在第一階段從桃園、南投、彰化等地區的溫室採集二十四個土壤樣品分析土壤化學特性。結果顯示EC值並不是主要限制植物生長因子，主要原因為養分不平衡，特別是土壤中累積高含量的有效性NO3-1-N、 Ca、 Mg、 K或 Na。而鹽分對EC的相關性順序為NO3-N (R2=0.8974) > Na (R2=0.8906) > Mg (R2=0.6505) > Ca (R2=0.6035) > K (R2=0.1811) > NH4-N (R2=0.0081)，且對於前四種離子顯著差異P值均<0.001。
在第二階段，是有無接種根瘤菌的複因子試驗(由生長於沿海的田菁篩選出)，及添加氯化鈉調整不同鹽度(2、4、6、8、及12 dS m-1)。田菁盆栽試驗是在溫室進行有四重複，並且在種植後的四十天由田菁根部的根瘤篩出根瘤菌。結果顯示接種外源根瘤菌減少根瘤生成，特別是EC值高於6 dS m-1。然而接種處理會增加地上部的氯離子濃度，但只有在低鹽度處理下增加硫酸根的濃度。種植田菁的確能大量降低土壤EC值，特別是減少對土壤鈉與鉀的濃度。在二十五天前增加鹽度的處理會顯著降低植株高度，但在四十天時則沒有統計上的顯著差異是由於所有的處理EC值均小於2 dS m-1，甚至EC12的處理。
在第三階段，兩種不同驗度處理(BK未添加及EC12)及有無接種由第二階段EC2及EC12處理所篩選出的原土根瘤菌菌A (Rhizobium sp. R8-557) 和菌B (Rhizobium sp. R55)。在此階段有五處理分別為BK、 BK+A、BK+B、EC12+A、及EC12+B。結果顯示除了EC12+B的處理之外，有接原土根瘤菌增加50%以上的根瘤數，且在株高及乾重也反應出顯著增加。然而，第三十天在EC12+B的土壤中Ca、Mg、Na、Cl-1及 SO4-2濃度比EC12+A處理顯著較低。在耐EC試驗中顯示菌B比菌A有較高的耐鹽性，此結果指出接種高耐鹽性的根瘤菌不能增加田菁生長但有效增加鹽害移除。
在第四階段為在種植花溫室的田間試驗包含BK和BK+A的處理。結果顯示土壤EC大量降低，但是BK+A的處理降低EC程度較小，可能是由於接種菌A是由EC2 dS m-1篩出並不適合生存於原田間土壤EC4 dS m-1。
The area of salt-affected soil has been increased worldwide and concerning more on housing agriculture activities. An effective and low input method on the remediation of salt-affected soil is expecting, especially demanded for the housing agriculture. Sesbania has been known as a good tropic green manure for its fast growth, it also has high potential in salt uptake. This thesis is divided into four sections to study the efficiency of sesbania (sesbania cannabia (Retz.) Pers.) on desalination of salt-affected soils with or without inoculation of rhizobia.
In the first phase 24 soil samples were collected from Taoyuan, Nanto, and Changhua counties’s greenhouses to analyze soil chemical properties. Results showed that electric conductivity (EC) is not the main cause on plant growth limiting factor, the main causes are the nutrient unbalances, especially on high accumulation of available NO3-1-N, Ca, Mg, K, or Na in soils. The correlation order of salts on EC is NO3-N (R2=0.8974) > Na (R2=0.8906) > Mg (R2=0.6505) > Ca (R2=0.6035) > K (R2=0.1811) > NH4-N (R2=0.0081)，and the significance p values are all <0.001 for the first four ions.
In the second phase, a factorial experiment with rhizobia (selected from sesbania grown in sea coast inoculation or not) and greenhouse spiking NaCl into different EC levels (2, 4, 6, 8, and 12 dS m-1). The sesbania pot experiment was conducted under net house with 4 replications. The rhizobia were also screened from the sesbania root nodules at 40 days after seeding. Results showed that the inoculation of nonnative rhizobia reduced the formation of root nodules, especially as the EC higher than 6 dS m-1. However, the inoculation increased the Cl-1 concentration of shoot, but increase SO4-2 concentration only at low EC treatments. The plantation of sesbania really dramatically reduced soil EC, especially reducing on soil Na and K concentrations. Increasing soil EC significantly reduced sesbania plant height before 25 days after seeding (DAS), but no statistical difference at 40 DAS due to the soil EC all reduced to lower than 2 dS m-1, even the EC12 treatment.
In the third phase, two soil EC levels (BK with no spiking and EC12) inoculation with rhizobia screened from the second phase as inoculants A (Rhizobium sp. R8-557) and B (Rhizobium sp. R55) from EC2 and EC12 treatments with no inoculation of native rhizobium. In this phase the five treatment are BK, BK+A, BK+B, EC12+A, and EC12+B. The results showed that the inoculation of these two native inoculants increased the root nodules more than 50 % except the EC12+B treatment. Accordingly, that responded to the plant height and dry weight of sesbania significantly increased except EC12+B treatment. However, the concentrations of soil Ca, Mg, Na, Cl-1, and SO4-2 in EC12+B treatment were significantly lower than those in EC12+A treatment at 30DAS. In EC tolerance experiment showed that B inoculant is more tolerant than A inoculant to high EC. This result indicates that the inoculation of high EC tolerant rhizobium may not increase the sesbania growth, but readily increase it salt removal.
The fourth experiment only BK and BK+A treatments was conducted in a flower growing green house. Results showed that the soil EC was dramatically reduced, and the reduced level of BK+A was lower. This may be due to the inoculants screened form EC2 soil will be not competitive in the soil with original EC at 4 dS m-1.
The conclusion could be droved that the sesbania is a good phytoremediation plant for salt-affected soils, and the inoculation of rhizobium should be from the native induced ones, otherwise negative effect may be found.