Study Site and Weather Conditions The field experiments were conducted in a continuous rice– rice–rice cropping system for nine consecutive seasons during 2009 to 2012 on farmers’ fields. There are three rice-growing seasons (i.e., Aus, Aman, and Boro) in a year. Aus and Aman are monsoon seasons where rice is mainly grown as rainfed crop, while Boro is a dry season and crop production completely depends on irrigation water. The experiments were spread over 35 upazilas (subdistricts) of eight districts (Barisal, Bagerhat, Borguna, Jhalakati, Madaripur, Patuakhali, Pirojpur, and Shariatpur) in southern Bangladesh. Shariatpur and Madaripur districts fall under agro-ecological zone (AEZ)-12 while the remaining districts fall under AEZ-13 (BBS, 2011). The region comprises an extensive area of tidal floodplain to the South and Southwest of the country. The climate is a humid subtropical monsoon. Average annual rainfall ranges from 1700 mm in the West to 3300 mm in the Southeast, primarily received from June to October. The entire region lies within the cyclone zone. In this region, soils are generally fertile silty clay with low to medium organic matter content, high acidity, high cation exchange capacity (CEC) and K, and low to medium Zn, B, and S (FRG, 2012). Experimental Design and Treatments A total of 115 field trials were established during nine ricegrowing seasons over 4 yr. Out of 115 experiments, 31 experiments were established in 2009 (11 and 20 in Aus and Aman, respectively), 50 in 2010 (23, 16, and 11 in Boro, Aus, and Aman, respectively), 28 in 2011 (11, 7, and 10 in Boro, Aus, and Aman, respectively), and six in 2012 (Boro). The four fertilizer treatments, namely broadcast PU (control), UB, and NPK of sizes 2.4 and 3.4 g, were arranged in a randomized complete block design in each farmer’s field (each farm represent a block). Farmers’ fields were considered as replications. Out of four treatments, three treatments were deep placement of fertilizer briquettes and one was broadcast PU. Nitrogen rates for the Aus and Aman seasons were the same, while rates were higher during the Boro season. During the Boro season, urea briquettes of 2.70 g (78 kg N ha–1, UB N78) were used instead of the 1.8 g (52 kg N ha–1, UB N52) urea briquettes used during the Aus–Aman seasons. Similarly, two NPK of 2.4 g (87 kg N ha–1, NPK N87) were used during the Boro season, while one briquette (43.5 kg N ha–1, NPK N43) was used in the Aus–Aman seasons. Treatment with NPK of 3.4 g (57 kg N ha–1, NPK N57) was the same for all the seasons. Urea briquettes of 1.8 g (Aus and Aman) and 2.7 g (Boro) at 62,500 placement sites per ha (40 by 40 cm spacing) were equivalent to 52 and 78 kg N ha–1, respectively. The NPK of 2.4 and 3.4 g were prepared using PU, diammonium phosphate (DAP), and muriate of potash (MOP). The P and K rates were lower in NPK compared to broadcast application. The equivalent N rates for 2.4 and 3.4 g NPK were 43.5 and 57 kg N ha–1, respectively. The placement of two 2.4 g NPK resulted in an N rate of 87 kg N ha–1. In addition to field experiments, two fertilizer treatments (broadcast PU and deep placement of UB) were tested in 440 demonstration plots across eight districts during 2009–2010 (Table 2) to compare the grain yield between UDP and farmers’ broadcast PU with farmers’ management practice. Plot size for each treatment was 200 m2. In each demonstration field, deep placement of urea briquettes and other cultural practices were similar with field experiments, while broadcast urea was applied in two to three splits (early tillering, maximum tillering, and panicle initiation stages) following farmers’ management practice. Crop Management Rice seedlings (two to three per hill) of 20 to 30 (Aus), 25 to 35 (Aman), and 35 to 45 (Boro) d old were transplanted at 20 by 20 cm spacing. Prilled urea was applied as broadcast in three equal splits at basal, 25 to 30 d after transplanting (DAT) and 5 to 7 d before panicle initiation. Urea and NPK (one to two per placement site) were applied 5 to 7 d after transplanting. The briquettes were deep placed at a 7- to 10-cm soil depth between four alternate hills of rice. Phosphorus (triple superphosphate) and K (muriate of potash, MOP) fertilizers were applied basally in all the plots (except plots receiving NPK) during final land preparation at 14 to 19 kg P–K ha–1 during the Aus–Aman seasons and at 20–36 P–K ha–1 during the Boro season. All the other management practices were similar for all the treatments. In farmers’ fields during Aus and Aman seasons, PU was applied in two splits, 60% of total N at early tillering and 40% at late tillering stages. During Boro season it was applied in three splits, 40% at 10–15 DAT, 30% each during maximum tillering stage and 5 to 7 d before panicle initiation stage. The average (± SD) N rates for Aus, Aman and Boro seasons were 95 ± 14, 92 ± 9 and 124 ± 9 kg N ha–1, respectively. Rice Yield and Yield Parameter Effective tillers (panicle) and grain yield (14% moisture content) from experimental plots were recorded at maturity from 1 and 5 m2 area, respectively. Grain yield in each demonstration plot was recorded from two subsamples (10 m2 each). Moreover, to evaluate the response of deep placement of UB compared to broadcast PU under farmers’ management practices, crop cuts were done (as in demonstration plots) from 1395 farmers’ fields across eight districts where farmers cultivated rice following both broadcast PU and deep placement of UB. The harvest area was same for all the growing seasons and years. Partial factor productivity for nitrogen (PFPN) or an index of total economic outputs relative to the use of all N sources (indigenous soil N and applied fertilizer N) was calculated as kg grain yield per kg N applied (Olk et al., 1998; Ladha et al., 2005). Partial factor productivity for N (PFPN) = kg grain/kg applied N Grain yield in each treatment was due to applied N plus indigenous N. Similarly, the average N rate in the respective district was used while calculating PFPN in farmers’ field crop cuts. Yield increment and nutrient savings due to FDP over broadcast PU were estimated. Net benefit in US$ per ha–1 was calculated based on partial budget analysis