In a red clover breeding program one of the goals is to create cultivars that are resistant to abiotic stress and thus indirectly affect the increase in yield of green mass and the adaptation of cultivars. Selection of red clover was performed at the Faculty of Agriculture, University of Zagreb, under low temperature or drought conditions and conditions without abiotic stress. In this paper, populations developed under stress conditions were compared to populations developed under stress-free conditions and control cultivar. The hypothesis was that populations of red clover developed in abiotic stress conditions differ in morphological and agronomic properties from populations developed in stress-free conditions and that there has been a change in the allelic frequency due to a one selection cycle in a low temperature conditions. In addition to agro-morphological traits evaluated through field experiments, DNA-based genetic variability was analyzed using AFLP markers in all populations from field experiments, and SSR markers were used to investigate the change of the allelic frequency from two original populations (cultivars) to populations (reselection) after a single selection cycle under low temperature conditions. Plant material for three-year field trials included four breeding populations originating from Canada (K1, K2, K3 and K4), two populations selected under dry conditions (Lp1, Lp2), two populations selected under low temperature conditions (PS1, PS2), one breeding population reselected from the PS2 population under dry conditions (PS2rr), two Japanese cultivars selected under low temperature conditions ('Sapporo', 'Hokuseki') and standard cultivar ('Viola'). In the field trials, 18 agro-morphological traits and quality traits were measured and observed. Twelve entries were distributed in randomized block design with four repetitions. Two cuts were performed during the first, three cuts during the second and two cuts during the third experimental year. Analysis of variance (ANOVA) revealed significant differences between breeding populations versus cultivars and populations developed under stress-free conditions in most agro-morphological traits and fodder quality traits. Populations were significantly different in stem thickness, central leaflet width, number of days to flowering, stem hair density and direction, central leaflet mark shape and intensity. Also, the populations differed in structural carbohydrates content and fodder relative feed value in both years of measurement of these traits. Stress induced increase in the content of water-soluble carbohydrates was confirmed during the extremely dry second year of growing, and the relationship between the content of water-soluble carbohydrates and the resistance to abiotic stress was also confirmed. After two years of growing, breeding populations Lp1, PS1 and PS2 had highest yields of green matter and air-dry matter and after three years highest yields had breeding populations Lp1, PS1 and PS2. Strong positive correlations has been found in the yield of green and air dry matter of the between first and second year of growing of between the second and third of growing. Also, the yield of green matter was positively correlated with the thickness of the stem, central leaflet width and the content of acid detergent fibers. These correlations indicate that that the most productive forms can be selected even in the first year of vegetation, and also that indirect improvement of several traits could be achieved through selection of one trait. All populations from field trials were used in the AFLP marker analysis. Four combinations of AFLP primers amplified 503 markers, of which 83% were polymorphic. The average level of genetic diversity by Nei for all populations was 0,236. Lowest level of genetic diversity was recorded in control cultivar ‘Viola’. Cluster analysis separated the populations into two groups. In one group there are both local populations, control cultivar ‘Viola’ and three breeding populations originating from Canada. The second group consists of other breeding populations and cultivars, with the exception of the K1 population, all selected in low temperature conditions. Molecular variance analysis (AMOVA) showed a high level of variability that was consisted of between population variability (7,73 %), and within population variability (92,27 %). Genetic differentiation index showed moderate differentiation between populations, ranging from 0,030 to 0,152. Results of AFLP analysis are consistent with work of other authors with this molecular markers on red clover. SSR markers analysis included cultivar 'Reichersberger' (P1) with its reselection (PS1) and cultivar 'Croatia' (P3) with its reselection (PS3). Both reselections were created under stressful low temperature conditions in a single selection cycle. Analysis revealed that the frequency of alleles from the original populations (P1 and P3) to the populations after a single selection cycle (PS1 and PS3) was increased. There was also a change in allele distribution after selection. The fraction of moderate and high frequency alleles was increased in populations after selection, while the fraction of low frequency alleles was decreased. Waples test revealed 12 non-neutral loci whose frequency was altered by selection in populations P1 → PS1 and nine in populations P3 → PS3. Calculation of expected heterozygosity and Shannon index revealed a reduction in genetic variability after selection. Based on the results of analysis of agromorphological traits, and due to its good yield and quality, four breeding populations Lp1, Lp2, PS1 and PS2, developed in the abiotic stress conditions could be marked as promising material for future breeding programs. AFLP markers had proved to be a good tool for measuring genetic variability and population differentiation in red clover. SSR markers analysis showed reduced genetic variability through decreased heterozygosity in populations after strong selection. This decreased heterozygosity, due to the plasticity of the diploid red clover, was still high enough not to cause bottlenecks in breeding populations for the future selections.