In many regions of the world, fires are an important and highly variable source of air pollutant emissions, and they thus constitute a significant if not dominant factor controlling the interannual variability of the atmospheric composition. This paper describes the 41‐year inventory of vegetation fire emissions constructed for the Reanalysis of the Tropospheric chemical composition over the past 40 years project (RETRO), a global modeling study to investigate the trends and variability of tropospheric ozone and other air pollutants over the past decades. It is the first attempt to construct a global emissions data set with monthly time resolution over such a long period. The inventory is based on a literature review, on estimates from different satellite products, and on a numerical model with a semiphysical approach to simulate fire occurrence and fire spread. Burned areas, carbon consumption, and total carbon release are estimated for 13 continental‐scale regions, including explicit treatment of some major burning events such as Indonesia in 1997 and 1998. Global carbon emissions from this inventory range from 1410 to 3140 Tg C/a with the minimum and maximum occurring in 1974 and 1992, respectively (mean of 2078 Tg C/a). Emissions of other species are also reported (mean CO of 330 Tg/a, NO_xof 4.6 Tg N/a, CH₂O of 3.9 Tg/a, CH₄of 15.4 Tg/a, BC of 2.2 Tg/a, OC of 17.6 Tg/a, SO₂of 2.2 Tg/a). The uncertainties of these estimates remain high even for later years where satellite data products are available. Future versions of this inventory may benefit from ongoing analysis of burned areas from satellite data going back to 1982.

Land suitability evaluation in a raster GIS environment is conceptualized as a multiple-criteria decision-making (MCDM) problem. A combination of MCDM techniques selected for implementing the methodology included value and priority assessment techniques for scaling the interval and ordinal data respectively, and compromise programming (CP) to aggregate the unidimensional evaluations. The contribution of the proposed methodology to handle problems of scaling and dependence that often affect expert-based suitability analyses is discussed. A case-study of habitat evaluation for the endangered Mount Graham red squirrel is presented. The multiple-criteria models resulting from the CP analysis of an expert's perception of the habitat preference structure of the red squirrel are compared with data of actual habitat use. The predictive power of the models is good and sensitivity analysis based on the distance-metric parameter p of CP reveals some interesting differences between alternative strategies for data aggregation.