Range resolution of a lidar system is determined by the laser pulse length and the electronic bandwidth of detectors and the relevant circuitry. Long-pulse lidars such as differential absorption lidars and Doppler lidars based on $CO_2$ laser have range resolution of from hundreds of meters up to a few kilometers. To resolve small-scale inhomogeneity or to analyze the near-range signal with long-pulse lidars, the effects of laser pulse length should be eliminated. Some hardware techniques for improving range resolution have been reported, including the pulse clipping technique and the pulse compression technique, but they necessarily cause the transmitted pulse energy to be reduced or make the lidar system complicated. These facts have led to the development of software approaches to remove the convolution effect of the long laser pulse on the lidar return signals. In this thesis, the effect of the laser pulse length on the lidar signal is investigatd and three different deconvolution methods are introduced to derive a highly range-resolved signal from a long pulse lidar signal. Their performance together with the possible errors are also investigated numerically by computer simulation. One of three methods is a deconvolution technique utilizing matrix formulation of deconvolution equation, in which the deconvolution problem becomes a task to find a inverse matrix. While this algorithm is simple and straightforward, it is sensitive to the noise in the original lidar data and it will be successful only if the numerical values of inverse matrix elements are well-bounded. Another method is so called iterative error-reduction method. In this method, the lidar signal itself with certain temporal shift is set to be the start profile for the unknown maximally resolved profile in the proposed technique, and then is corrected in proportion to the difference between the lidar return calculated with the assumption and the real one. The same process is repeated until the correction is smaller than tolerance. Using this technique, the induced fast noise component can be controlled. This technique is good for a laser pulse with a single peak and a long tail, but fails for more complex laser pulse having two peaks or more, for example. The third technique, called iterative Fourier deconvolution technique, is a combination of Fourier deconvolution technique and iterative error-reduction technqiue. The advantage of this technique is to be able to adopt well-established FFT algorithms and to filter out the high frequency noise. Therefore, the last will probably be one of the most promising techniques for deconvolution of long pulse lidar signals. This technique will be a useful software tool for data correction as well as improvement of range resolution if it is adopted as a preliminary step in lidar data processing.