The first observations related to the lightning activity in Brazil were performed in the beginning of the 1960s, when the keraunic level, that is, the number of days per year on which thunder is heard at a given location, began to be reported at different parts of the country. These observations covered three decades and provided the earliest information for the whole country. Based on these observations, it was concluded that in most parts of Brazil thunderstorms occur in more than 50 days per year, with a large fraction of the country experiencing thunderstorms in more than 100 days per year. The maximum value of 140 thunderstorm days per year apparently occurs in the Amazon region.
 At the end of the 1980s, the first lightning observations in Brazil using a ground network were made in the Southeast region. During this study, the lightning network in the Southeast was composed of 14 sensors (4 Impact T-141, 4 LPATS-III and 6 LPATS-IV sensors). The location of these sensors and the region of study (from 14°S to 24°S of latitude and from 42°W to 52°W of longitude) are shown in Figure 1. The estimated detection efficiency of the network in the region of study was assumed to be 80%. Only positive flashes with peak current above 15 kA were considered in this study to avoid a possible contamination by intracloud flashes.
 In 1999, a four T-141 ES Impact-sensor lightning network was installed in the North region of the country through a collaboration program between INPE and NASA [Blakeslee et al., 1999]. The location of these sensors and the region of study (from 6°S to 16°S of latitude and from 57°W to 67°W of longitude) are also shown in Figure 1. Although there are fewer sensors in the North region, their higher sensitivity allow us to consider the same estimated detection efficiency (80%) for a region with similar area. It is also assumed the threshold of 15 kA for positive flashes as above.
 After 1995, with the new technology of optical sensors on board orbiting satellites, other methods to detect the lightning activity in Brazil became available [Christian et al., 1999]. In 1997, the Lightning Imaging Sensor (LIS), the second sensor of this new generation of sensors, was launched on board the TRMM satellite in a lower altitude orbit than the previous sensor (the Optical Transient Detector – OTD). When the satellite sensors pass over Brazil, however, they are subjected to the influence of the South Atlantic magnetic anomaly (SAMA), a large region covering part of South America and Atlantic ocean, where the Earth's magnetic field has its lowest intensity [for a review, see Pinto Jr., 1993 and references therein]. The energetic charged particles from the inner radiation belt in the SAMA may produce pulses in the sensor output, which may be confused with lightning events. Considering that the core of the SAMA is presently in the South region of Brazil and, in consequence, closer to the Southeast region than the North region, some difference in the influence of the SAMA on the LIS data in both regions may be expected. The effect of the SAMA on the LIS results, however, is expected to be small and it will be neglected in this study. A more detailed study of the influence of the SAMA on the LIS data is currently being made by the MSFC group. In addition, due to the lack of discrimination between cloud-to-ground and cloud flashes, the LIS observations should be seen as representing the total lightning activity. Another aspect in the LIS data that should be considered is its diurnal bias. In order to avoid this sampling limitation, the comparative analysis presented in this paper is based on data blocks (or windows) of 49 days that match the natural precession cycle of the satellite. Four windows were considered: 01 Oct. 1999 to 18 Nov. 1999 (W1), 19 Nov. 1999 to 06 Jan. 2000 (W2), 07 Jan. 2000 to 24 Feb. 2000 (W3), and 25 Feb. 2000 to 13 Apr. 2000 (W4). In each window the entire local time cycle is sampled evenly. A comparison between the LIS data with the network data gives an opportunity to estimate the intracloud lightning activity in the North and Southeast regions. In this study, the detection efficiency of the LIS sensor was assumed to be 90% in both regions (R. Blakeslee, private communication). We assume that a possible geographic variability in the LIS detection efficiency in both regions (if any) is small, so that we can neglect its effect on the results. Such variability might occur in association with a difference in the storm optical depth in the two regions [Boccippio et al., 2001]. Nevertheless, there is no reason to believe that a significant difference exists. We also assume that any differences from the intracloud and CG LIS detection efficiency can be neglected [Goodman et al., 1988]. Recent results based on a limited case study [Thomas et al., 2000], however, have indicated that the CG detection efficiency may be lower than the intracloud efficiency. The influence of a possible difference in the intracloud and CG detection efficiency on our results will be discussed later.