| Popis: |
Surveys of soil fertility and quality provide essential information, at multiple scales, for policy and management decisions on agricultural and environmental questions. Such surveys are expensive, and national-scale agricultural or geochemical surveys are conducted infrequently by public sector organisations. Where resources are scarce, the costs of management, sample collection, analysis, and data management can be prohibitive. This is likely to create a particular barrier to understanding how soil fertility may influence crop production in contexts where small-scale production is prevalent. Private sector soil laboratories conduct soil fertility analyses on thousands of samples annually: samples are often collected in a systematic way by experienced technical staff. The simple step of collecting location coordinates can greatly extend the utility of the data, beyond the immediate value to the farmer and without an added burden on them. Here we use a dataset of soil organic matter (SOM), plant-available phosphorus (P) and exchangeable potassium (K), from c. 27,000 georeferenced sample points collected in 2014-2017 in Pakistan by Fauji Fertilizer Company (FFC). Geostatistical methods are used to model the spatial variation of these soil properties and predict their values at unsampled locations. The geostatistical model allows the probability that soil properties fall below recommended thresholds for wheat production, and these probabilities are displayed as maps using "calibrated phrases" for probability ranges in the legend. The resulting maps provide a comprehensive overview of key agronomic data across ~194,000 km 2 of cropland, realised as a direct consequence of sample location data collection being implemented by FFC. There is spatial structure in the SOM and K data, but not in the P data. These maps can be used to educate agronomists and farmers about typical soil fertility conditions in their region, and where soil analysis is a priority. They can also support more strategic planning of fertiliser needs in Pakistan where widespread nutrient deficiencies are found. This research demonstrates the potential for this approach to be used in many wider geographic contexts where there are few data from one-off survey activities. Additional benefits address concerns about data privacy have thus far limited their integration into a spatially disaggregated output that could benefit wider society. |
