Because of the serious soil erosion, the soil moisture and nutrients deficiency, the slowly recovery and scarcity of natural vegetation in the hill-gully Loess Plateau, the ecological effects of litter layer are particularly important for litter layer effective reduction on surface runoff and mitigate soil erosion. Thus, in this study, the litter accumulation of communities and main herb species in different site conditions was investigated in the hill-gully Loess Plateau, water holding capacity of litter of main herb species was measured by the immersion method, potential nutrient return capacity of litter of main herb species was analyzed by measuring the nutrients content, lignin and cellulose content, then the effects of main species litter on soil fertility improvement was discussed. It's important to understand the mechanism and efficacy of the resistance of plants to soil erosion. The main results were as follows: 1) the accumulation of community litter was between 73.74-175.26 g/m², and was significantly different in slope-gully sites (P<0.05), ordered as shady gully slope > hilltop > sunny gully slope > shady hilly slope > sunny hilly slope. While in micro-sites ordered as under vegetation > ephemeral gully > fish-scale pit > bare land on the sunny slope, ephemeral gully > fish-scale pit > under vegetation > bare land on the shady slope and under vegetation > bare land on the hilltop. 2) The maximal water holding capacity of litter of main species was up to 1.22-4.34 times of its dry mass, and it was extremely significantly different among species (P<0.01), ordered as the defoliation of Bothriochloa ischaemun > the defoliation of Artemisia gmelinii > the deadwood of Bothriochloa ischaemun > the defoliation of Lespedeza davurica > the litter of Stipa bungeanaon > the deadwood of Lespedeza davurica > the deadwood of Artemisia gmelinii. 3) The organic carbon content (7.35%-40.33%) and total nitrogen content (0.61%-1.60%) of species were extremely significant different among species (P<0.01), but they were not significantly different in slope-gully sites. 4) Indicators that influenced litter decomposition (such as lignin content, cellulose content, lignin/N ratio, C/N ratio) were extremely significant different among species (P<0.01). The lignin content (1.00%-8.20%) and cellulose content (3.16%-14.06%) and lignin/N ratio (0.78-12.48) were extremely significantly different in slope-gully sites (P<0.05), while C/N ratio (5.61-57.41) was not significantly different. 5) The potential nutrient return capacity of Artemisia gmelinii defoliation was the highest, followed by the litter of Lespedeza davurica and Bothriochloa ischaemun, and the litter of Stipa bungeanaon was the lowest. In conclusion, the water holding capacity and the potential nutrient return capacity of Bothriochloa ischaemun litter on the sunny slope and the defoliation of Artemisia gmelinii and Lespedeza davurica on the hilltop and shady slope were the highest among species in slope-gully sites, they could effectively improve the soil fertility. While, the litter of Stipa bungeanaon water holding capacity and the potential nutrient return capacity both were the lowest among species in slope-gully sites. The soil fertility effects of different species should be considered when the vegetation was used to control soil erosion. The species should be chosen which not only could control soil erosion but also significantly improve soil fertility, and then the soil quality could be slowly improved. Therefore, Bothriochloa ischaemun, Artemisia gmelinii and Lespedeza davurica were the good choice to control soil erosion on the slopes.