Abstract
The objective of this work is on one hand to show the vulnerability of forest landscapes in relation to climate variability at the scale of Tanzania and on the other hand to analyze forest dynamics in order to carry out a prospective study of the dynamics of land use in the forest reserves of Pugu and Kazimzumbwi. Analysis of the TRMM data over the period from 2001 to 2013 has allowed revealing a seasonal and inter-annual variability in precipitation across the country. The inter-annual precipitation maps have made it possible to distinguish the years with low rainfall (2003, 2005, 2012 ), the years of high rainfall (2002, 2007, 2006, 2011) and the years of intermediate rainfall (2001, 2007, 2008, 2009, 2013). It has also help to distinguish 11 types of rainfall regimes marked by different patterns of seasonal variability at the scale of Tanzania. There are oppositions between the rainfall regimes of the central savannah region on one hand marked by low annual heights over an important period of seven dry months, also more affected by inter-annual variability, and the northern, southern and eastern forest regions are more humid and presenting low deficits of heights inter-annual rainfall. The sensitivity of plant phenology to rainfall variability has been analyzed by the examination of the spatio-temporal relationships between the standardized vegetation index NDVI-MODIS and rainfall (rain TRMM). The rain / NDVI correlation maps show an opposition between the dry regions of the center marked by savannah landscapes highly vulnerable to rainfall variability and the southern regions of moist forests, mountains and coastal regions, mangrove forests Reacting poorly to this rainfall variability. In the savannah regions of the center, the intensity of rain / NDI dependence is measured by a correlation coefficient of 0.70. A monitoring of the analysis of human pressures on forest reserves was carried out using the example of the Pugu and Kazimzumbwi forests during the period 1995-2015 using SPOT 6 (high resolution) and LANDSAT imagery. The land use classifications were realized from the object oriented method. The forest review shows that in 2015 (55% of which 32% is dense forest), from the two reserves only the reserve of Pugu still preserves nearly the half of its surface in forest while the reserve forest of Kazimzumbwi contains only 5% of its area. Over the entire period studied, the sub-period 2009-2014 was the most critical in terms of forest loss. In fact, within five years the forest reserves of Pugu and Kazimzumbwi have almost lost the double of their area. Based on the increased vulnerability of human pressures in the Pugu and Kazimzumbwi forest reserves and their periphery, a multicriteria analysis has made it possible to identify areas of high and low human pressures. The most vulnerable areas remain those located close to the communication axes and cities. Consequently forest reserves are more vulnerable in their eastern parts, close to major roads and major urban centers such as Pugu and Kazimzumbwi. The use of a model for prospective modeling in 2050 has required the integration of the explanatory variables of the observed changes and the land use maps of 1995 and 2014. The model is validated from a predicted map and a real map. The result shows an exact simulation at 72%, based on this hypothesis of an increase in anthropogenic human pressures on the two forest reserves over time; we have predicted the land use map of 2050 under the effect of explanatory variables. This prospective modeling therefore envisages, by 2050, an expansion and densification of artificial surfaces, notably at the north-eastern periphery of the reserve of Pugu and on the south in the kazimzumbwi reserve. This growth in artificial surfaces will result in a significant decline in existing forest areas within reserves.