Daily water temperatures in the fish tanks were monitored by means of a maximum-minimum digital thermometer

The operation of both MAS was identical until the water temperature started dropping below 20 ◦C. At that moment,two different strategies were employed. In one of the MAS,the water was heated by means of a selfconstructed thermo-solar panel coupled with the system, in order to avoid water temperatures below 13 ◦C. If temperatures outside the greenhouse were close to zero degrees, two submersible heaters were placed in the fish tank, which were in operation for 8− 10 hours at night. The second MAS stopped working as an aquaponic, given that the fishes were removed to avoid heating the water. Therefore, it was transformed into a hydroponic installation in which only the grow bed tank, the NFT and the sump were functioning, with a total volume of 180 L. In order to provide the nutrients required for the hydroponic production of the plants, a bio-fertiliser obtained from a small vermicompost facility installed next to the greenhouse was employed. For the monitoring of the production of the different crops, information about the exact date of harvest, number of vegetables or fruits obtained and their weight was recorded. The fish biomass was also monitored by weighing the fish in each tank. An initial weighing was performed to obtain the baseline and then ten more measurements took place, distributed during the year. The fish were also weighed when they were harvested. Calcium hydroxide was used to increase the pH when it dropped below 6. Venturi devices and air compressors helped to maintain optimal oxygen levels. Simple and inexpensive water test kits that employ colorimetric methods were used for the measurement of the water chemical parameters. 

Three complete water analyses were performed in both MAS during the experiment. To do so, water samples were taken from the sump and analysed by spectrophotometry. When nitrate concentrations in the water exceeded 80 ppm, flower pots for sale new plants were put into the system. Foliar applications of K2SO4 and MnSO4 were performed when required. An EDDHA Sequestrene 138 Fe chelated iron solution  was directly added to the water. Seaweed fertiliser  at 10 mL L− 1 was used only for 3 weeks, sprayed on leaves, at the beginning of the operation of the aquaponic facilities, between April and May, because the fish biomass being still low, the nutrient level for plants in water was also low. FERRAMOL  at 5 g m-2 was used against the snails and slugs that appeared during the autumn. Nettle slurry  at 15 mL L− 1 was used as a foliar application against the white fly that affected lettuces in autumn. Other routines periodically performed were: observations of the health status of fish and plants, the control of the water levels  and flows,and maintenance and cleaning of equipment. Potassium soap  at 20 mL L− 1 mixed with Neem oil  at 1 g L-1 was used against aphids  from July to November. Slaked lime  was added in the sumps to raise the pH from 6–6.2 to 6.4–6.5. Micronised sulfur  was applied foliarly at 0.3 g L-1 against fungal pathogens  and occasionally against red spiders that appeared in a very small quantity in strawberry plants.The total horticultural production, number of plants and fruits, average production per plant and average weight of fruits or leaves in the different hydroponic sub-systems are shown in Tables 2 and 3. Both MAS had a similar annual total production per cropping area: 38.96 and 38.31 kg m− 2 for MAS1 and MAS2, respectively.

NFT was the most productive hydroponic sub-system, mainly due to the lettuce production. However, GB was the sub-system with the highest production per cropping area,followed by NFT  and DWC. The NFT subsystem had a higher density of plant species, with 20 plants m− 2. The plant density in the GB sub-system was 11.7 plants m− 2,the values being lower in the DWC. Cucumber and broccoli were the only plants that were cultivated in the three sub-systems, the GB being the one with the highest production for both species. The most productive species were courgette, cucumber and Roma tomato in the GB; lettuce, chard, Italian frying pepper and Roma tomato in the NFT; and pumpkin in the DWC. The plants that took, on average, the least time to be ready for harvest in MAS 1 were stevia,courgette  and cucumber. For MAS2, they were stevia,lettuce  and basil. On the other hand, the plants that took the longest to be produced in MAS1 were strawberry,water melon  and goat horn pepper. For MAS2, they were pumpkin,water melon  and goat horn pepper. The distribution of the production during the year  for all the species cultivated is also detailed in the Supplementary Information. Lettuce was the only plant species grown year-round in the GB and NFT subsystems, and therefore had the highest annual production with 68.7  and 64.5 kg. Also, cucumber and courgette had a high production. On the contrary, onion and cabbage were the species which contributed less to total production. The presence of aphids slightly affected the chards and cucumber; however, they were able to grow and give satisfactory production. The strawberry had a low production due to late planting, outside its natural season. The large-fruited and typical summer plants in Andalusia such as water melon, pumpkin and melon produced very few fruits in September and October. The water melon and melon plants were also very affected by the attack of the aphids, although they did bear fruit. The months with more production were April, July and August, accounting for 46 % of the total in the whole year, May and February being those with a lower production. In the case of May, this was because it was the system’s first month of operation, when the fish biomass was still low.

The total tilapia production was 33.5 kg in MAS1 and 29.28 kg in MAS2. The fish harvests occurred in August,September, October and April. The evolution of the fish biomass is shown in Fig. 5. The largest fish harvest was carried out in the fifth weighing: 43 fishes were extracted from MAS1 and 41 fishes from MAS2 as high fish biomass,plus high temperatures reduced dissolved oxygen levels in water to critical levels. In the seventh weighing, 16 fish were extracted from MAS1 and in MAS2 all the remaining fish  were harvested, so it became a completely hydroponic system. The mean weight per harvested fish was 317 g for MAS1 and 302 g for MAS2. Among the 31 fish finally harvested in MAS1 at the end of April 2019 after 1 year of growth, 14 of them weighed below 350 g. There was no predation of small fish; however, a total of 5 fishes died in MAS1 and 9 fishes in MAS2. Table 4 shows the average values for the growth indicators in each MAS. Fig. 6 shows the evolution over time of the plants harvest, fish biomass, nitrates and amount of fish feed.One of the main challenges of aquaponic production is maintaining an adequate water quality so the plants, fish and microorganisms can function correctly. Therefore, the key to a successful aquaponic system, especially if it is a micro-scale one, is to be able to keep the physicochemical water parameters within the limits that those three populations need. This is not easy in coupled systems, because the quality of the water circulating in the hydroponic subsystem is the same as the aquaculture subsystem. However, opting for decoupled systems on a small scale is not feasible due to their complex maintenance  which is not suitable for a single family but on a large scale such as a commercial environment. For instance, the pH is one of the most difficult factors to control as the optimal values for fish and plant production are different. During our experiment, the pH was maintained with values of 6.45 ± 0.47  and 6.62 ± 0.51,given that with values below 6, bacterial nitrification is considerably reduced. Hence, when the pH decreased, calcium hydroxide had to be added. At the time of transforming MAS2 from aquaponic to hydroponic,the pH, NO3 − and EC levels fluctuated rapidly. The pH quickly rose from 6.8-7 to 7.5 in the first hours, so hydrochloric acid was added to lower it to the previous levels. In aquaponic systems, tower garden the balance between the fish biomass and the plant production is very important in order to both maximise the production and to maintain an adequate quality of the water. Also, plant nutrients uptake depends, among other factors, on the pH.

If it is not optimal, the nutrients can be accumulated in the water. This is a drawback in coupled aquaponic systems. In decoupled systems, pH values can be better adapted without affecting fish and bacteria. Not only pH levels but also dissolved oxygen is important for the nutrients balance. Wongkiew et al.  stated that low levels of dissolved oxygen can lead to nitrogen loss mainly via nitrifier denitrification. They considered values of 3.8 mg L− 1 as low levels, while those around 7 mg L− 1 where high. In our experiment, dissolved oxygen ranged between 5 and 11 mg L− 1,being below 5 mg L− 1 only in August. The relation between the plants harvest, fish biomass, nitrates and amount of fish feed is clear. For instance, in MAS1, the maximum nitrate levels were 155 mg L− 1,corresponding with a peak in fish feed  and biomass. Indicating that the rate of nitrogen input to the system  was higher than the rate of assimilation by plants. The same was reported by  in aquaponic systems based on lettuce, pak choi and chives. This suggests that a greater surface could be dedicated to plant cultivation during summer. As an alternative, more fruiting vegetables could be introduced, these having a higher demand for nutrients than green leafy vegetables and also producing more kg at harvest. This is important because it would mean an increase in the annual plant production. In contrast, since the beginning of December 2018, the nitrate levels dropped, though 50 mg L− 1 were maintained with a low feed consumption  until the end of February 2019, when a minimum of 21 mg L− 1 was reached with a fish biomass of 8.34 kg and 24.9 g of feed. The decrease of nitrates was not so fast because of the expanded clay used in the GB sub-system, which could act as a “nutrient sponge” where a significant amount of solids accumulated and continued to be mineralised in well-oxygenated areas due to the emptying and filling effect of the bell-syphon.The total production during the study period accounting both for tilapia and plant production was nearly 211 kg in MAS1 and 204 kg in MAS2, employing a maximum cropping area of 4.56m2 and a fish tank volume of 0.95 m3,occupying a surface of 1.2 m2. Therefore, though the vegetal production per cropping area was slightly above 38 kg m− 2,the yields per production unit area  were around 36 kg m− 2. Still, in order to compare the yields per unit area obtained with other methods of production, the entire surface required for the production should be considered. Including service corridors and the area devoted to the clarifier and other equipment, the surface to be considered for each MAS was 16 m2. In our study the total production  was 415 kg of produce in a 45 m2 greenhouse,which means 9.2 kg m− 2. This value more than doubles that of the average yield per unit area in the social urban vegetable gardens in Seville. Sommerville et al.  estimated that an FAO-type MAS could produce 360 lettuce heads, 54 kg of tomatoes and 30 kg of fish during a year with a cropping area of 3 m2. Although these authors do not specify the average weight per lettuce head, if we estimate 250 g per head, the total production would be 90 kg of lettuces that, added to the tomatoes and fish, would result in 174 kg year− 1,slightly lower than the amount obtained in our study, though the fish production was very similar. The fact that our work involved 22 different species of vegetables, herbs and fruits makes it difficult to establish comparisons in terms of total plant production with other research works related with aquaponics in which only one crop was produced.