Biodiversity data

Filling biodiversity knowledge gaps requires prioritisation of efforts not only to compile addi-tional data but also to evaluate and enhance the quality of the data already available and to make it accessible. Works from Ballesteros-Mejia et al. (2013); Marques et al.(2018);Stropp et al.

(2016) as well as the work in Chapter2are recent examples for African countries.

Many developing countries are understudied and present a severe lack of species-occurrence data (Peterson et al., 2015), which is worsened by the poor dissemination of these research data. Thus, improving knowledge of the biodiversity of poorly documented countries can only be achieved by allocating resources to expand and promote national and international initia-tives, with a strong emphasis on capacity-building of national and local institutions. Positive progress has been made in this direction. For example, Biodiversity Information for Devel-opment (BID) is a multi-year programme funded by the European Union and led by GBIF to increase the amount of biodiversity information available in the nations of sub-Saharan Africa, the Caribbean and the Pacific (https://bid.gbif.org). Of the 23 projects financed thus far, Mozambique is participating in an “African Insect Atlas”, which aims to unleash the poten-tial of insects in conservation and sustainability research (https://www.gbif.org/project/

82632/african-insect-atlas).

3.5 Conclusion

It is most important to fill knowledge gaps on species occurrence and distribution, especially if the aim is to expand the taxonomic extent of conservation planning. A conservation plan-ning based on accurate species occurrence data is even more crucial in countries where high poverty rates, sporadic armed conflicts, intensive exploration of natural resources and extreme weather events accrue. Deprived of reasonable information regarding species occurrence, it is unmanageable to concentrate efforts to preserve diversity and guide conservation actions.

Based on primary species occurrence data, which span the years from 1845 to today, we identified provinces in Mozambique that are poorly documented regarding terrestrial mammal fauna (e.g., Niassa, Cabo Delgado, Nampula and Tete). These provinces are vastly explored for oil, coal, hydrocarbons and minerals (Guedes et al., 2018), presenting severe challenges for biodiversity conservation. Moreover, the high population growth observed in the northern provinces is associated with agricultural development and habitat degradation (Niquisse et al., 2017;Timberlake,2011). Given that habitat loss is a leading cause of biodiversity decline, there is an urgency to study and survey the provinces identified in this study since some economic activities, such as mine-exploration and plantation forestry, without proper impact studies may lead to irreversible biodiversity loss (Ceballos et al., 2017; Chaudhary et al., 2016). Hence, we encourage action from the scientific community and government authorities to continue improving the country’s biodiversity knowledge.

Finally, the assessment of the knowledge gaps from primary species occurrence data showed to be a powerful strategy to generate information that is essential to species conservation and management plan, particularly for understudied countries.

3.6 Supplementary figures

Figure 3.S1: Visualization of the number of unique records across Mozambique based on grids of dif-fering spatial resolutions: a) 0.1º; b) 0.5º; c) 1º.

Figure 3.S2: Relationship between the number of unique records (i.e. unique combination of date, loca-tion of collecloca-tion and species name) per grid cell and (a) number of species, and between the number of unique records and the estimates of inventory completeness (“Comp.”) ob-tained according to according to the three methods tested in this study: (b)Sousa-Baena et al.(2014), (c)Chao and Jost(2012) and (d) the curvilinearity of species accumulation curves according toYang et al.(2013)

Figure 3.S3: Visualization of the records’ density patterns estimated based on the isotropic Gaussian kernel of the whole inventory, and per mammal group.

Figure 3.S4: Bias estimates to “distance to protected areas” for A) the whole inventory, and B) “mammal size”. Bias estimates were calculated following Kadmon et al (2004) for each distance interval from “interval 1” for short distance to “interval 4” for largest distance. The dashed lines mark the range of values where no bias is expected (between -1.64 and 1.64). If the boxplots are within the lines than the number of localities is as expected from a random sampling scheme. Boxplots above and below this area are an over or under-representation of records’ localities in that interval, respectively.

Figure 3.S5: Bias estimates to “distance to main cities” for A) the whole inventory, and B) the mammal groups. Bias estimates were calculated following Kadmon et al (2004) for each distance interval from “interval 1” for short distance to “interval 4” for largest distance. The dashed lines mark the range of values where no bias is expected (between -1.64 and 1.64). If the boxplots are within the lines than the number of localities is as expected from a random sampling scheme. Boxplots above and below this area are an over- or under-representation of records’ localities in that interval, respectively. The cities included in this factor are the provinces capitals.

Figure 3.S6: Bias estimates to “distance to main primary roads” for A) the whole inventory, and B) the mammal groups. Bias estimates were calculated following Kadmon et al (2004) for each distance interval from “interval 1” for short distance to “interval 4” for largest distance.

The dashed lines mark the range of values where no bias is expected (between -1.64 and 1.64). If the boxplots are within the lines than the number of localities is as expected from a random sampling scheme. Boxplots above and below this area are an over- or under-representation of records’ localities in that interval, respectively.

Figure 3.S7: Spatial visualisation of the (A) distance in the bioclimatic space between the country’s cells and (B) geographical distance from the well-known cells regarding terrestrial mam-mal sampling in Mozambique, at 0.25º resolution. The country’s bioclimatic space was defined by the following variables: Mean temperature of the wettest quarter, Temperature seasonality, and Precipitation of the driest quarter. These variables were obtained from the WorldClim database (Fick and Hijmans, 2017). Cells that fit the criteria of well-known grid cells are marked with a dark point in (A) and with a cross in (B).

Figure 3.S8: Sensitivity analysis for different assignment methods of ecoregions to grid cells: (a) for each ecoregion with cover in Mozambique; and (b) for each ecoregion with cover in Mozambique, with the less extensive ecoregions aggregated in biomes. Sensitivity was measured as the ratio between the proportion of cells within a specific ecoregion using an-other assignment methods and the assignment method used in our study. For this study, we assign the terrestrial ecoregions (and associated biome) by overlaying ecoregions map onto the country’s grid, and the ecoregion that overlaps each cell centroid is assigned to the cells (“Cell centroid method”). Two assignment methods were tested and compared to “Cell centroid rule”: “Maximum area rule” and “Majority rule”. In the “Maximum area rule” the largest ecoregion is assigned to the cells; and in the “Majority rule” the ecoregion that over-laps by at least 50 per cent is assigned to the cell, or, when multiple ecoregions overlap a cell, the largest overlapping area must be greater than the area in the cell that is not covered by any ecoregion. Values of sensitivity close to 1 reveal a result that is robust independently of the method chosen for assigning cells to ecoregions. The ecoregions and biomes con-sidered for Mozambique in this study followed a last comprehensive assessment of African terrestrial biomes, ecoregions and habitats (Burgess et al.,2004). Spatial data was down-loaded from WWF Terrestrials Ecoregions of the World dataset (www.worldwildlife.

org/publications/terrestrial-ecoregions-of-the-world).

Figure 3.S9: Barplot showing the number of country’s cells and the number of gap cells at 0.25º resolu-tion occupied by each ecoregion with different polygon-cell assignment rules.

4