Natural temperature gradient (NTG) can be a significant problem in thermal sap flow measurements, particularly in dry environments with sparse vegetation. To resolve this problem, we propose a novel correction method called cyclic heat dissipation (CHD) in its thermal dissipation probe (TDP) application. The CHD method is based on cyclic, switching ON/OFF power schema measurements and a three-exponential model, extrapolating measured signal to steady state thermal equilibrium. The extrapolated signal OFF represents NTG, whereas the extrapolated signal ON represents standard TDP signal, biased by NTG. Therefore, subtraction of the OFF signal from the ON signal allows defining the unbiased TDP signal, finally processed according to standard Granier calibration. The in vivo Kalahari measurements were carried out in three steps on four different tree species, first as NTG, then as standard TDP and finally in CHD mode, each step for ∼1-2 days. Afterwards, each tree was separated from its stem following modified Roberts' (1977) procedure, and CHD verification was applied. The typical NTG varying from ∼0.5 °C during night-time to -1 °C during day-time, after CHD correction, resulted in significant reduction of sap flux densities (J(p)) as compared with the standard TDP, particularly distinct for low J(p). The verification of the CHD method indicated ∼20% agreement with the reference method, largely dependent on the sapwood area estimate. The proposed CHD method offers the following advantages: (i) in contrast to any other NTG correction method, it removes NTG bias from the measured signal by using in situ, extrapolated to thermal equilibrium signal; (ii) it does not need any specific calibration making use of the standard Granier calibration; (iii) it provides a physical background to the proposed NTG correction; (iv) it allows for power savings; (v) it is not tied to TDP, and so can be adapted to other thermal methods. In its current state, the CHD data processing is not yet fully automated.