lowlands in Thailand.
The relations between ImageJ-estimated and WinRHIZO-estimated root length using Mean and Triangle algorithms are depicted in Fig. 2. The ImageJ-estimated root length using the Tri-angle algorithm has a very good fit to WinRHIZO-estimated root length (Fig. 2a). However, using the Mean algorithm, the ImageJ-estimatedrootlengthcorrespondstoWinRHIZOestimatesforlarge root lengths per image but not for small ones (Fig. 2b).
Using 2/3 times the root length with ImageJ for the Triangle thresholding algorithm for all samples, both root length and spe-cific root length (i.e., root length/root weight) estimated using this protocol were closely matched with data estimated using WinRHIZO (R2 =0.990, 0.982, respectively, Fig. 3); the regression slopes were almost unity (the regression lines were, respectively, y=1.02x−2.23 and y=1.08x−41.9; Fig. 3). Similar results were obtained also for estimation of the root length of rice grown under various hydrological regimes in the pots (Kato, unpublished data). Fine root development and consequent increased specific root length might be an adaptive growth characteristic in drought-affected rice fields (Suralta and Yamauchi, 2008). Importantly, the