Publicações

Several studies have been devoted to dynamic and statistical downscaling for both climate variability and climate change. This paper introduces an application of temporal neural networks for downscaling global climate model output and autocorrelation functions. This method is proposed for downscaling daily precipitation time series for a region in the Amazon Basin. The downscaling models were developed and validated using IPCC AR4 model output and observed daily precipitation. In this paper five AOGCMs for the twentieth century (20C3M) (1970-1999) and three SRES scenarios (A2, A1B, and B1) were used. The performance in downscaling of the temporal neural network was compared to that of an autocorrelation statistical downscaling model with emphasis on its ability to reproduce the observed climate variability and tendency for the period 1970-1999. The model test results indicate that the neural network model significantly outperforms the statistical models for the downscaling of daily precipitation variability.

O albedo da superfície, definido como a razão entre a radiação solar refletida e a radiação solar incidente, é uma medida indireta da energia solar absorvida pela superfície. É um importante parâmetro que afeta o balanço de radiação terrestre e, portanto, é considerado em estudos do clima global e regional. O presente trabalho teve como objetivo validar o produto albedo MOD43B3 do sensor MODIS e estudar a variabilidade espaço-temporal do albedo em uma área de caatinga nativa do Nordeste do Brasil. Para isso foram consideradas medidas de radiação solar em uma área de caatinga no município de Petrolina-PE. Os resultados mostraram que os valores de albedo estimados do sensor MODIS foram próximos daqueles encontrados a partir de medidas de campo, apresentando erros em torno de 0,02 em relação às medições. Verificou-se também que na área de caatinga considerada, embora o albedo da superfície não tenha apresentado uma variação temporal considerável, a variação espacial é bastante significativa, sendo que os valores de albedo variaram de 0,10 a 0,26 (em média). Por fim, as informações obtidas sobre a variabilidade espaço-temporal do albedo da superfície poderão ser incorporadas em modelos de superfície vialibilizando estudos mais realistas dos impactos climáticos das mudanças da cobertura da terra, em especial da desertificação no Nordeste do Brasil.

Brazilian Northeast region is covered largely by vegetation adapted to the arid conditions and with varied physiognomy, called caatinga. It occupies an extension of about 800.000 km that corresponds to 70% of the region. In recent decades, considerable progress in understanding the micrometeorological processes has been reached, with results that were incorporated into soil-vegetation-atmosphere transfer schemes (SVATS) to study the momentum, energy, water vapor, carbon cycle and vegetation dynamics changes of different ecosystems. Notwithstanding, the knowledge of the parameters and physical or physiological characteristics of the vegetation and soil of the caatinga region is very scarce. So, the objective of this work was performing a calibration of the parameters of the SSiB model for the Brazilian Northeast Caatinga. Micrometeorological and hydrological data collected from July 2004 to June 2005, obtained in the Agricultural Research Center of the Semi-Arid Tropic (CPATSA), were used. Preceding the calibration process, a sensibility study of the SSiB model was performed in order to find the parameters that are sensible to the exchange processes between the surface and atmosphere. The results showed that the B parameter, soil moisture potential at saturation (øs), hydraulic conductivity of saturated soil (ks) and the volumetric moisture at saturation (ès) present high variations on turbulent fluxes. With the initial parameters, the SSiB model showed best results for net radiation, and the latent heat (sensible heat) flux was over-estimated (under-estimated) for all simulation periods. Considering the calibrated parameters, better values of latent flux and sensible flux were obtained. The calibrated parameters were also used for a validation of the surface fluxes considering data from July 2005 to September 2005. The results showed that the model generated better estimations of latent heat and sensible heat fluxes, with low root mean square error. With better estimations of the turbulent fluxes, it was possible to obtain a more representative energy partitioning for the caatinga. Therefore, it is expected that from this calibrated SSiB model, coupled to the meteorological models, it will be possible to obtain more realistic climate and weather forecasts for the Brazilian Northeast region.

Model GL version 1.2 runs operationally at CPTEC/INPE assessing daily solar radiation from GOES 10 VIS imagery. Besides basic hypotheses about atmospheric variables, the fundamental local variable is reflectance at a pixel observed by satellite (Ceballos et al. 2004). Daily irradiation is obtained by time integration of local irradiance, therefore a geostationary satellite is the best instrument for this task. Northeastern Brazil is equally observed by GOES 10 and MSG satellites, with similar (although opposite) viewing angles. This fact suggests that both satellites could assess solar radiation in this region with similar performance. This paper presents results of a comparison between those estimates. It is shown that GOES and MSG provide similar daily cycle of solar irradiance for several regional sites, compatible with pyranometer measurements. Both satellites provide close results for instantaneous irradiance in nearly cloudless sites. In cloudy conditions, absolute differences can be higher than 100 W/m**2 (about 10-20% or even more of assessed irradiance); this makes evident how delicate can be comparisons over small areas. On the other hand, daily totals (mean irradiance over 24 hours) are quite similar, with standard deviation of about 20 W/m**2 related to ground truth and absolute mean deviation of about 10-15 W/m**2 (5% or less than daily value). However, MSG results tend to be higher than GOES (it is suggested a systematic difference of about +20 W/m**2). The somewhat better fitting of MSG results suggest to improve reflectance assessed for GOES 10 VIS images, which need corrections due to VIS sensor degradation. Nevertheless, the overall results make evident the confidence for using GL model and MSG VIS imagery, in order to extend solar radiation assessments to Atlantic Ocean, poorly covered by GOES images.

Biome models of the global climate-vegetation relationships indicate that most of the Brazilian Amazon has potential for being covered by tropical forests. From current land-use processes observed in the region, however, several areas of natural forests were lead to different states dominated by degraded and secondary forests or even savannas. This may happen because deforestation and presence of ignition sources frequently cause nearby forests to burn. Fire is a major disturbance in tropical forests, which may substantially increase the mortality of trees and other native species. In a first attempt to generally represent the potential for tropical forest degradation due to fires in Brazilian Amazon, we analysed large-scale data on fire occurrence and climate factors. Fire-activity information are based on remote-sensing detections with the Tropical Rainfall Measuring Mission – Visible and Infrared Scanner (TRMM-VIRS). Climate information are summarized by a wetness index used as an estimate of dry/wet climate conditions, and a soil-moisture seasonality index. Our results have important implications for evaluating the resilience of tropical forests in the Brazilian Amazonia by presenting general relations between climate and the occurrence of fires/deforestation, which can help to refine the estimates of the allocation of forest-savanna boundaries based on the likelihood for land-use fires in the region.

Fire research is key for understanding the ecosystem functions of Amazonia and the sustainability of land-use practices in the region. Fires are significant disturbances to biogeochemical cycles, and have important links to land use and climate. Major methods for studying fires include fieldwork, remote sensing and modeling. The potential benefits from the synergy of these methods are numerous. Occasional ground-based fire activity information can be supplemented by frequent data from satellites. Remote-sensing fire products can be improved using accuracy assessments based on ground observations. Models rely on calibrations and validations using both ground- and satellite-based data, and are the only tools for projecting future fire activity and effects. The review of several studies representing fire research methods with emphasis in their synergy show that different methods typically provide information at different spatial and temporal scales, and there is an important tradeoff between datasets resolution and domain. Ground-based methods can provide information at high spatial and temporal resolution, but they do not usually provide data for large domains. Remote sensing and modeling studies can provide information for larger domains, but their data are usually at lower resolution, which complicate their interpretation for studying fires. Studies based on the synergy of all methods have a great potential to significantly advance fire research. One important outcome expected from this type of studies is the possibility of building a class of models that can be used for multi-scale applications, ranging from warning systems to climate and Earth System modeling. While combinations of multiple methods are desirable, important challenges exist including technological, expertise and resource needs. The perspectives for overcoming these challenges are promising because the recent advancements in technology and increases in multidisciplinary expertise, but will depend on sustained or increased budgets for supporting relevant research.

This work presents a general architecture for building and simulating agent-based models that use real-world geospatial data, take into account all the ways geospatial data can feed these models. We focus on how data can be used to create an initial arrangement for the model, as if it was a static representation. We have as hypothesis that the Generalized Proximity Matrix (GPM) is a foundation for setting up the relations between the entities of an agent-based model for simulating geospatial phenomena.

Following the extensive conclusions from the IPCC (Intergovernmental Panel on Climate Change) Fourth Assessment Report and other reports (including INPEs report 2007 – ww.cptec.inpe.br/mudancas_climaticas) we have credible evidence that the climate is changing across the world. But it is important to note that while the current versions of atmosphereocean general circulation models (AOGCMs) have the ability to simulate well the state of the global climate at the large and continental scales, there are significant variations between these models in future climate projections of precipitation and temperature changes at the regional scale, including those for South America. One of the top priorities for narrowing gaps between current knowledge and policymaking needs is the quantitative assessment of the sensitivity, adaptive capacity and vulnerability of human and natural systems to climate change. Vital for such assessments are reliable estimates of current and future climate variability at the regional scale which can be readily used to assess the sensitivity of these systems to climate change. Often an important requirement for these assessments is for the climate data to be provided at high spatial and temporal resolution, and the main method for providing these data regionally is dynamical downscaling, i.e. output from global climate models is used to drive a high resolution regional climate model. Regional models 1provide improved spatial detail, but in order to improve reliability of projections, it is essential to run multiple realizations, to take uncertainties into account. There has been much effort to quantify the range of uncertainties that are known to exist in global climate model projections and dynamical downscaling allows a detailed exploration of these. Important for the interpretation of any downscaled projections is to assess the regional-scale climate and climate changes in the global projections. This can guide the selection of suitable global models for driving the regional model where the quality of global model control simulations and the identification of global model large-scale projected changes which are considered reliable would be relevant information. As a starting point for this, in the present paper, we present the results of an ensemble simulation of the HadCM3 climate model, where each ensemble member incorporates different but plausible versions of the parameterizations of important physical processes. This is used to assess the potential impacts of climate change on precipitation and temperature over South America and explore the range in projections obtained via the modifications to the model parameterizations.

Land changes are determined by a complex web of biophysical and socio-economic factors that interact in time and space, in different historical and geographical contexts, creating different trajectories of change. It is peopleís response to economic opportunities mediated by institutional factors that drives changes. In this paper we discuss how to incorporate such institutional land tenure categories in land change models. Our hypothesis is that this is an essential step in the direction of constructing regional models to represent the heterogeneity of actors and processes. We implemented the conceptual proposal using the TerraME modeling Environment. Through a case study we analyze how the existence of different rules of territory use affects the landscape dynamics at the regional level.

Land changes are the result of a complex web of interaction between human and biophysical factors, which act over a wide range of temporal and spatial scales. In this paper we conceptualize spatial relations among geographic objects at different scales. We analyze two types of spatial relations: hierarchical, which handles the interaction of nested objects at different scales (countries, states and municipalities, for example); and a network-based relation, which handles action-at-a-distance types of interaction (market chains, for instance) of objects at different scales (such as farms in Central Amazonia to soybean market consumers at the global scale). We implemented such relations in the Terralib environment, in which they can be constructed using selected strategies, and then used in dynamic models. We exemplify the use of such concepts in a real-world case study in the Brazilian Amazonia. We conclude that combining hierarchical and network-based spatial relations provide a comprehensive conceptual framework to include top-down and bottom-up interactions and feedbacks in multi-scale landchange models.

However the spatial resolution of GCMs remains quite coarse , in the order of 300 x 300 km, and at scale , the regional and local details of the climate which are influenced by spatial heterogeneities in the regional physiography are lost. Therefore, there is the need to convert the GCM outputs into a reliable data set with higher spatial resolution, with daily rainfall and temperature time series at the scale of the watershed or a region to which the climate impact is going to be onvestigated. The methods used to convert GCM outputs into local meteorological variables required for reliable climate modeling are usually referred to as downcaling techniques (Hewinton e Crane, 1996).

As Histoty embraces the beginning of a new millennium, old problem still constitute enormous challenges to the popoulation in general, and to the academic world in particular. It is now widely accepted that General Circulation Models (CCMs) represent the most satisfactory technique to answer these challengs (IPCC, 1996). Numerical models (General Circulation Models or GCMs), representing physical processes in the atmosphere, ocean, cryosphere and land surface, are the most advanced tools currently. avaible for simulating the respose of the global climate system ti increasing greenhouse gas concentrations. This work analyses the performance of the IPCC models (CCma, CCSRNIES, CSIRO, GFDL, HACM3, and others) in simulate the present and future climate pattern of the rainfall over the South America Continent. It general the models get to reproduce the phase of the annual cycle of the rainfall. In this work was used four common metric are reviewed, the Heidke skill score, relative operating characteristic (ROC) skill score, equitable threat score, and the rank analog.